|
ThAT1 Regular Session, Auditorium |
|
Planning, Scheduling and Coordination 4 |
Chair: Sláma, Jakub | Czech Technical University in Prague |
Co-Chair: Nekovar, Frantisek | Czech Technical University in Prague |
|
10:00-10:20, Paper ThAT1.1 | |
>Multi-Tour Set Traveling Salesman Problem in Planning Power Transmission Line Inspection |
|
Nekovar, Frantisek | Czech Technical University in Prague |
Faigl, Jan | Czech Technical University in Prague |
Saska, Martin | Czech Technical University in Prague |
Keywords: Planning, Scheduling and Coordination
Abstract: This paper concerns optimal power transmission line inspection formulated as a proposed generalization of the traveling salesman problem for a multi-route one-depot scenario. The problem is formulated for an inspection vehicle with a limited travel budget. Therefore, the solution can be composed of multiple runs to provide full coverage of the given power lines. Besides, the solution indicates how many vehicles can perform the inspection in a single run. The optimal solution of the problem is solved by the proposed Integer Linear Programming (ILP) formulation, which is, however, very computationally demanding. Therefore, the computational requirements are addressed by the combinatorial metaheuristic. The employed greedy randomized adaptive search procedure is significantly less demanding while providing competitive solutions and scales better with the problem size than the ILP-based approach. The proposed formulation and algorithms are demonstrated in a real-world scenario to inspect power line segments at the electrical substation.
|
|
10:20-10:40, Paper ThAT1.2 | |
>Chance-Constrained Motion Planning Using Modeled Distance-To-Collision Functions |
|
Johnson, Jacob | UCSD |
Yip, Michael C. | University of California, San Diego |
Keywords: Planning, Scheduling and Coordination, Probability and Statistical Methods
Abstract: This paper introduces Chance Constrained Gaussian Process-Motion Planning (CCGP-MP), a motion planning algorithm for robotic systems under motion and state estimate uncertainties. The paper's key idea is to capture the variations in the distance-to-collision measurements caused by the uncertainty in state estimation techniques using a Gaussian Process (GP) model. We formulate the planning problem as a chance constraint problem and propose a deterministic constraint that uses the modeled distance function to verify the chance-constraints. We apply Simplicial Homology Global Optimization (SHGO) approach to find the global minimum of the deterministic constraint function along the trajectory and use the minimum value to verify the chance-constraints. Under this formulation, we can show that the optimization function is smooth under certain conditions and that SHGO converges to the global minimum. Therefore, CCGP-MP will always guarantee that all points on a planned trajectory satisfy the given chance-constraints. The experiments in this paper show that CCGP-MP can generate paths that reduce collisions and meet optimality criteria under motion and state uncertainties. The implementation of our robot models and path planning algorithm can be found on GitHub.
|
|
10:40-11:00, Paper ThAT1.3 | |
>Improving the Reliability of Pick-And-Place with Aerial Vehicles through Fault-Tolerant Software and a Custom Magnetic End-Effector |
> Video
|
|
Kobilarov, Marin | Johns Hopkins University |
Garimella, Gowtham | Johns Hopkins University |
Sheckells, Matthew | Johns Hopkins University |
Kim, Soowon | Johns Hopkins University |
Baraban, Gabriel | Johns Hopkins University |
Keywords: Planning, Scheduling and Coordination, Software Architecture for Robotic and Automation, Factory Automation
Abstract: Aerial manipulation is an emerging field in robotics with various potential applications such as transport and delivery, agriculture, and, infrastructure inspection. To deploy aerial vehicles in the real world, the safety and reliability of these systems is paramount. Motivated by the need for safety and reliability, this work proposes a software framework that has built-in robustness to algorithmic failures and hardware faults. The framework allows users to build complex applications while reasoning about faults that can happen at different stages of an aerial manipulation task and specifying fallback actions to return to normal operating mode. The aerial manipulator is further endowed with a magnetic gripper that can handle positional errors arising from perception and control uncertainties. We also introduce a bias estimator for measuring the contact forces and sensor bias. We demonstrate how the estimator can be used to detect either completion or failures across several tasks. We demonstrate the reliability of the proposed framework on two tasks: package sorting task (e.g. as might be used in a distribution center) and sensor placement task (for infrastructure inspection). We show different failure modes that can occur and how our aerial manipulation system recovers from them.
|
|
11:00-11:20, Paper ThAT1.4 | |
>Priority-Based Distributed Coordination for Heterogeneous Multi-Robot Systems with Realistic Assumptions |
> Video
|
|
Cecchi, Michele | Università Di Pisa |
Paiano, Matteo | University of Pisa |
Mannucci, Anna | Örebro University |
Palleschi, Alessandro | University of Pisa |
Pecora, Federico | Örebro University |
Pallottino, Lucia | Università Di Pisa |
Keywords: Planning, Scheduling and Coordination, Logistics, Intelligent Transportation Systems
Abstract: A standing challenge in current intralogistics is to reliably, effectively yet safely coordinate large-scale, heterogeneous multi-robot fleets without posing constraints on the infrastructure or unrealistic assumptions on robots. A centralized approach, proposed by some of the authors in prior work, allows to overcome these limitations with medium-scale fleets (i.e., tens of robots). With the aim of scaling to hundreds of robots, in this paper we explore a de-centralized variant of the same approach. The proposed framework maintains the key features of the original approach, namely, ensuring safety despite uncertainties on robot motions, and generality with respect to robot platforms, motion planners and controllers. We include considerations on liveness and solutions to prevent or recover from deadlocks in specific situations are reported and discussed. We validate the approach empirically with simulated, large, heterogeneous multi-robot fleets (up to 100 robots tested) operating both in benchmark and realistic environments.
|
|
11:20-11:40, Paper ThAT1.5 | |
>Risk-Aware Trajectory Planning in Urban Environments with Safe Emergency Landing Guarantee |
|
Sláma, Jakub | Czech Technical University in Prague |
Váňa, Petr | Czech Technical University in Prague |
Faigl, Jan | Czech Technical University in Prague |
Keywords: Planning, Scheduling and Coordination
Abstract: In-flight aircraft failures are never avoidable entirely, inducing a significant risk to people and properties on the ground in an urban environment. Existing risk-aware trajectory planning approaches minimize the risk by determining trajectories that might result in less damage in the case of failure. However, the risk of the loss of thrust can be eliminated by executing a safe emergency landing if a landing site is reachable. Therefore, we propose a novel risk-aware trajectory planning that minimizes the risk to people on the ground while an option of a safe emergency landing in the case of loss of thrust is guaranteed. The proposed method has been empirically evaluated on a realistic urban scenario. Based on the reported results, an improvement in the risk reduction is achieved compared to the shortest and risk-aware only trajectory. The proposed risk-aware planning with safe emergency landing seems to be suitable trajectory planning for urban air mobility.
|
|
ThAT2 Special Session, Rhone 1 |
|
On Enhancing Predictive Capabilities into and Around Advanced Process
Control |
Chair: Borodin, Valeria | Mines Saint-Etienne |
Co-Chair: Roussy, Agnès | Mines Saint-Etienne |
Organizer: Borodin, Valeria | Mines Saint-Etienne |
Organizer: Juge, Michel | STMicroelectronics |
Organizer: Le Cunff, Delphine | STMicroelectronics |
Organizer: Roussy, Agnès | Mines Saint-Etienne |
|
10:00-10:20, Paper ThAT2.1 | |
>A Comparative Evaluation of Deep Learning Anomaly Detection Techniques on Semiconductor Multivariate Time Series Data (I) |
|
Tchatchoua, Philip | Aix Marseille Université, Université De Toulon, CNRS, LIS (UMR 7 |
Graton, Guillaume | Ecole Centrale De Marseille |
Mustapha, Ouladsine | Aix-Marseille Université |
Juge, Michel | STMicroelectronics |
Keywords: Failure Detection and Recovery, AI-Based Methods, Semiconductor Manufacturing
Abstract: In industrial processes, keeping equipment units in good operating conditions while reducing maintenance costs is one of the most important objectives to improve productivity. One of the ways to do so is to early detect equipment dysfunction. This can be done by analyzing the massive amounts of data collected via numerous sensors during production activities. Thanks to the advantages of distributed architecture and computation efficiency improvements, deep learning methods have gained much interest and have been investigated by researchers for thorough industrial data analysis, notably for anomaly detection. A comparative evaluation of data-driven deep learning methods used to detect anomalies occurring during equipment processing is proposed. This evaluation is done on a total of six methods, their ability to detect anomalies on raw sensor data, collected on semiconductor machines, with variable correlations and temporal dependencies is discussed. The evaluation gives an insight on the industrial performances on the methods and shows how the supervised learning methods outperform the other models with less training time but need labelled data meanwhile some semi-supervised learning methods have good detection performances with training done on normal data only. Index Terms— Anomaly detection, data-driven methods, deep learning, multivariate analysis, raw sensor data, semiconductor manufacturing.
|
|
10:20-10:40, Paper ThAT2.2 | |
>Virtual Metrology for Semiconductor Manufacturing: Focus on Transfer Learning (I) |
|
Clain, Rebecca | Mines Saint-Etienne |
Borodin, Valeria | Mines Saint-Etienne |
Juge, Michel | STMicroelectronics |
Roussy, Agnès | Mines Saint-Etienne |
Keywords: Semiconductor Manufacturing, AI-Based Methods, Process Control
Abstract: Nowadays, virtual metrology models for semiconductor manufacturing aim to be scalable. A virtual Metrology (VM) system is intended to cover a wide spectrum of production contexts. However, due to the large numbers of possible combinations of recipes, tools and chambers, it is intractable to model each context separately. This work presents a VM modeling approach based on the paradigm of transfer learning in a fragmented production context. The approach exploits a 2-Dimensional Convolutional Neural Network (2D-CNN) architecture, namely Spatial Pyramid Pooling Network (SPP-net), to perform the transfer learning from source to target domains with input of different sizes. We implemented several transfer learning strategies on a benchmark dataset provided by the Prognostics and Health Management competition in 2016. The main key points of the proposed approach are discussed based on the findings gathered from the numerical analysis.
|
|
10:40-11:00, Paper ThAT2.3 | |
>Analysis of the Degradation of a Machine in Semiconductor Manufacturing for Maintenance Decision Support (I) |
|
Moritz, Alexandre | MINES Saint-Etienne |
Dauzere-Peres, Stephane | Mines Saint-Etienne |
Ben-Ammar, Oussama | MINES Saint-Etienne |
Vialletelle, Philippe | STMicroelectronics |
|
|
11:00-11:20, Paper ThAT2.4 | |
>Out-Of-Control Detection in Semiconductor Manufacturing Using One-Class Support Vector Machines (I) |
|
Rabhi, Ilham | Mines Saint-Etienne, STMicroelectronics |
Roussy, Agnès | Mines Saint-Etienne |
Pasqualini, Francois | STMicroelectronics |
Alegret, Cyril | STMicroelectronics |
Keywords: Process Control, Machine learning, Big-Data and Data Mining
Abstract: Semiconductor manufacturing is a continuously challenging and competitive industry. It is important to detect any anomalies in the production facilities, or fabs, as they occur to avoid defect accumulations and loss of performance. In this paper we present a literature review of classification methods and detailed the chosen method which is One Class-Support Vector Machine (OC-SVM). This method is used for out-of-control detection in semiconductor manufacturing. The method is tested via an application using industrial data of the studied fab.
|
|
11:20-11:40, Paper ThAT2.5 | |
>Virtual Metrology to Eliminate Test Wafers Measurements on Copper Electroplating Deposition (I) |
|
Doinychko, Anastasiia | Université Grenoble Alpes, Laboratoire d'Informatique De Grenobl |
Amato, Umberto | Italian National Research Council |
Raitsyn, Stanislav | NVIDIA |
Perna, Stefania | STMicroelectronics |
Blundo, Franco | STMicroelectronics |
Genua, Caterina | STMicroelectronics |
Vinciguerra, Daniele | STMicroelectronics |
La Magna, Antonino | CNR-IMM |
Torres, J Andres | Siemens EDA |
Rosenbaum, Alex | NVIDIA |
Amini, Massih-Reza | Université Grenoble Alpes, Laboratoire d'Informatique De Grenobl |
Vasquez, Patrizia | STMicroelectronics |
Keywords: Semiconductor Manufacturing, Machine learning
Abstract: There is a major effort in semiconductor manufacturing to develop a methodology that provides necessary in-line control for each produced integrated circuit (IC). A virtual metrology (VM) task is a key solution for monitoring critical wafer parameters in-between scheduled measurement check-ups. Its goal is to provide an estimation of metrology values as a function of the process state features. In state-of-the-art literature, linear and non-linear methods are proposed to solve VM tasks but for only specific tool conditions regarding different case studies such as chemical vapor deposition, factory-wide control, etch depth prediction, and more. However, every new process setting may require different or adapted to the case VM model. This work provides an overview of different machine learning algorithms for VM task in an application with copper electroplating deposition (Cu ECD) process data. It proposes a method for feature extraction that allows reducing the size of the feature vector required for analysis but still keeping similarity in representation with the original data set. It shows that gradient boosted decision trees and stepwise regression models better perform and generalize compared with widely used in literature neural networks.
|
|
11:40-12:00, Paper ThAT2.6 | |
>On Feature Selection for Virtual Metrology |
|
Djedidi, Oussama | Ecole Des Mines Saint-Etienne |
Borodin, Valeria | Mines Saint-Etienne |
Juge, Michel | STMicroelectronics |
Roussy, Agnès | Mines Saint-Etienne |
Keywords: Semiconductor Manufacturing, Big-Data and Data Mining, Machine learning
Abstract: In semiconductor manufacturing, the increase of low-cost and high-volume demand for integrated circuits drives technological advances and high levels of automation, by increasingly strengthening the process control. Metrology tools are used to perform identical measurements of a set of key manufacturing parameters to help controlling production environments. Physical measurements done on wafers are costly and time-consuming, leading to the development of virtual metrology. Based on the available historical observations, virtual metrology aims to predict one or several parameters of wafers according to their current states. One of the main challenges occurring when building a virtual metrology model is the feature selection. Raw data collected on production machines and measurements sampled at high frequencies result in large datasets. Feature selection aims to preserve only the relevant features to accurately predict the metrology values. In this work, we propose a feature selection algorithm combining machine learning and a genetic algorithm. Extensive computational experiments have been conducted to evaluate the performance of the proposed approach on an industrial dataset.
|
|
ThAT3 Regular Session, Rhone 2 |
|
Collaborative Robots in Manufacturing |
Chair: Henaff, Patrick | Université De Lorraine, CNRS, INRIA, LORIA, F-54000 Nancy, France |
Co-Chair: Chaikovskaia, Mari | LIMOS, INP Clermont Auvergne |
|
10:00-10:20, Paper ThAT3.1 | |
>Non-Linear Stiffness Modeling of Multi-Link Compliant Serial Manipulator Composed of Multiple Tensegrity Segments |
|
Zhao, Wanda | Ecole Centrale Nantes |
Pashkevich, Anatol | Ecole Des Mines De Nantes |
Chablat, Damien | Laboratoire Des Sciences Du Numérique De Nantes |
Keywords: Collaborative Robots in Manufacturing, Product Design, Development and Prototyping, Optimization and Optimal Control
Abstract: The paper focuses on the stiffness modeling of a new type of compliant manipulator and its non-linear behavior while interacting with the environment. The manipulator under study is a serial mechanical structure composed of dual-triangle segments. The main attention is paid to the initial straight configuration which may suddenly change its shape under the loading. It was discovered that under the external loading such manipulator may have six equilibrium configurations but only two of them are stable. In the neighborhood of these configurations, the manipulator behavior was analyzed using the Virtual Joint Method (VJM). This approach allowed us to propose an analytical technique for computing a critical force causing the buckling and evaluate the manipulator shape under the loading. A relevant simulation study confirmed the validity of the developed technique and its advantages in non-linear stiffness analysis.
|
|
10:20-10:40, Paper ThAT3.2 | |
>Reinforcement Learning-Based Learning from Demonstrations for Collaborative Robots |
|
Li, Weidong | Wuhan University of Technology |
Keywords: Collaborative Robots in Manufacturing, Human Factors and Human-in-the-Loop
Abstract: Learning from Demonstrations (LfD) can support a human operator to control a collaborative robot (cobot) in an intuitive means. Gaussian Mixture Model and Gaussian Mixture Regression (GMM and GMR) are useful tools for implementing such a LfD approach. However, well-performed GMM/GMR require a series of demonstrations without trembling and jerky features, which is challenging to achieve in practical applications. To address this issue, in this paper, an improved Reinforcement Learning (RL)-based approach for GMM/GMR is devised to carry out a variety of complex tasks. The innovations of the research are twofold: firstly, a Gaussian noise strategy is designed to scatter demonstrations with trembling and jerky features to better support the optimization of GMM/GMR; Secondly, an improved RL-based optimization algorithm is developed to eliminate potential under-/over-fitting GMM/GMR. A cases study was conducted to verify the approach. Experimental results and comparative analyses showed that this developed approach exhibited good performances in computational efficiency and solution quality.
|
|
10:40-11:00, Paper ThAT3.3 | |
>Towards Context-Aware Natural Language Understanding in Human-Robot-Collaboration |
|
Haase, Tobias | German Aerospace Center (DLR) |
Schönheits, Manfred | German Aerospace Center (DLR) |
Keywords: Collaborative Robots in Manufacturing, Human-Centered Automation
Abstract: Human-robot-collaboration still poses a challenge to contemporary robotic systems. Besides functional safety, usability and user experience are topics of current research. This paper describes an approach that combines contextual-awareness and natural language understanding (NLU) to convey human intents to a robotic system. A user study is then conducted using the system in a simplified collaborative assembly scenario and the findings are discussed.
|
|
11:00-11:20, Paper ThAT3.4 | |
>Sizing of a Fleet of Cooperative Robots for the Transport of Homogeneous Loads |
|
Chaikovskaia, Mari | LIMOS, INP Clermont Auvergne |
Gayon, Jean-Philippe | LIMOS, INP Clermont Auvergne |
Chebab, Zine Elabidine | MecaBotiX |
Fauroux, Jean-Christophe | MecaBotiX |
Keywords: Collaborative Robots in Manufacturing, Logistics, Optimization and Optimal Control
Abstract: We consider the problem of determining the number of robots necessary to transport a set of homogeneous loads in a given time interval from a zone A to a zone B, at minimum cost. The cost is function of the number of robots and of the distance travelled by robots. The operations are divided into several phases: loading, loaded travel, unloading, empty travel and battery charging. The case of non-cooperative robots is considered for which we derive a closed-form expression for the optimal number of robots. We then consider the case of cooperative robots where loads can be carried either by a single robot (mono-robot) or by several robots that cooperate (poly-robot). The fleet sizing problem can be formulated as a mathematical programming. We distinguish several scenarios, depending on the respective carrying capacity of mono-robots and poly-robots. Finally, the infinite horizon problem is also addressed, which models a fleet of vehicles and leads to simpler results.
|
|
11:20-11:40, Paper ThAT3.5 | |
>Conceptual Design and Feasibility Test of Foldable Robotic Arms for Collaborative Work : FRAC |
|
Suthar, Bhivraj | Chungnam National University |
Jung, Seul | Chungnam National University |
Keywords: Physically Assistive Devices, Mechanism Design in Meso, Micro and Nano Scale, Collaborative Robots in Manufacturing
Abstract: A new type of wearable robot is introduced to provide supplementary arms for assisting the worker's tasks. The supplementary arm is metamorphic called Foldable Robotic Arms for Collaboration (FRAC) to extend the operational degrees-of-freedom. FRAC provides additional two arms for a worker to augment the tasks with the environment where the additional arms are necessary. Initially, the FRAC is folded for saving the workspace to allow more space for human arms to interact with the environment. It is unfolded when the worker needs extra arms to hold the object. The strategy for assisting the worker is described, and kinematics and simulation analysis of the folding and unfolding are performed. FRAC is developed for its feasible functionality by performing the conceptual drilling and screwing operation for construction work with the FRAC.
|
|
11:40-12:00, Paper ThAT3.6 | |
>Dynamic Oscillators to Compensate Master Devices Imperfections in Robots Teleoperation Tasks Requiring Dynamic Movements |
|
Menges, Baptiste | Laboratoire Lorrain De Recherche En Informatique Et Ses Applicat |
Henaff, Patrick | Université De Lorraine, CNRS, INRIA, LORIA, F-54000 Nancy, Fra |
Guenard, Adrien | CNRS |
Keywords: Collaborative Robots in Manufacturing, Telerobotics and Teleoperation, Robust/Adaptive Control
Abstract: This paper presents the benefits of using dynamic oscillators when teleoperated robot are controlled with master devices generating discontinuous control signals. A comparative study is carried on the example of a brushing task of a cast iron casting trough. The results show that an oscillator can compensate imperfections of master devices while reducing the power consumed by the robot. This solution can be interesting in the implementation of a robust master device in severe industry environments.
|
|
ThAT4 Special Session, Rhone 3A |
|
Data Driven Automation in Decision Making for Intelligent Systems |
Chair: Qin, Wei | Shanghai Jiao Tong University |
Co-Chair: Zürn, Manuel | Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW), University of Stuttgart |
Organizer: Qin, Wei | Shanghai Jiao Tong University |
Organizer: Pan, Ershun | Shanghai Jiao Tong University |
Organizer: Zhou, Yaoming | Shanghai Jiao Tong University |
|
10:00-10:20, Paper ThAT4.1 | |
>Performance Evaluation of Real-Time ROS2 Robotic Control in a Time-Synchronized Distributed Network |
|
Puck, Lennart | FZI Forschungszentrum Informatik |
Keller, Philip | FZI Forschungszentrum Informatik |
Schnell, Tristan | FZI Forschungszentrum Informatik |
Plasberg, Carsten | FZI Forschungszentrum Informatik |
Tanev, Atanas | FZI Forschungszentrum Informatik |
Heppner, Georg | Forschungszentrum Informatik |
Roennau, Arne | FZI Forschungszentrum Informatik, Karlsruhe |
Dillmann, Rüdiger | FZI - Forschungszentrum Informatik - Karlsruhe |
Keywords: Domain-specific Software and Software Engineering, Software, Middleware and Programming Environments, Control Architectures and Programming
Abstract: Modern robots are mainly controlled by monolithic black-box controllers provided by the individual manufacturers. In research institutions the first version of the Robot Operating System (ROS1) is widely used for different applications. However, ROS1 lacks real-time capable communication. The ongoing development of ROS2 promises to break this paradigm. By employing Data Distribution Service (DDS) as a middleware the modular architecture aims at providing real-time capabilities. This study assesses the current prospects and limitations of ROS2. It gains novel insights towards improved and, in particular, reliable results regarding latencies and jitter. To this end, the allocation and transmission of ROS2 messages is evaluated in an example application for robotic control. An oscilloscope is applied for external validation of the measurements in such a time-synchronized distributed network. The complete application is set up from non-real-time object detection towards real-time control via ROS2 and EtherCAT. An in-depth evaluation of the ROS2 communication stack on a single host and in distributed setups is included. With real-time safe memory allocation and highly privileged ROS2 processes real-time capabilities are ensured.
|
|
10:20-10:40, Paper ThAT4.2 | |
>Modelling and Prediction of Injection Molding Process Using Copula Entropy and Multi-Output SVR (I) |
|
Sun, Yanning | Shanghai Jiao Tong University |
Chen, Yu | Shanghai Jiao Tong University |
Wang, Wuyin | Shanghai Jiao Tong University |
Xu, Hongwei | Shanghai Jiao Tong University |
Qin, Wei | Shanghai Jiao Tong University |
Keywords: Big-Data and Data Mining, AI-Based Methods, Machine learning
Abstract: Optimization and parameter adjustment of an injection molding (IM) process depend largely on a good modelling and prediction of industrial process, which has been received considerable attention in recent years. However, IM process is a typical multivariate production process with multiple product quality indices. It poses a great challenge for multi-output quality prediction problem to select key process variables as input with good interpretability. This study proposes a multivariate quality prediction method for IM process using copula entropy (CE) and multi-output support vector regression (MSVR). First, copula entropy is employed to characterize the association relationships between each process variable and the set of quality indices, thus key process variables can be selected by ranking CE. Then, the quantitative relationship between key process variables and quality indices is established by MSVR. Finally, the proposed method is tested by the experiment on a real IM process dataset. This study will provide an important reference for modelling and prediction of IM process and other multi-output problems.
|
|
10:40-11:00, Paper ThAT4.3 | |
>Covariance Matrix Adaptation Based Tuning of Mass Spectrometry Parameters Using Experimental Probability Distributions (I) |
|
Gioioso, Marisa | Waters Corporation |
Kurmi, Akshay | Northeastern University |
Keywords: Learning and Adaptive Systems, Probability and Statistical Methods, Machine learning
Abstract: The operation of a mass spectrometry instrument, used in analytical chemistry, for custom applications requires the careful tuning of several instrument settings by an expert. In this work, we developed a model that allows the instrument to tune itself. The approach employs a fast, adaptive evolutionary algorithm, the Covariance Matrix Adaptation evolutionary strategy, to tune a mass spectrometry instrument. By developing a scheme for normalizing the values of the outcome variables (resolution, intensity and peak shape of a calibrant peak signal) based on their experimental probability distributions, we combined the outcomes into a single score that was used as the fitness score for the search algorithm. This approach resulted in a more thorough examination of the search space, but in an economical amount of time by being adaptive, resulting in a more stable tuning, no matter the initial state of the settings involved.
|
|
11:00-11:20, Paper ThAT4.4 | |
>The Design of Hybrid Hub-And-Spoke Networks for Large-Scale Dynamic Express Logistics: A Case Study of Chinese Express (I) |
|
Li, Yue | Shanghai Jiao Tong University |
Zhuang, Zilong | Shanghai Jiao Tong University |
Qin, Wei | Shanghai Jiao Tong University |
Keywords: Logistics, Intelligent Transportation Systems, Hybrid Logical/Dynamical Planning and Verification
Abstract: With the continuous growth of express demand, the impact of its fluctuations is becoming more and more significant. The traditional H&S networks cannot respond intelligently to the demand changes. Therefore, a new network design method combining information and automation needs to be developed urgently. This paper considers a multi-hub version and proposes a hybrid network to dynamically design the hubs' locations and the straight connections between nodes. A mixed integer linear programming model is formulated, and a two-stage genetic algorithm is developed to solve the small and large-scale instances of the hybrid H&S network design problem. The MILP and heuristic algorithm are tested on instances provided by an express delivery giant from China. Experimental results verify the effectiveness and economy of the hybrid H&S network and the ability of proposed heuristic algorithm to quickly find the approximate optimal solution.
|
|
11:20-11:40, Paper ThAT4.5 | |
>Learning to Rank High Closeness Centrality Nodes in a Given Network Based on RankNet Method (I) |
|
Chen, Yu | Shanghai Jiao Tong University |
Zhuang, Zilong | Shanghai Jiao Tong University |
Qin, Wei | Shanghai Jiao Tong University |
Keywords: Machine learning, Intelligent Transportation Systems, Logistics
Abstract: Identifying nodes with location advantage in large-scale networks is crucial to the decision making of spatial strategic deployment. Many intelligent decisions can benefit from the discovery of the most central nodes, including the planning of airline hub, transport hub and regional logistics hub. Closeness centrality is a widely used node centrality metric measuring how central a node is. However, the exact calculation of closeness centrality is not feasible for large networks due to its high computational complexity. In this paper, we propose a new approach for fast estimation of closeness rank without calculating the closeness centrality of all nodes in a given network. A small number of nodes are used as training samples and several low-complexity metrics are used as input features. We apply a neural network-based learning to rank method to train the model. Experimental results demonstrate that our model dramatically outperforms the regression model proposed in previous work on synthetic and large real-world networks.
|
|
11:40-12:00, Paper ThAT4.6 | |
>Kinematic Trajectory Following Control for Constrained Deformable Linear Objects |
|
Zürn, Manuel | Institute for Control Engineering of Machine Tools and Manufactu |
Wnuk, Markus | Institute for Control Engineering of Machine Tools and Manufactu |
Hinze, Christoph | Institute for Control Engineering of Machine Tools and Manufactu |
Lechler, Armin | University Stuttgart |
Verl, Alexander | University of Stuttgart |
Xu, Weiliang | The University of Auckland |
Keywords: Control Architectures and Programming, Software Architecture for Robotic and Automation, Process Control
Abstract: Abstract—deformable linear objects (DLOs) such as hoses, cables or wires are often a limiting factor in robotic manipulation. For an industrial robot control, their nonlinear dynamics combined with their large variations in material parameters make it challenging to handle. Therefore, DLOs are usually manipulated manually in the industry, offering a huge potential for automation. This paper contributes to the field of robotic handling of DLOs by presenting and evaluating a novel model based controller and observer concept which is based on a multibody simulation. The presented closed-loop framework of a kinematic trajectory following controller with an online point cloud processing observer is first evaluated separately in the simulation and then in an experiment, showing its potential for cable routing scenarios.
|
|
ThAT5 Regular Session, Rhone 3B |
|
Robotics and Automation in Construction |
Chair: Fourie, Dehann | Massachusetts Institute of Technology and Woods Hole Oceanographic Institution |
Co-Chair: Yu, Wen | CINVESTAV-IPN |
|
10:00-10:20, Paper ThAT5.1 | |
>A MAV Platform for Indoors and Outdoors Autonomous Navigation in GPS-Denied Environments |
|
Reyes-Munoz, Julio Alberto | The University of Texas at El Paso |
Flores Abad, Angel | University of Texas at El Paso |
Keywords: Power and Energy Systems automation, Industrial and Service Robotics, Automation in Construction
Abstract: As Micro Aerial Vehicles (MAV) technology becomes mature and more capable, its use for automating risky and repetitive tasks has been explored, even in challenging environments. This has resulted in a wide array of expensive platforms specialized for solving specific problems that sometimes do not generalize to other applications. In this work we develop a MAV system, UTAINS, capable of navigating in GPS-denied environments for indoors and outdoors applications.The main design target for the platform is autonomous aerial inspections of power plants, so we first introduce a survey about flying robots for inspection and exploration tasks in GPS-denied environments, presenting the proposed systems in terms of their autonomy level. Furthermore, we present the design of a modular low-cost system that can easily scale for a wide range of applications, both from a software and a hardware perspective. The software architecture is subdivided in modules that implement the key competences for autonomous navigation: motion control, perception, localization, and planning. We show the successful implementation of a design aimed for power plants inspection in GPS-denied environments through real flights following a planned trajectory, both in indoors and outdoors scenarios.
|
|
10:20-10:40, Paper ThAT5.2 | |
>Multimodal Navigation-Affordance Matching for SLAM |
|
Terblanche, Johan | IDT4 |
Claassens, Samuel David | Semisorted Technologies |
Fourie, Dehann | Massachusetts Institute of Technology and Woods Hole Oceanograph |
Keywords: Robotics and Automation in Construction, AI-Based Methods, Probability and Statistical Methods
Abstract: In robotics and mapping prior knowledge of an environment can be included as virtual assets to a simultaneous localization and mapping (SLAM) solution. Borrowing the concept of affordances from robotic manipulation (i.e. virtual/interactive object models/primitives), this work addresses the fundamental duality in discrepancies between virtual and physical structures for localization and mapping. We propose a multimodal/non-Gaussian solution as a fundamental mechanism to leverage navigation-affordances assets during the localization and mapping process while simultaneously identifying any mismatches from the physical object. This allows the localization and mapping state-estimate more robust access to non-conventional and imperfect prior information about the environment, while computationally identifying assumed model discrepancies from imperfect sensor data. We use non-Gaussian factor graphs as modeling language to incorporate navigation-affordances with multi-sensor data similar to SLAM methods. We illustrate the approach with synthesized and real-world data from the construction industry where digital assets (such as drawings or models) are good proxies for how navigation-affordances can be generated and used in general.
|
|
10:40-11:00, Paper ThAT5.3 | |
>Learning of Causal Observable Functions for Koopman-DFL Lifting Linearization of Nonlinear Controlled Systems and Its Application to Excavation Automation |
|
Selby, Nicholas Stearns | MIT |
Asada, Harry | MIT |
Keywords: Robotics and Automation in Construction, Reinforcement Learning
Abstract: Effective and causal observable functions for low-order lifting linearization of nonlinear controlled systems are learned from data by using neural networks. While Koopman operator theory allows us to represent a nonlinear system as a linear system in an infinite-dimensional space of observables, exact linearization is guaranteed only for autonomous systems with no input, and finding effective observable functions for approximation with a low-order linear system remains an open question. Dual-Faceted Linearization uses a set of effective observables for low-order lifting linearization, but the method requires knowledge of the physical structure of the nonlinear system. Here, a data-driven method is presented for generating a set of nonlinear observable functions that can accurately approximate a nonlinear control system to a low-order linear control system. A caveat in using data of measured variables as observables is that the measured variables may contain input to the system, which incurs a causality contradiction when lifting the system, i.e. taking derivatives of the observables. The current work presents a method for eliminating such anti-causal components of the observables and lifting the system using only causal observables. The method is applied to excavation automation, a complex nonlinear dynamical system, to obtain a low-order lifted linear model for control design.
|
|
11:00-11:20, Paper ThAT5.4 | |
>BALTO: A BIM-Integrated Mobile Robot Manipulator for Precise and Autonomous Disinfection in Buildings against COVID-19 |
> Video
|
|
Giusti, Andrea | Fraunhofer Italia Research |
Magnago, Valerio | University of Trento |
Siegele, Dietmar | Fraunhofer Italia Research |
Terzer, Michael | Fraunhofer Italia Research Scarl |
Follini, Camilla | Fraunhofer Italia Research S.c.a.r.l |
Garbin, Simone | Fraunhofer Italia Research |
Marcher, Carmen | Fraunhofer Italia Research, Free University of Bozen-Bolzano |
Steiner, Dieter | Fraunhofer Italia Research |
Schweigkofler, Alice | Fraunhofer Italia |
Riedl, Michael | Fraunhofer Italia |
Keywords: Robotics and Automation in Construction, Software Architecture for Robotic and Automation
Abstract: We present a novel mobile robot manipulator named BALTO, which can precisely and autonomously disinfect critical items of buildings, yet be easily deployed in different environments by non-expert operators. Conventional approaches based on greedy disinfection actions can be dangerous for surrounding humans, or provide non-controlled coverage of critical items such as door-handles, push-bars or switches. Other robotic solutions providing precision disinfection can require time-consuming geometric and semantic mapping of the environment, placement of non-natural landmarks, or the presence of an operator on-site for programming to adapt to the use case. Our approach is different: we integrate preliminary knowledge of the building within the robot control system itself, using Building Information Modelling (BIM) data readily available from modern construction processes. BIM models completely describe a building and its components with the corresponding geometric and semantic data. We use these data to provide the robot with the components to disinfect and their location, as well as the description of the environment in which it operates, without requiring any prior mapping procedure. Precise disinfection actions are enabled by an automatic calibration procedure based on a learned 3D model of the target. We finally present results of experiments performed in operational environment that verify the effectiveness of our proposed solution.
|
|
ThAT6 Regular Session, St Clair 1 |
|
Intelligent Transportation Systems |
Chair: Garaix, Thierry | Ecole Nationale Superieure De Mines De Saint Etienne |
Co-Chair: Seiler, Konstantin M | The University of Sydney |
|
10:00-10:20, Paper ThAT6.1 | |
>A Data-Driven Multi-Device Collaborative Control Method in Coal Transportation System |
|
You, Bo | Institute of Automation, Chinese Academy of Sciences |
Kang, Liu | University of Chinese Academy of Sciences |
Wang, Shuaishuai | University of Chinese Academy of Sciences |
Li, Xue-en | Chinese Academy of Sciences |
Keywords: Intelligent Transportation Systems, Big-Data and Data Mining, Machine learning
Abstract: The coal transportation system comprises several belt conveyors, which includes lots of electrical equipment. The devices are interconnected, and a problem with one of them will cause all devices to be paralyzed. So rapid and accurate multi-device collaborative control plays an important role in high safety performance and production efficiency. Traditional multi-device collaborative control algorithms depend on complex modeling with large-scale, complex constraints, uncertainties, and multi-objective conditions. Here, we propose a data-driven multi-device collaborative control method. In this paper, the neural network algorithm is selected to model the relationship between the equipment control operation and the environment, equipment status, and human activities, thus forming the multi-device collaborative control operation knowledge to guide the multi-device collaborative control in the actual operation process. Furthermore, experiments on real production datasets demonstrate the proposed approach can realize multi-device collaborative control in the coal transportation system, meeting the three goals of safety, energy-saving, and high efficiency simultaneously.
|
|
10:20-10:40, Paper ThAT6.2 | |
>An Autonomous Driving Framework for Long-Term Decision-Making and Short-Term Trajectory Planning on Frenet Space |
|
Moghadam, Majid | University of California, Santa Cruz |
Elkaim, Gabriel Hugh | UC Santa Cruz |
Keywords: Intelligent Transportation Systems, Motion and Path Planning, Collision Avoidance
Abstract: In this paper, we present a hierarchical framework for decision-making and planning on highway driving tasks. We utilized intelligent driving models (IDM and MOBIL) to generate long-term decisions based on the traffic situa- tion flowing around the ego. The decisions both maximize ego performance while respecting other vehicles’ objectives. Short-term trajectory optimization is performed on the Frenet space to make the calculations invariant to the road’s three- dimensional curvatures. A novel obstacle avoidance approach is introduced on the Frenet frame for the moving obstacles. The optimization explores the driving corridors to generate spatiotemporal polynomial trajectories to navigate through the traffic safely and obey the BP commands. The framework also introduces a heuristic supervisor that identifies unexpected situations and recalculates each module in case of a potential emergency. Experiments in CARLA simulation have shown the potential and the scalability of the framework in implementing various driving styles that match human behavior.
|
|
10:40-11:00, Paper ThAT6.3 | |
>A Deep Reinforcement Learning Approach for Long-Term Short-Term Planning on Frenet Frame |
|
Moghadam, Majid | University of California, Santa Cruz |
Alizadeh, Ali | Istanbul Technical University |
Tekin, Engin | University of California Santa Cruz |
Elkaim, Gabriel Hugh | UC Santa Cruz |
Keywords: Intelligent Transportation Systems, Motion and Path Planning, Reinforcement
Abstract: Tactical decision-making and strategic motion planning for autonomous highway driving are challenging due to predicting other road users' behaviors, diversity of environments, and complexity of the traffic interactions. This paper presents a novel end-to-end continuous deep reinforcement learning approach towards autonomous cars' decision-making and motion planning. For the first time, we define both states and action spaces on the Frenet space to make the driving behavior less variant to the road curvatures than the surrounding actors' dynamics and traffic interactions. The agent receives time-series data of past trajectories of the surrounding vehicles and applies convolutional neural networks along the time channels to extract features in the backbone. The algorithm generates continuous spatiotemporal trajectories on the Frenet frame for the feedback controller to track. Extensive high-fidelity highway simulations on CARLA show the superiority of the presented approach compared with commonly used baselines and discrete reinforcement learning on various traffic scenarios. Furthermore, the proposed method's advantage is confirmed with a more comprehensive performance evaluation against 1000 randomly generated test scenarios.
|
|
11:00-11:20, Paper ThAT6.4 | |
>Automated Taxi Queue Management at High-Demand Venues |
|
Ji, Mengyu | Singapore Management University |
Cheng, Shih-Fen | Singapore Management University |
Keywords: Intelligent Transportation Systems, Planning, Scheduling and Coordination
Abstract: In this paper, we seek to identify an effective management policy that could reduce supply-demand gaps at taxi queues serving high-density locations where demand surges frequently happen. Unlike current industry practice, which relies on broadcasting to attract taxis to come and serve the queue, we propose more proactive and adaptive approaches to handle demand surges. Our design objective is to reduce the cumulative supply-demand gaps as much as we could by sending notifications to individual taxis. To address this problem, we first propose a highly effective passenger demand prediction system that is based on the real-time flight arrival information. By monitoring cumulative passenger arrivals, and control for factors such as the flight's departure cities, we demonstrate that a simple linear regression model can accurately predict the number of passengers joining taxi queues. We then propose an optimal control strategy based on a Markov Decision Process to model the decisions of notifying individual taxis that are at different distances away from the airport. By using a real-world dataset, we demonstrate that an accurate passenger demand prediction system is crucial to the effectiveness of taxi queue management. In our numerical studies based on the real-world data, we observe that our proposed approach of optimal control with demand predictions outperforms the same control strategy, yet with Poisson demand assumption, by 43%. Against the status quo, which has no external control, we could reduce the gap by 23%. These results demonstrate that our proposed methodology has strong real-world potential.
|
|
11:20-11:40, Paper ThAT6.5 | |
>Flow-Achieving Online Planning and Dispatching for Continuous Transportation with Autonomous Vehicles |
|
Seiler, Konstantin M | The University of Sydney |
Palmer, Andrew William | Emesent |
Hill, Andrew John | University of Sydney |
Keywords: Intelligent Transportation Systems, Optimization and Optimal Control, Planning, Scheduling and Coordination
Abstract: In large-scale industrial applications, goods must be continuously transported between locations, which in the absence of conveyor systems is by a fleet of individual vehicles. This article introduces flow-achieving scheduling tree (FAST), an online dispatching algorithm that allows vehicles to efficiently operate as a team to maximize the system's throughput while meeting a production schedule. A high-performance model is developed for high-fidelity prediction of vehicle interactions and system performance. It is subsequently optimized using a self-tuning variant of Monte Carlo tree search (MCTS) to make agile dispatch decisions in real time. The method is validated using an open-cut mine site and is shown to outperform a commonly used algorithm in this domain.
|
|
11:40-12:00, Paper ThAT6.6 | |
>Traffic Management of a CBTC Suburban Railway Line: A Simulation-Optimization Approach |
|
Meunier, Hugo | Mines Saint-Etienne and SNCF RESEAU |
Baro, Sylvain | SNCF RESEAU |
Borodin, Valeria | Mines Saint-Etienne |
Dauzere-Peres, Stephane | Mines Saint-Etienne |
Pochet, Juliette | SNCF RESEAU |
Keywords: Intelligent Transportation Systems, Planning, Scheduling and Coordination
Abstract: This paper focuses on the real-time traffic management of an open suburban railway line equipped with a Communication-Based Train Control (CBTC) system. Being dedicated to operating in highly dense areas and to maximizing the traffic capacity, CBTC systems are characterized by complex functional architectures. Aiming to improve the performance of a CBTC suburban railway line, a simulation-optimization approach is proposed to support the efficient real-time management of CBTC trains subject to disturbances. The simulation module reproduces the system dynamics, while satisfying signaling and controlling constraints specific to a CBTC system. The optimization module solves a local train retiming problem in case of disturbances, which minimizes the gap between the current and planned timetables. Based on a conjunctive graph, the train retiming problem is formulated as a linear programming model. Numerical experiments on real-life instances have been conducted to evaluate and validate the proposed approach.
|
|
ThAT7 Special Session, St Clair 2 |
|
Recent Advances in Degradation Data Analysis |
Chair: Zhao, Xiujie | Tianjin University |
Co-Chair: Maus, Maren | RWTH Aachen University |
Organizer: Chen, Piao | Delft University of Technology |
Organizer: Zhao, Xiujie | Tianjin University |
|
10:00-10:20, Paper ThAT7.1 | |
>Sequential Degradation-Based Burn-In Test with Multiple Periodic Inspections (I) |
|
Hu, Jiawen | University of Electronic Science and Technology of China |
Keywords: Probability and Statistical Methods, Diagnosis and Prognostics, Big-Data and Data Mining
Abstract: Burn-in has been proven effective in identifying and removing defective products before they are delivered to customers. Most existing burn-in models adopt a one-shot scheme, which may not be sufficient enough for identification. Borrowing the idea from sequential inspections for remaining useful life prediction and accelerated lifetime test, this study proposes a sequential degradation-based burn-in model with multiple periodic inspections. At each inspection epoch, the posterior probability that a product belongs to a normal one is updated with the inspected degradation level. Based on the degradation level and the updated posterior probability, a product can be disposed, put into field use, or kept in the test till the next inspection epoch. We cast the problem into a partially observed Markov decision process to minimize the expected total burn-in cost of a product, and derive some interesting structures of the optimal policy. Then, algorithms are provided to find the joint optimal inspection period and number of inspections in steps. A numerical study is also provided to illustrate the effectiveness of our proposed model.
|
|
10:20-10:40, Paper ThAT7.2 | |
>A Novel Deep Dual Network with Unsupervised Domain Adaptation for Transfer Fault Prognosis across Different Machines (I) |
|
Cheng-Geng, Huang | Sun Yat-Sen University |
|
|
10:40-11:00, Paper ThAT7.3 | |
>Detection and Differentiation of Replay Attack and Equipment Faults in SCADA Systems |
|
Li, Dan | Georgia Institute of Technology |
Gebraeel, Nagi | Georgia Tech |
Paynabar, Kamran | Georgia Tech |
Keywords: Cyber-physical Production Systems and Industry 4.0, Diagnosis and Prognostics
Abstract: Supervisory control and data acquisition (SCADA) systems are widely used for industrial control of critical infrastructures such as power plants and manufacturing systems. There is abundant evidence of SCADA systems being subject to cyber-attacks. With increasing interest in industrial digitization, cybersecurity of SCADA systems is poised to be ever more important. Equipment faults and cyberattacks can manifest themselves in a similar fashion, i.e., they can exhibit similar signatures. This paper focuses on methods capable of distinguishing equipment faults from bonafide cyber-attacks. Specifically, we consider a relatively sophisticated form of cyber-attack known as the “replay attack”. We derive mathematical formalisms that distinguish the replay attack from several classes of equipment faults and verify our methodology through an extensive numerical study.
|
|
11:00-11:20, Paper ThAT7.4 | |
>Maintenance Policies for Balanced Systems Subject to Degradation (I) |
|
Zhao, Xiujie | Tianjin University |
|
|
11:20-11:40, Paper ThAT7.5 | |
>Individualized Degradation Modeling and Prognostics in a Heterogeneous Group Via Incorporating Intrinsic Covariate Information |
|
Kim, Minhee | University of Wisconsin - Madison |
Song, Changyue | Stevens Institute of Technology |
Liu, Kaibo | University of Wisconsin - Madison |
Keywords: Diagnosis and Prognostics, Probability and Statistical Methods, Data fusion
Abstract: This study focuses on individualized degradation modeling and prognostics for a heterogeneous group, where each individual unit shows a distinct degradation process. Existing degradation models usually treat each unit separately and do not fully utilize the distinct characteristics of each individual. In this study, we propose a generic framework to handle the heterogeneity across units by effectively leveraging the intrinsic covariate information, which is closely related to the unit’s degradation process. Specifically, we employ a multivariate Gaussian process to nonparametrically establish the relation between the covariate information and degradation process. Through modeling the unit similarities based on the covariates, efficient information transfer among units is enabled for better degradation modeling and prognostics, as the collected degradation signals from one unit can be shared with the entire heterogeneous group. A theoretical justification for the proposed model is also investigated. Numerical studies are presented to evaluate the parameter estimation accuracy and the sensitivity of the proposed method, which demonstrate the advantage of the proposed method over existing benchmark approaches.
|
|
11:40-12:00, Paper ThAT7.6 | |
>Degradation Functions for Railway Station Equipment Quality Based on Maintenance-Influenced Data (I) |
|
Maus, Maren | RWTH Aachen University |
Lampe, Felix | VIA | RWTH Aachen University |
Friesen, Nadine | VIA Der RWTH Aachen |
Heckmann, Mark | DB Station&Service AG |
Elfert, Lea | DB Station&Service AG |
Nießen, Nils | RWTH Aachen University, Institute of Transport Science |
Keywords: Planning, Scheduling and Coordination, Diagnosis and Prognostics, Probability and Statistical Methods
Abstract: To find the optimal maintenance policies, the DB Station&Service AG – a railway infrastructure company which manages the majority of the railway stations in Germany – needs to describe the cause-effect-relationship between the funds for maintenance and the quality of the station equipment. To determine the influence of funding on the infrastructure quality, it is necessary to predict the cost and effect of maintenance measures as well as the behaviour of unmaintained items. However, the available degradation data do not contain any condition measurements of items which are not influenced by maintenance. Here, we provide a model which allows to describe the average maintenance intervals and the degradation during these intervals based on such maintenance-influenced data. This model can be used to predict the development of the equipment quality and is intended to be used to determine the necessary budget for a specific equipment quality or vice versa.
|
|
ThAT8 Regular Session, Rhone 4 |
|
Computer Vision in Automation 2 |
Chair: Palermo, Francesca | Queen Mary University of London |
Co-Chair: Stricker, Ronny | Ilmenau University of Technology |
|
10:00-10:20, Paper ThAT8.1 | |
>A Novel Geometric Calibration Method for Active Stereovision System |
> Video
|
|
Liu, Jierui | Institute of Automation, Chinese Academy of Sciences |
Liu, Xilong | Chinese Academy of Sciences |
Cao, Zhiqiang | Institute of Automation, Chinese Academy of Sciences |
Li, Zhonghui | Beijing Engo Technology Co., Ltd |
Yu, Junzhi | Chinese Academy of Sciences |
Keywords: Calibration and Identification, Computer Vision for Manufacturing, Foundations of Automation
Abstract: In this paper, a novel active stereovision calibration method is proposed to determine the time-varying extrinsic parameters of the cameras. This method mainly focuses on the rotation of cameras around the corresponding spatial axes with deviation. In the offline calibration, the installation matrices and baseline matrix are determined by introducing geometrical characteristics that represents the relationship of rotation axis and the transformation of camera rotation. On this basis, the extrinsic parameters can be updated online. The proposed method solves the problem where the spatial relationship between each camera and its rotation axis can be arbitrary, which has been verified by the simulations as well as tracking and measurement experiments on the developed active stereovision platform.
|
|
10:20-10:40, Paper ThAT8.2 | |
>Visual 3D Perception for Interactive Robotic Tactile Data Acquisition |
|
Jain, Siddarth | Mitsubishi Electric Research Laboratories (MERL) |
Corcodel, Radu Ioan | Mitsubishi Electric Research Laboratories |
Vanbaar, Jeroen | MERL |
Keywords: Computer Vision in Automation, Force and Tactile Sensing
Abstract: In this paper, we present a novel approach for tactile saliency computation on 3D point clouds of unseen object instances, where we define salient points as those that provide informative tactile sensory information with robotic interaction. Our intuition is that the local 3D surface geometries of objects contain characteristic information both in terms of texture and shape which can provide important discriminating information for tactile interactions. We solve the problem by taking as input a 3D point cloud of an object and develop a geometric approach which computes the tactile saliency map for the object without requiring pre-training. We furthermore develop a formulation to compute grasps using the tactile saliency for prehensile probing manipulation. We demonstrate our framework with evaluation on a variety of household objects in real-world experiments. Since it is difficult to manually define a ground truth tactile saliency measure, we evaluate our approach by having a human subject provide saliency information as baseline in pilot experiments. Results show good performance of our algorithm both in terms of the computation of tactile saliency and its usefulness to acquire informative tactile sensory data with a real-world robot.
|
|
10:40-11:00, Paper ThAT8.3 | |
>Road Surface Segmentation - Pixel-Perfect Distress and Object Detection for Road Assessment |
|
Stricker, Ronny | Ilmenau University of Technology |
Aganian, Dustin | University of Technology Ilmenau |
Sesselmann, Maximilian | LEHMANN + PARTNER GmbH |
Seichter, Daniel | Ilmenau University of Technology |
Engelhardt, Marius | Ilmenau University of Technology |
Spielhofer, Roland | AIT Austrian Institute of Technology GmbH |
Hahn, Matthias | AIT Austrian Institute of Technology GmbH |
Hautz, Astrid | VIA IMC GmbH |
Debes, Klaus | Ilmenau University of Technologies |
Gross, Horst-Michael | Ilmenau University of Technology |
Keywords: Computer Vision in Automation, Environment Monitoring and Management, Domain-specific Software and Software Engineering
Abstract: Visual road assessment, which is carried out by many countries, involves the evaluation of millions of surface images. This exhaustive task is usually done manually and therefore is costly in terms of time and prone to failure. Different methods for automatic distress detection have been presented in the literature recently. However, most of the approaches are focused on crack detection only. This paper focuses on detecting multiple distress types and object classes on asphalt roads, aiming to fully automate distress detection on road surfaces in Austria, Switzerland, and Germany using image segmentation with neural networks. The paper introduces a distress and object catalog developed by experts of the involved countries that guarantees convertibility into federal distress catalogs. We evaluate the performance gain of different neural network architectures and advanced training techniques by conducting extensive experiments.
|
|
11:00-11:20, Paper ThAT8.4 | |
>Object Pose Estimation Via Pruned Hough Forest with Combined Split Schemes for Robotic Grasp |
|
Dong, Huixu | Carnegie Mellon University |
Prasad, Dilip | Nanyang Technological University |
Chen, I-Ming | Nanyang Technological University |
Keywords: Computer Vision in Automation, Manipulation Planning, Logistics
Abstract: Robotic grasp in complex open-world scenarios requires an effective and generalizable perception. Estimating object’s pose is needed in a variety of practical grasping scenarios. Here we present a novel approach of pose estimation of texture-less and textured objects. The algorithm utilizes a single RGB-D image to exploit depth invariant, oriented point pair feature as well as local contextual sensitivity in cluttered environments. To enhance the performance of the voting process and improve learning efficiency, we employ a global pruning algorithm that reduces the risk of over-fitting and simplifies the structure of decision trees after compensating for the complementary information among multiple trees by optimizing a designed global objective function. Finally, we also refine the pose obtained from the above stage. The proposed approach of estimating 6D (Degree of Freedom) poses of textured and texture-less objects is evaluated on publicly available datasets against the recent works under various conditions. It illustrates that our framework is superior to these recent works. Further, we perform extensive qualitative experiments of robotic grasp to illustrate the proposed approach can be applied to practical scenarios.
|
|
11:20-11:40, Paper ThAT8.5 | |
>Towards an Interpretable Deep Driving Network by Attentional Bottleneck |
> Video
|
|
Kim, Jinkyu | UC Berkeley |
Bansal, Mayank | Waymo |
Keywords: Computer Vision for Automation, Deep Learning Methods
Abstract: Deep neural networks are a key component of behavior prediction and motion generation for self-driving cars. One of their main drawbacks is a lack of transparency: they should provide easy to interpret rationales for what triggers certain behaviors. We propose an architecture called Attentional Bottleneck with the goal of improving transparency. Our key idea is to combine visual attention, which identifies what aspects of the input the model is using, with an information bottleneck that enables the model to only use aspects of the input which are important. This not only provides sparse and interpretable attention maps (e.g. focusing only on specific vehicles in the scene), but it adds this transparency at no cost to model accuracy. In fact, we find improvements in accuracy when applying Attentional Bottleneck to the ChauffeurNet model, whereas we find that the accuracy deteriorates with a traditional visual attention model.
|
|
11:40-12:00, Paper ThAT8.6 | |
>Multi-Modal Robotic Visual-Tactile Localisation and Detection of Surface Cracks |
> Video
|
|
Palermo, Francesca | Queen Mary University of London |
Rincon Ardila, Liz Katherine | Tokyo University of Agriculture and Technology |
Oh, Changjae | Queen Mary University of London |
Althoefer, Kaspar | Queen Mary University of London |
Poslad, Stefan | Queen Mary University of London |
Venture, Gentiane | Tokyo University of Agriculture and Technology |
Farkhatdinov, Ildar | Queen Mary University of London |
Keywords: Computer Vision for Automation, Environment Monitoring and Management, Robotics and Automation in Construction
Abstract: We present and validate a method to detect surface cracks with visual and tactile sensing. The proposed algorithm localises cracks in remote environments through videos/photos taken by an on-board robot camera. The identified areas of interest are then explored by a robot with a tactile sensor. Faster R-CNN object detection is used for identifying the location of potential cracks. Random forest classifier is used for tactile identification of the cracks to confirm their presences. Offline and online experiments to compare vision only and combined vision and tactile based crack detection are demonstrated. Two experiments are developed to test the efficiency of the multi-modal approach: online accuracy detection and time required to explore a surface and localise a crack. Exploring a cracked surface using combined visual and tactile modalities required four times less time than using the tactile modality only. The accuracy of detection was also improved with the combination of the two modalities. This approach may be implemented also in extreme environments since gamma radiation does not interfere with the sensing mechanism of fibre optic-based sensors.
|
|
ThAT9 Special Session, St Clair 3A |
|
Simulation Optimization in New Information Age 1 |
Chair: Luo, Jun | Shanghai Jiao Tong University Antai College of Economics & Management |
Co-Chair: Peng, Yijie | Peking University |
Organizer: Jia, Qing-Shan | Tsinghua University |
Organizer: Luo, Jun | Shanghai Jiao Tong University Antai College of Economics & Management |
Organizer: Pedrielli, Giulia | Arizona State University |
Organizer: Peng, Yijie | Peking University |
|
10:00-10:20, Paper ThAT9.1 | |
>Data Efficient Learning of Implicit Control Strategies in Water Distribution Networks (I) |
|
Candelieri, Antonio | University of Milano-Bicocca |
Ponti, Andrea | University of Milano-Bicocca |
Archetti, Francesco | University of Milano-Bicocca |
Keywords: Optimization and Optimal Control, Modelling, Simulation and Validation of Cyber-physical Energy Systems, Model Learning for Control
Abstract: Bayesian Optimization is proposed for data-efficient learning of optimal control strategies aimed at minimizing the energy related costs for operating pumps in a water distribution network. The control strategies are defined as pressure-based rules, whose parameters are the decision variables of the optimization problem. A probabilistic description is used to model the optimization problem from parameters to energy cost. The probabilistic model is learned from data obtained by testing a set of parameters via software hydraulic simulation. Bayesian Optimization selects the next values of the parameters to evaluate in a principled way, proving to be able to find globally optimal control strategies, within relatively few trials (i.e., software simulation runs). The proposed Bayesian Optimization framework deals with a quite general formalization of the control problem, including constraints, also black box. Relevant results on a real-life water distribution network are reported, also in comparison with Pure Random Search and Genetic Algorithms.
|
|
10:20-10:40, Paper ThAT9.2 | |
>A Strategy for Autonomous Source Searching Using Gaussian Mixture Model to Fit the Estimate of the Source Location (I) |
|
Ji, Yatai | National University of Defense Technology |
Chen, Bin | National University of Defense Technology |
Wang, Yiping | National University of Defense Technology |
Zhao, Yong | National University of Defense Technology |
Zhu, Zhengqiu | National University of Defense Technology |
Keywords: Sensor-based Control, Cognitive Automation, Formal Methods in Robotics and Automation
Abstract: Quickly and accurately locating an unknown emitting source in turbulence is an important research. Previous studies have demonstrated that cognitive strategies can be used to effectively search the source. However, it takes a lot of computation to determine the action by solving the reward function. This paper proposes a fresh searching algorithm named MEGI-taxis, which has a less computational effort than cognitive strategies. It employs the Gaussian Mixture Model (GMM) to extract the information to decide the action instead of the reward function. The area of the Maximum Effective Gaussian dIstribution (MEGI) determined by the GMM is considered as the most possible area of the source. The searcher guided by the MEGI-taxis algorithm has a move tendency towards the MEGI and it explores the area of the MEGI in square search pattern. The results of experiment demonstrate that the MEGI-taxis algorithm has a better performance than cognitive strategies in mean search time (MST) and success rate (SR). The proposed strategy is also more computationally efficient.
|
|
10:40-11:00, Paper ThAT9.3 | |
>Intelligent Yard Crane Scheduling in a New Automated Container Terminal (I) |
|
Xinjia Jiang, Xinjia | Nanjing University of Aeronautics and Astronautics |
Keywords: Intelligent Transportation Systems, Optimization and Optimal Control, Collaborative Robots in Manufacturing
Abstract: In traditional automated container terminals, storage blocks are laid perpendicular to the apron. The yard crane (YC) frequently moves along a block for pick-ups or drop-offs at the seaside/land-side end. To reduce the wasted time for such YC movements, a new design has been proposed to install a rail-mounted ground trolley along each block for container delivery. The Ground Trolley based Automated Container Terminal (GT-ACT) is proven to be very effective in operation, because it reduces gantry moves. However, the interaction between YC and GT increases the planning difficulty of container handling. At the operational level, the container handling efficiency highly depends on the YC scheduling decisions. An interesting YC scheduling problem with GT collaboration is formally discussed in this study.
|
|
11:00-11:20, Paper ThAT9.4 | |
>Service System Resource Allocation Optimization Via Simulation (I) |
|
Chen, Weiwei | Rutgers University |
Gao, Siyang | City University of Hong Kong |
Chen, Wenjie | City University of Hong Kong |
Du, Jianzhong | City University of Hong Kong |
Keywords: Modelling, Simulation and Optimization in Healthcare, Optimization and Optimal Control, Simulation and Animation
Abstract: Service systems often observe performance metrics expressed in the form of probabilistic measures; for example, a public service office may aim to achieve the goal of serving at least 80% of the customers in a waiting time below 15 minutes. Properly allocating limited resources in such systems holds a key to optimize their performance. Further, the systems are best evaluated via simulation due to their complex dynamics and lack of convenient mathematical structures. This note focuses on a class of such resource allocation problems where the objective and constraints are probabilistic measures. To this end, we develop an optimal computing budget allocation (OCBA) formulation to minimize the expected opportunity cost in selecting the optimal resource allocation solution. The asymptotic optimality conditions of the OCBA formulation are derived, based on which an iterative algorithm is developed and its finite-time convergence is discussed. Numerical experiments demonstrate the effectiveness of the proposed algorithm for solving the proposed resource allocation problems via simulation.
|
|
11:20-11:40, Paper ThAT9.5 | |
>Missile Defense Decision-Making under Incomplete Information Using the Artificial Neural Network (I) |
|
Qu, Tianci | Chinese Academy of Sciences |
Xiong, Gang | Institute of Automation, Academy of Sciences |
Dong, Xisong | Institute of Automation, Chinese Academy of Sciences |
Li, Wei | Institute of Automation,Chinese Academy of Sciences |
Tao, Hao | China Ship Development and Design Center |
Yan, Jun | Institute of Automation, Academy of Sciences |
Shen, Zhen | Institute of Automation, Chinese Academy of Sciences |
Wang, Fei-Yue | University of Arizona |
Keywords: Agent-Based Systems, Discrete Event Dynamic Automation Systems, Learning and Adaptive Systems
Abstract: The Anti Saturation Attack is a vital problem in the field of military defense. When confronted with a large scale missile attack in a short duration, how to optimally allocate interceptor missiles to minimize the loss of total assets until the end of the attack has been studied by many researchers. Recently, deep reinforcement learning methods have been applied to achieve a suboptimal defense policy. Yet there still remains some problems. The convergence process is dependent on the computing resources and the accumulated policies can not be well explained. Moreover, when the attack parameters in the environment is changed, the model needs to be trained from the beginning again, which limits the usage in real time decision-making scenarios. To this end, we propose a hybrid method, which speeds up the approximation to the optimal policy and can explain the attack patterns from the statistical point of view. Furthermore, the light-weight structure makes it possible to deploy in real-time cases. First, an integer convex optimization model is set up to generate initial data and feed the neural network. Second, a neural network and TD methods are implemented to predict the final asset value. Last but not least, the potential usage of the proposed method is testified in a simulation environment where different battlefield parameters and experiments have been set up and conducted. With comparison of other two methods (Heuristic and Deep Q-Network), experiments indicate the hybrid method outperforms the other two methods, with convergence speed nearly 20% faster than the DQN method and expectation of final assets value 12% higher when training in the same episodes.
|
|
11:40-12:00, Paper ThAT9.6 | |
>Performance Analysis and Improvement of Serial Lines with Quality Inspection and Rework (I) |
|
Song, Yun-Lei | Northwestern Polytechnical University |
Wang, Jun-Qiang | Northwestern Polytechnical University |
Li, Yang | Tongji Univeristy |
Keywords: Discrete Event Dynamic Automation Systems, Optimization and Optimal Control, Process Control
Abstract: We study serial lines with quality inspection and rework, in which defective parts are sent back to a production machine for processing once they are identified by quality inspection machines. We propose a systematic analysis of the production dynamic, and present a data-driven method to quantitatively evaluate the impact of various disruptions, including machine breakdown, quality failure, on production. A data-driven method is integrated into an optimization method that determines the optimal quality inspection allocation to trade off the costs incurred production loss and quality-related investments. Simulation studied are conducted to validate the proposed methods.
|
|
ThAT10 Regular Session, St Clair 3B |
|
Factory Automation |
Chair: Date, Hisashi | University of Tsukuba |
Co-Chair: Fischer, Juliane | Technical University of Munich |
|
10:00-10:20, Paper ThAT10.1 | |
>Development of Fabric Feed Mechanism Using Horizontal Articulated Dual Manipulator for Automated Sewing |
> Video
|
|
Tajima, Shungo | University of Tsukuba |
Date, Hisashi | University of Tsukuba |
Keywords: Factory Automation, Computer Vision for Manufacturing
Abstract: This study addresses the problem of automating the sewing process in the garment industry. While automatic feeding during sewing has been achieved, overlapping and feeding two a pair of cloths still require a human operator. The difficulty lies in manipulating a soft material. Instead of using a generic robot finger, we use a horizontal dual-arm manipulator with rollers as end effectors, which always hold down the cloth during manipulation. We validated our system through experiments with a prototype that can feed and sew a single piece of cloth in a single operation.
|
|
10:20-10:40, Paper ThAT10.2 | |
>Measuring the Overall Complexity of Graphical and Textual IEC 61131-3 Control Software |
|
Fischer, Juliane | Technical University of Munich |
Vogel-Heuser, Birgit | Technical University Munich |
Schneider, Heiko | Technical University of Munich |
Langer, Nikolai | Brückner Maschinenbau GmbH & Co. KG |
Felger, Markus | Teamtechnik Maschinen Und Anlagen GmbH |
Bengel, Matthias | Teamtechnik Maschinen Und Anlagen GmbH |
Keywords: Factory Automation, Control Architectures and Programming
Abstract: Software implements a significant proportion of functionality in factory automation. Thus, efficient development and the reuse of software parts, so-called units, enhance competitiveness. Thereby, complex control software units are more difficult to understand, leading to increased development, testing and maintenance costs. However, measuring complexity is challenging due to many different, subjective views on the topic. This paper compares different complexity definitions from literature and considers with a qualitative questionnaire study the complexity perception of domain experts, who confirm the importance of objective measures to compare complexity. The paper proposes a set of metrics that measure various classes of software complexity to identify the most complex software units as a prerequisite for refactoring. The metrics include complexity caused by size, data structure, control flow, information flow and lexical structure. Unlike most literature approaches, the metrics are compliant with graphical and textual languages from the IEC 61131-3 standard. Further, a concept for interpreting the metric results is presented. A comprehensive evaluation with industrial software from two German plant manufacturers validates the metrics' suitability to measure complexity.
|
|
10:40-11:00, Paper ThAT10.3 | |
>OPC UA in Support of Distributed Intelligence for a Smart Factory |
|
Xu, Xun | University of Auckland |
Keywords: Factory Automation, Intelligent and Flexible Manufacturing, Cyber-physical Production Systems and Industry 4.0
Abstract: Smart factories in the context of Industry 4.0 reply on a decentralised control strategy to be supported by distributed intelligence. The key to such a feature is an extensible and flexible communication architecture that enables semantic machine-to-machine communications. This research concerns with an exploratory study of OPC Unified Architecture (OPC UA) as an enabling protocol for the realisation of decentralised control in a smart factory. Various OPC UA communication regimes are discusses, including Request-Response (client-server) and Publish-Subscribe models, and the concepts of aggregated servers, server-to-server, client-to-client.
|
|
11:00-11:20, Paper ThAT10.4 | |
>Tracking & Tracing of Metal Parts in Manufacturing Using UHF RFID As Backbone for Industry 4.0 |
|
Kim, Minjung | The University of Auckland |
Anithottam, Jacob | University of Auckland |
Polzer, Jan | The University of Auckland |
Keywords: Factory Automation, Intelligent Transportation Systems, Logistics
Abstract: A reliable product tracking and tracing system is the backbone of Industry 4.0. In addition, tracking and tracing of products has many advantages like the access to complete and updated information in real time, the elimination of paper support/ transcript errors and the immediate visibility of finished workings and stocks. However, many tracking and tracing technologies are suffering due to harsh environments such as high-temperature variations, high humidity, excessive mechanical stresses, and exposure to aggressive chemicals. These conditions are commonly found in manufacturing of metal parts. To overcome such conditions and achieve satisfying tracking and tracing of metal parts in manufacturing, Radio Frequency Identification (RFID) is emerging as a promising solution although there is a considerable challenge in using RFID on metals. As Ultra High Frequency (UHF) RFID is found to be applicable to metals, experimental tests in several, realistic scenarios are conducted to find suitable tags out of 32 off-the-shelf on-metal UHF RFID tags. The paper includes a recommendation for commercially available on-metal RFID tags for use in tracking and tracing metals based on the test results.
|
|
11:20-11:40, Paper ThAT10.5 | |
>Equational Model Guided by Real-Time Sensor Data to Monitor Industrial Robots |
|
Osmani, Aomar | Laboratoire D'informatique De Paris Nord |
Alizadeh, Pegah | Léonard De Vinci Pôle Universitaire, Research Center, 92 916 Par |
Rodrigues, Christophe | Léonard De Vinci Pôle Universitaire, Research Center, 92 916 Par |
Keywords: Factory Automation, Intelligent and Flexible Manufacturing, Industrial Robots
Abstract: The monitoring of industrial robots is often ensured by generic simulators which model the equational aspect of the target machines. We propose an original approach to complete the equational simulator of a milling machine using the accumulated data from the used sensors. This approach creates a specific simulator for each machining situation by taking the triplet (material, cutting tool, workpiece) into account. This improvement brings great added value to the industrial experts and improves the efficiency of industrial robots. It allows them to better follow and interpret the behavior of machines during the milling process. In addition to correct the simulator using real data, our method detects also the anomalies during the real manufacturing performance and fixes the minor bugs along the observed real data during its continuous simulation mimicry. The additional interest of our model remains the precise definition of the complementary model between the real system and the equational simulator. This makes it possible, by using an inductive approach to search for regularities in the model in order to better interpret the structural differences between the model and the system and to better understand the situations linked to their functionalities or undesirable situations. The intensive experiments on real data validate our model and open up many perspectives for future works.
|
|
11:40-12:00, Paper ThAT10.6 | |
>DeltaCharger: Charging Robot with Inverted Delta Mechanism and CNN-Driven High Fidelity Tactile Perception for Precise 3D Positioning |
> Video
|
|
Okunevich, Iaroslav | Skolkovo Institute of Science and Technology |
Trinitatova, Daria | Skolkovo Institute of Science and Technology |
Kopanev, Pavel | Skolkovo Institute of Science and Technology |
Tsetserukou, Dzmitry | Skolkovo Institute of Science and Technology |
Keywords: Automation Technologies for Smart Cities, Deep Learning Methods, Factory Automation
Abstract: DeltaCharger is a novel charging robot with an Inverted Delta structure for 3D positioning of electrodes to achieve robust and safe transferring energy between two mobile robots. The embedded high-fidelity tactile sensors allow to estimate the angular, vertical and horizontal misalignments between electrodes on the charger mechanism and electrodes on the target robot using pressure data on the contact surfaces. This is crucial for preventing short circuit. In this paper, the mechanism of the developed prototype and evaluation study of different machine learning models for misalignment prediction are presented. The experimental results showed that the proposed system can measure the angle, vertical and horizontal values of misalignment from pressure data with an accuracy of 95.46%, 98.2%, and 86.9%, respectively, using a Convolutional Neural Network (CNN). DeltaCharger can potentially bring a new level of charging systems and improve the prevalence of mobile autonomous robots.
|
|
ThAT11 Regular Session, St Clair 4 |
|
Motion and Path Planning 4 |
Chair: Clever, Debora | ABB Corporate Research Center |
Co-Chair: Wang, Qianqian | The Chinese University of Hong Kong |
|
10:00-10:20, Paper ThAT11.1 | |
>Multi-Robot Online Terrain Coverage under Communication Range Restrictions – an Empirical Study |
> Video
|
|
Gautam, Avinash | Birla Institute of Technology and Science |
Soni, Ankit | BITS Pilani |
Shekhawat, Virendra Singh | BITS Pilani |
Mohan, Sudeept | Birla Institute of Technology and Science |
Keywords: Robot Networks, Motion and Path Planning, Autonomous Agents
Abstract: Communication in a multi-robot system is vital as it facilitates coordination. The performance of a multi-robot system improves with coordination. Many state-of-the-art approaches ignore intermittent connectivity, which is inevitable due to communication range restrictions. In this paper, the assumption of global communication is dropped, and the robots are restricted to communicate in a pre-specified communication range as in a realistic scenario. A comparative empirical study of five different state-of-the-art approaches which assume that the communication is omnipresent is conducted. The performance of each algorithm is evaluated by varying the communication range with a different sized robot team both in simulation and on a physical multi-robot test-bed. Finally, the impact of communication range restrictions on the performance of the approaches under evaluation is discussed.
|
|
10:20-10:40, Paper ThAT11.2 | |
>Micromanipulation Using Reconfigurable Self-Assembled Magnetic Droplets with Needle Guidance |
|
Wang, Qianqian | The Chinese University of Hong Kong |
Yang, Lidong | The Chinese University of Hong Kong |
Zhang, Li | The Chinese University of Hong Kong |
Keywords: Manipulation Planning, Motion and Path Planning, Optimization and Optimal Control
Abstract: Dynamic self-assembly is a promising approach for inducing the collective behavior of agents to perform coordinated tasks at small scales. However, efficient pattern formation and navigation in environments with complex conditions remain a challenge. In this article, we propose a strategy for micromanipulation using dynamically self-assembled magnetic droplets with needle guidance. An iron needle was controlled by a three-degree-of-freedom (3-DoF) manipulator and magnetized by precessing magnetic fields. The process of self-assembly was optimized based on real-time vision feedback and a genetic algorithm. Affected by the locally induced field gradient near the needle, reconfigurable assembled magnetic droplets were formed beneath the air-liquid interface with high time efficiency, and the geometric center of the pattern was determined. Following the magnetized needle, assembled patterns were navigated along preplanned paths and exhibited reversible pattern expansion and shrinkage. Moreover, cargo can be trapped and caged by exploiting the induced fluid flow around the assembled droplets. To perform cargo transportation tasks in a multiple-obstacle environment, an optimal path planner with obstacle-avoidance capability was designed based on the particle swarm optimization (PSO) algorithm. Experiments demonstrated effective pattern formation, navigation, cargo trapping, and obstacle-avoidance transportation. The proposed method opens new prospects of using a dynamically self-assembled pattern as an untethered end-effector for micromanipulation.
|
|
10:40-11:00, Paper ThAT11.3 | |
>SLInKi: State Lattice Based Inverse Kinematics -- a Fast, Accurate, and Flexible IK Solver for Soft Continuum Robot Manipulators |
> Video
|
|
Chiang, Shou-Shan | Worcester Polytechnic Institute |
Yang, Hao | Worcester Polytechnic Institute |
Skorina, Erik | Worcester Polytechnic Institute |
Onal, Cagdas | WPI |
Keywords: Compliant Joints and Mechanisms, Motion and Path Planning, Foundations of Automation
Abstract: Soft continuum robots offer unique properties that cannot be achieved using rigid linkage based robot manipulators. Their dexterity and intrinsic compliance deliver the ability to navigate constrained environments and operate in unprecedented ways. Although Jacobian velocity matrix based method is a widely used approach to solve inverse kinematics (IK) problems for traditional rigid robots, the drawbacks of this method emerge obviously while solving IK problems of continuum robots, such as high computational cost with no solution guarantees. Attempts to provide alternative solutions suffer from limitations due to the computational complexity and vast functional workspace of continuum manipulator postures. Here, we propose a heuristic approach, State Lattice based Inverse Kinematics Solver (SLInKi), which is inspired by concepts originally developed for solving path-finding problems to solve the IK problem of a soft continuum robot. The algorithm implementation is intuitive, runs in real time, and combines the strengths of two algorithms in a unique package that surpasses existing methods in adjustability and efficiency. Several simulation case studies and real robot experiments demonstrate that the proposed approach is flexible, computationally efficient, and highly accurate as compared to the state of the art.
|
|
11:00-11:20, Paper ThAT11.4 | |
>Building Skill Learning Systems for Robotics |
> Video
|
|
Lutter, Michael | Technische Universitaet Darmstadt |
Clever, Debora | ABB Corporate Research Center |
Kirsten, René | ABB Corporate Research |
Listmann, Kim Daniel | ABB Schweiz AG |
Peters, Jan | Technische Universität Darmstadt |
Keywords: Learning and Adaptive Systems, Motion and Path Planning, Assembly
Abstract: Skill-generating policies have enabled robots to perform a wide range of applications as for example assembly tasks. However, the manual engineering effort for such policies is fairly high and the environment is frequently required to be rather deterministic. For expanding robot deployment to low-volume manufacturing two challenges need to be addressed. First, the robot should acquire the skill-generating policy not from a robot programmer but rather from an expert on the task and second, the robot needs to be able to operate in unstructured environments. In this paper we present a learning approach that combines imitation learning and reinforcement learning to provide a tool for intuitive task teaching followed by self-optimization of the system. The presented approach is applied to a dual-arm assembly task using a real robot and appropriate simulation models. Whereas pure imitation learning does not result in an acceptable success rate for the considered example, after 400 episodes of reinforcement learning the robot can successfully solve the assembly task.
|
|
ThIPT6 Special Session, St Clair 1 |
|
Data-Driven Modeling, Analysis and Control of Advanced Manufacturing
Systems |
Chair: Pei, Zhi | Zhejiang University of Technology |
Co-Chair: Gou, Yi-Xing | Northwestern Polytechnical University |
Organizer: Wang, Jun-Qiang | Northwestern Polytechnical University |
Organizer: Li, Yang | Tongji Univeristy |
Organizer: Yan, Chao-Bo | Xi'an Jiaotong University |
Organizer: Pei, Zhi | Zhejiang University of Technology |
Organizer: Jia, Zhiyang | Beijing Institute of Technology |
Organizer: Zhang, Ding | Guangdong University of Technology |
Organizer: Gou, Yi-Xing | Northwestern Polytechnical University |
|
13:30-13:50, Paper ThIPT6.1 | |
>Parameter Identification for Synchronous Two-Machine Exponential Production Line Model (I) |
|
Sun, Yuting | University of Connecticut |
Zhang, Liang | University of Connecticut |
Keywords: Intelligent and Flexible Manufacturing, Manufacturing, Maintenance and Supply Chains, Sustainable Production and Service Automation
Abstract: Production system modeling refers to the process of constructing valid and high-fidelity mathematical models that are capable of capturing the behavior of job flow in the manufacturing systems. During the modeling process, model parameter identification is the most critical step. This step, however, often involves a significant amount of complex and non-standardized work. To tackle this problem, we propose to reversely compute the production system model parameters based on standard manufacturing system performance metrics. In this paper, we consider a two-machine production line model, in which the machines follow the exponential reliability model and have identical processing speed, and formulate a constrained optimization problem with the objective of finding the optimal machine parameters which can fit the system performance metrics the best. To solve this problem, barrier method with BFGS quasi-Newton algorithm and cyclic coordinate descent method with proximal point update are developed. The accuracy of these two methods in estimating machine parameters and performance metrics are computed and compared through extensive numerical experiments. Although barrier method is much more efficient in terms of computation time, the risk of getting trapped in local optima exists due to the lack of convexity. On the other hand, the numerical experiments demonstrate that the coordinate descent method reaches the global optimal solution for all the cases. Therefore, an ensemble strategy is recommended to ensure a high accuracy in parameter estimation with acceptable computation time.
|
|
13:50-14:10, Paper ThIPT6.2 | |
>Event-Based Modeling and Analysis of Serial Production Lines with Quality Scrap (I) |
|
Gou, Yi-Xing | Northwestern Polytechnical University |
Wang, Jun-Qiang | Northwestern Polytechnical University |
Li, Yang | Tongji Univeristy |
Keywords: Discrete Event Dynamic Automation Systems, Optimization and Optimal Control, Process Control
Abstract: Quality inspection plays an important role in manufacturing systems as an effective measurement to ensure high product quality and throughput. In this research, we study serial production lines with quality scrap, in which defective parts are scrapped once identified by quality inspection machines. An event-based model is established to quantitatively analyze the impact of machine breakdown and quality failure events on throughput. Based on the model, an optimization method is proposed to optimize the allocation of quality inspection machines such that the overall system cost, including the production loss cost and quality inspection cost, can be minimized. Numerical studies are performed to validate the effectiveness of the proposed method.
|
|
14:10-14:30, Paper ThIPT6.3 | |
>Intelligent Fault Diagnosis for Large-Scale Rotating Machines Using Binarized Deep Neural Networks and Random Forests |
|
Li, Huifang | Beijing Institute of Technology |
Hu, Guangzheng | Beijing Institute of Technology |
Li, Jianqiang | Beijing University of Technology |
Zhou, MengChu | New Jersey Institute of Technology |
Keywords: Diagnosis and Prognostics, Cloud Computing For Automation, Machine learning
Abstract: Recently, deep neural network (DNN) models work incredibly well, and edge computing has achieved great success in real-world scenarios, such as fault diagnosis for large-scale rotational machinery. However, DNN training takes a long time due to its complex calculation, which makes it difficult to optimize and retrain models. To address such an issue, this work proposes a novel fault diagnosis model by combining binarized DNNs (BDNNs) with improved random forests (RFs). First, a BDNN-based feature extraction method with binary weights and activations is designed to reduce the model runtime without losing the accuracy in extracting features. Its generated features are used to train an RF based fault classifier to relieve the information loss caused by binarization. Second, to compensate for the possible classification accuracy reduction resulting from the similar binarized features of two instances with different labels, we replace a Gini index with ReliefF in training RFs to further enhance the separability of features extracted by BDNN. Third, an edge computing-based fault diagnosis paradigm is proposed to increase diagnostic efficiency, where our diagnosis model is deployed distributedly on a number of edge nodes close to the end rotational machines in distinct locations. Extensive experiments are conducted to validate the proposed method on the data sets from rolling element bearings, and the results demonstrate that, in almost all cases, its diagnostic accuracy is competitive to the state-of-the-art DNNs and even higher due to a form of regularization in some cases. Benefited from the relatively lower computing and storage requirements of BDNNs, it is easy to be deployed on edge nodes to realize real-time fault diagnosis concurrently.
|
|
14:30-14:50, Paper ThIPT6.4 | |
>Urban Drone Delivery within an On-Demand Platform (I) |
|
Fang, Tao | Zhejiang University of Technology |
Pei, Zhi | Zhejiang University of Technology |
Keywords: Automation Technologies for Smart Cities, Intelligent Transportation Systems, Logistics
Abstract: The application of autonomous unmanned aerial mobility in urban on-demand delivery is a very challenging endeavor. In this study, a mixed integer programming model is proposed based on the realistic UAV delivery operation scenario. Different from the traditional pickup and delivery formulations, this model not only considers the time window constraints, but also considers the spatial conflicts incurred by the practical operation of the UAV. To further strengthen the model, a series of valid inequalities are applied to the model. Then a branch and cut algorithm is proposed to solve this problem. To further accelerate the solving process, a greedy insertion heuristic method is designed to obtain an initial upper bound. Experimental results show that the algorithm proposed in this paper can be used to solve real life on-demand drone delivery problem within reasonable computing time.
|
|
14:50-15:10, Paper ThIPT6.5 | |
>MPI-Based System 2 for Determining LPBF Process Control Thresholds and Parameters |
|
Adnan, Muhammad | National Cheng Kung University, Institute of Manufacturing Infor |
Yang, Haw-Ching | National Kaohsiung Univ. of Sci. and Tech |
Kuo, Tsung-Han | National Cheng Kung University |
Cheng, Fan-Tien | National Cheng Kung University |
Tran, Hong-Chuong | National Cheng Kung University |
Keywords: Control Architectures and Programming, Intelligent and Flexible Manufacturing, AI-Based Methods
Abstract: Determining thresholds of the primary control loops (System 1) of an additive manufacturing (AM) process is challenging when realizing System 1 with fast and intuitive capability for adapting to different metal powers, machine configurations, and process parameters. This paper presents a secondary tuning loop (System 2) based on the convolution neural network (CNN) and long short-term memory (LSTM) to classify the types of melt-pool images (MPIs) from a coaxial camera online, suggest re-melting parameters, and determine the control thresholds of System 1 offline. Case studies indicate that the thresholds and parameters of System 1 including air flowing, powder coating, and laser printing of control loops of a laser powder bed fusion (LPBF) machine can be more deliberatively and logically decided by the proposed MPI-based System 2.
|
|
ThIPT7 Regular Session, St Clair 2 |
|
Collision Avoidance |
Chair: Zhang, Zengjie | Technical University of Munich |
Co-Chair: Mbakop, Steeve | Yncrea Hauts-De-France |
|
13:30-13:50, Paper ThIPT7.1 | |
>Safe Vessel Navigation Visually Aided by Autonomous UAV Systems |
> Video
|
|
le Fevre Sejersen, Jonas | Aarhus University |
Pimentel de Figueiredo, Rui | Aarhus University |
Kayacan, Erdal | Aarhus University |
Keywords: Collision Avoidance, Industrial and Service Robotics, Sensor Fusion
Abstract: In the maritime sector, safe vessel navigation is of great importance, particularly in congested harbors and waterways. The focus of this work is to estimate the distance between an object of interest and potential obstacles using a companion UAV. The proposed approach fuses GPS data with long-range aerial images. First, we employ semantic segmentation DNN for discriminating the vessel of interest, water, and potential solid objects using raw image data. The network is trained with both real and images generated and automatically labeled from a realistic AirSim simulation environment. Then, the distances between the extracted vessel and non-water obstacle blobs are computed using a novel GSD estimation algorithm. To the best of our knowledge, this work is the first attempt to detect and estimate distances to unknown objects from long-range visual data captured with conventional RGB cameras and auxiliary absolute positioning systems (e.g. GPS). The simulation results illustrate the accuracy and efficacy of the proposed method for visually aided navigation of vessels assisted by UAV.
|
|
13:50-14:10, Paper ThIPT7.2 | |
>Adaptive Speed Collision Avoidance for Dynamic Environments |
|
Dutta, Sourav | University at Albany |
Tran, Tuan | University at Albany |
Rekabdar, Banafsheh | Southern Illinois University Carbondale |
Ekenna, Chinwe | University at Albany |
Keywords: Collision Avoidance, Motion and Path Planning, Planning, Scheduling and Coordination
Abstract: This paper presents a dynamic collision avoidance framework using predictive approaches. This predictive model exploits temporal and spatial characteristics of dynamic obstacles to predict future events thus avoiding collisions. To achieve this we implement a novel combination of two Poisson processes with independent outcomes and test this on a series of dynamic events with a crazyflie drone. Our approach initiates a speed adjustment when needed, which is a natural human response to avoid collisions with obstacles in crowded and congested environments. Our results in simulation and real-world tests indicate that our method removes the need for replanning and reduces the cost of time which is a huge bottleneck in trajectory planning. Our method outperforms baseline algorithms in terms of well-accepted metrics like the number of collisions, percentage accuracy, and execution time.
|
|
14:10-14:30, Paper ThIPT7.3 | |
>Curve-Based Approach for Shape Reconstruction and Planning of a Mobile-Continuum Manipulator in Structured Environment |
> Video
|
|
Singh, Inderjeet | CRIStAL, CNRS UMR 9189, University of Lille1 |
Mbakop, Steeve | Yncrea Hauts-De-France |
Singh, Manarshhjot | Polytech Lille |
Benskrane, Ismail | University of Lille |
Merzouki, Rochdi | CRIStAL, CNRS UMR 9189, University of Lille1 |
Keywords: Collision Avoidance, Manipulation Planning, Modelling, Simulation and Optimization in Healthcare
Abstract: Continuum manipulators are a popular research topic due to their use in a wide variety of fields including medical, military and exploration. However, a mobile-continuum manipulator can have a higher dexterity and therefore even higher opportunities for application. This paper presents a unified geometric modeling approach for path planning and shape reconstruction of a mobile-continuum manipulator. For both components of the considered robotic system, namely: the shape for the continuum manipulator; the path planning for the reconstructed shape and the mobile base, Pythagorean Hodograph (PH) curves are proposed as a common modeling tool. The PH curves are optimized under various conditions, so as to be applicable for various tasks like path planning, shape reconstruction, and obstacle avoidance. The proposed unified modeling is tested on a mobile-continuum manipulator 'RobotinoXT' consisting of a 'Compact Bionic Handling Assistant (CBHA)' continuum manipulator mounted on a 'Robotino' mobile base.
|
|
14:30-14:50, Paper ThIPT7.4 | |
>Designing Safe Lane-Change Maneuvers Using an Explicit Path Planner |
|
Sharma, Yash Raj | Indian Institute of Science |
Ratnoo, Ashwini | Indian Institute of Science |
Keywords: Intelligent Transportation Systems, Motion and Path Planning, Collision Avoidance
Abstract: This paper addresses the problem of designing safe lane changing maneuvers for an autonomous vehicle. In the presence of neighboring vehicles, the work considers all collision possibilities in the scenario. As an explicit path planner, the four parameter logistic (4PL) curve is proposed as the solution. Analytic conditions for collision avoidance with respect to the neighboring vehicles are deduced in closed-form, and are expressed in the 2 dimensional design parameter space of 4PL curve. Feasible lane changing maneuvers are represented in this space as a set of points satisfying collision avoidance, and reachability constraints. Case studies are presented for various scenarios. The proposed method offers a deterministic, and computationally efficient approach for generating multiple feasible lane changing trajectories.
|
|
14:50-15:10, Paper ThIPT7.5 | |
>An Online Robot Collision Detection and Identification Scheme by Supervised Learning and Bayesian Decision Theory |
|
Zhang, Zengjie | Technical University of Munich |
Qian, Kun | The University of Tokyo |
Schuller, Björn | University of Augsburg |
Wollherr, Dirk | Technische Universität München |
Keywords: Collision Avoidance, Failure Detection and Recovery, Machine learning
Abstract: This paper is dedicated to developing an online collision detection and identification scheme for human-collaborative robots. The scheme is composed of a signal classifier and an online diagnosor, which monitors the sensory signals of the robot system, detects the occurrence of a physical human-robot interaction and identifies its type within a short period. In the beginning, we conduct an experiment to construct a data set that contains the segmented physical interaction signals with ground truth. Then, we develop the signal classifier on the data set with the paradigm of supervised learning. To adapt the classifier to the online application with requirements on response time, an auxiliary online diagnosor is designed using Bayesian decision theory. The diagnosor provides not only a collision identification result but also a confidence index which represents the reliability of the result. Compared to the previous works, the proposed scheme ensures rapid and accurate collision detection and identification even in the early stage of a physical interaction. As a result, safety mechanisms can be triggered before further injuries are caused, which is quite valuable and important towards a safe human-robot collaboration. In the end, the proposed scheme is validated on a robot manipulator and applied to a demonstration task with collision reaction strategies. The experimental results reveal that the collisions are detected and classified within 20 ms with an overall accuracy of 99.6%, which confirms the applicability of the scheme to collaborative robots in practice.
|
|
ThIPT8 Regular Session, Rhone 4 |
|
Human Factors and Human-In-The-Loop |
Chair: Yi, Jingang | Rutgers University |
Co-Chair: Wang, Yue | Clemson University |
|
13:30-13:50, Paper ThIPT8.1 | |
>Repairing Human Trust by Promptly Correcting Robot Mistakes with an Attention Transfer Model |
|
Luo, Ruijiao | Cognitive Robotics and AI Lab (CRAI), Kent State University |
Huang, Chao | Kent State University |
Peng, Yuntao | Cognitive Robotics and AI Lab (CRAI), Kent State University |
Song, Boyi | Cognitive Robotics and AI Lab (CRAI), College of Aeronautics And |
Liu, Rui | Kent State University |
Keywords: Failure Detection and Recovery, Human Factors and Human-in-the-Loop, Motion and Path Planning
Abstract: In human-robot collaboration (HRC), human trust in the robot is the human expectation that a robot executes tasks with desired performance. A higher-level trust increases the willingness of a human operator to assign tasks, share plans, and reduce the interruption during robot executions, thereby facilitating human-robot integration both physically and mentally. However, due to real-world disturbances, robots inevitably make mistakes, decreasing human trust and further influencing collaboration. Trust is fragile and trust loss is triggered easily when robots show incapability of task executions, making the trust maintenance challenging. To maintain human trust, in this research, a trust repair framework is developed based on a human-to-robot attention transfer (textbf{textit{H2R-AT}}) model and a user trust study. The rationale of this framework is that a prompt mistake correction restores human trust. With textbf{textit{H2R-AT}}, a robot localizes human verbal concerns and makes prompt mistake corrections to avoid task failures in an early stage and to finally improve human trust. User trust study measures trust status before and after the behavior corrections to quantify the trust loss. Robot experiments were designed to cover four typical mistakes, textit{wrong action, wrong region, wrong pose, }and textit{ wrong spatial relation}, validated the accuracy of textbf{textit{H2R-AT}} in robot behavior corrections; a user trust study with 252 participants was conducted, and the changes in trust levels before and after corrections were evaluated. The effectiveness of the human trust repairing was evaluated by the mistake correction accuracy and the trust improvement.
|
|
13:50-14:10, Paper ThIPT8.2 | |
>Postural Balance of Kneeling Gaits on Inclined and Elevated Surface for Construction Workers |
|
Chen, Siyu | Rutgers University |
Yu, Yi | Rutgers University |
Di, Chong | Rutgers University |
Stevenson, Duncan | Rowan University |
Trkov, Mitja | Rowan University |
Gong, Jie | Rutgers University |
Yi, Jingang | Rutgers University |
Keywords: Human Factors and Human-in-the-Loop, Virtual Reality and Interfaces, Human-Centered Automation
Abstract: We present a human postural balance study on quiet stance and kneeling gaits on inclined and high elevated surfaces for construction workers. To simulate the high elevation, an immersive mixed reality environment is built with an actual inclined roof surface to create somatosensory haptic feedback. We quantify the postural balance during quiet kneeling and stance through measurements of the center of pressure and sway motion of the upper-body under various inclined angles and heights. The results of center of pressure and trunk acceleration measurements show smaller postural sway during kneeling compared to standing. A mathematical model is also presented to help understand the experimental results and potentially provide design guidance for further intervention to prevent and mitigate the fall risk for construction workers. The model and controller parameters are optimized to precisely capture and explain the experimental results.
|
|
14:10-14:30, Paper ThIPT8.3 | |
>Isometric-Based Approach for Detecting Localized Muscular Fatigue During Complex Dynamic Manufacturing Operations |
|
Chand, Saahil | The University of Auckland |
McDaid, Andrew | The University of Auckland |
Lu, Yuqian | The University of Auckland |
Keywords: Human Factors and Human-in-the-Loop, Human-Centered Automation, Assembly
Abstract: Unstructured manufacturing environments with flexible production procedures will continue to require worker involvement due to the difficulty of automation. In the context of human-centricity, human physical well-being during manufacturing operations, in particular, muscular fatigue, must be reliably assessed and optimized on the shop floor. However, there is a lack of accurate and reliable methods of dynamic fatigue assessment for complex worker operations within manufacturing environments. To this end, we develop a novel solution for a dynamic fatigue assessment framework to determine the fatigue impact of complex manufacturing operations. The framework utilizes a fatigue index profile that defines the fatigue development trend of an individuals' target muscle group and alongside the force-generating capacity of the muscle group under the dynamic target operations. By inducing fatigue buildup through the dynamic operations, impact the force-generating capacity, and therefore, the localized muscle fatigue can be determined through the isometric contraction phase. Results indicate that analyzing isometric contractions during complex, dynamic operations can provide a reliable fatigue impact estimation. Furthermore, this method maximizes signal uniformity by leveraging the isometric contraction's static nature while centering the fatigue buildup on the dynamic target operations. Although the proposed method was trialed using a dynamic vertical handling operation, the framework can determine the fatigue impact of more complex manufacturing operations with minimal deviation to the presented methods.
|
|
14:30-14:50, Paper ThIPT8.4 | |
>Is the Leader Robot an Adequate Sensor for Posture Estimation and Ergonomic Assessment of a Human Teleoperator? |
> Video
|
|
Yazdani, Amir | University of Utah |
Sabbagh Novin, Roya | University of Utah |
Merryweather, Andrew | University of Utah |
Hermans, Tucker | University of Utah |
Keywords: Human Factors and Human-in-the-Loop, Telerobotics and Teleoperation
Abstract: Ergonomic assessment of human posture plays a vital role in understanding work-related safety and health. Current posture estimation approaches face occlusion challenges in teleoperation and physical human-robot interaction. We investigate if the leader robot is an adequate sensor for posture estimation in teleoperation and we introduce a new probabilistic approach that relies solely on the trajectory of the leader robot for generating observations. We model the human using a redundant, partially-observable dynamical system and we infer the posture using a standard particle filter. We compare our approach with postures from a commercial motion capture system and also two least-squares optimization approaches for human inverse kinematics. The results reveal that the proposed approach successfully estimates human postures and ergonomic risk scores comparable to those estimates from gold-standard motion capture.
|
|
14:50-15:10, Paper ThIPT8.5 | |
>A Personalized Computational Model for Human-Like Automated Decision-Making (I) |
|
Jiang, Longsheng | Clemson University |
Wang, Yue | Clemson University |
Keywords: Human Factors and Human-in-the-Loop, Human-Centered Automation
Abstract: We propose a computational model for enabling robots to automatically make decisions under risk in a human-like way. Human decision-making under risk is influenced by psychological effects, including regret effects, probability weighting effects, and range effects. On the basis of regret theory, we devise a mathematical decision-making model to encompass these psychological effects. To further quantify the model, we cast the model into a state-space representation and design a fuzzy logic controller to obtain desired preference data from individual decision makers. The data from each individual were used to train a personalized instance of the model. The resulting model is quantitative. It sheds light on the psychological mechanism of risk-attitudes in human decision-making. The prediction accuracy of the model was statistically tested. On average, the accuracy of our model is 74.7%, which is significantly close to the average accuracy of the subjects when they repeated their own previously made decisions (73.3%). Furthermore, when only the decisions that were repeated consistently by the subjects are examined, the average accuracy of our model is 86.6%.
|
|
ThIPT9 Special Session, St Clair 3A |
|
Simulation Optimization in New Information Age 2 |
Chair: Peng, Yijie | Peking University |
Co-Chair: Pedrielli, Giulia | Arizona State University |
Organizer: Jia, Qing-Shan | Tsinghua University |
Organizer: Luo, Jun | Shanghai Jiao Tong University Antai College of Economics & Management |
Organizer: Pedrielli, Giulia | Arizona State University |
Organizer: Peng, Yijie | Peking University |
|
13:30-13:50, Paper ThIPT9.1 | |
>Simulation-Based Model Learning for Optimization of Building Energy Management (I) |
|
Shen, Yuanjun | Xi'an Jiaotong University |
Gao, Feng | Xi'an Jiaotong University |
Liu, Yaping | Xi'an Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Wu, Jiang | Xian Jiaotong University |
Liu, Kun | Xi'an Jiaotong University |
Keywords: Model Learning for Control, Modelling, Simulation and Validation of Cyber-physical Energy Systems
Abstract: In this paper, we present a reinforcement learning based method for building energy management, which introduces a transition-model learning from building energy simulation software. The transition-model receives state from simulation model and returns cost-to-go to control strategy. The method is intended to combat two main gaps when integrating simulation program and optimization algorithm in building energy management, which are nonanalytical and nonlinear computation for simulation program and requirement of optimization algorithms for system dynamics. The method is realized on EnergyPlus, which interact with python program in functional mock-up interface (FMI) standard. The numerical test results show that the transition-model introduced can estimate cost-to-go to a certain extent and significantly improve efficiency of control.
|
|
13:50-14:10, Paper ThIPT9.2 | |
>Efficient Cluster Sampling for Morris Method (I) |
|
Shi, Wen | Central South University |
Chen, Xi | Virginia Tech |
Keywords: Simulation and Animation
Abstract: We provide a thorough investigation of the cluster sampling scheme for Morris’ elementary effects method (MM). We first study the sampling mechanism underpinning the two sampling schemes of MM (i.e., cluster sampling and non-cluster sampling) and unveil its nature as a two-level nested sampling process. We first study the budget allocation problem for cluster sampling under the analysis of variance framework and derive optimal budget allocations for efficient estimation of the importance measures. we then devise an efficient cluster sampling algorithm with two variants to achieve enhanced statistical properties. The numerical results are promising.
|
|
14:10-14:30, Paper ThIPT9.3 | |
>An Event-Based Optimization Method for Building Evacuation with Queuing Network Model (I) |
|
Zhang, Yuandong | Xi'an Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Wu, Jiang | Xian Jiaotong University |
Guan, Xiaohong | Xi'an Jiaotong University |
Keywords: Discrete Event Dynamic Automation Systems, Motion and Path Planning, Reinforcement
Abstract: This paper presents an online evacuation policy optimization framework for escape movement of large populations through geometrically complex building spaces with active guides. Considering the uncertainty of the evacuation process, a queuing network model is introduced to formulate the evacuation problem. Evacuation policy is expressed as an online updating guide to the evacuation direction of people in each node. Then an event-based optimization method is used to improve the evacuation policy, in which actions are taken only when local congestion statuses are changed, under reinforcement learning in partially observable environments. Numerical results show that this method is effective in improving the efficiency of evacuation.
|
|
14:30-14:50, Paper ThIPT9.4 | |
>Learning-Based Safety-Critical Motion Planning with Input-To-State Barrier Certificate (I) |
|
Jin, Xinze | Tsinghua University |
Jia, Qing-Shan | Tsinghua University |
Zhang, Tao | Tsinghua University |
Xia, Huaxia | Meituan |
Keywords: Motion and Path Planning, AI-Based Methods
Abstract: Motion planning in an effective and safe manner is a critical yet challenging task for autonomous driving. Learning-based framework as a new fashion in simulation and optimization has great potentials to develop a time-efficient navigation policy. To design a controller that addresses safety with explicit formulation, we incorporate control barrier function approach to generate a constrained optimization problem. The proposed method with uncertainty analysis helps to deal with disturbance towards motion planning task. Simulation results show that the algorithm produces improvement in the learning process and the adaptation to safety and performance.
|
|
14:50-15:10, Paper ThIPT9.5 | |
>Multi-Zone Indoor Temperature Prediction Based on Graph Attention Network and Gated Recurrent Unit (I) |
|
Zhou, Chunxiang | Xi'an Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Wu, Jiang | Xian Jiaotong University |
Liu, Kun | Xi'an Jiaotong University |
Guan, Xiaohong | Xi'an Jiaotong University |
Keywords: Modelling, Simulation and Validation of Cyber-physical Energy Systems, Machine learning, Environment Monitoring and Management
Abstract: Indoor temperature have significant influence on load forecasting, comfort control and security monitoring. Achieving accurate temperature prediction can provide key basic data for energy efficiency and building safety and comfort. In the case of multiple zones, the heat transfer process in adjacent zones can have an important impact on the dynamics of indoor temperature. This paper focuses on the influence of heat transfer process in multiple adjacent zones. To describe the interactions of temperature among the multiple zones, we consider the zones as nodes and the connected walls as edges based on actual layouts to construct the graph network. For the non-linearity of the heat transfer process, we propose a novel multi-zone indoor temperature prediction model based on graph attention mechanism and recurrent network to achieve one-step ahead and multi-step ahead temperature predictions. The accuracy of this model was further verified by using data generated from EnergyPlus simulations. The best predicted result had an RMSE value of 0.47, an MAE value of 0.37, and an R Squared value of 0.94.
|
|
15:10-15:30, Paper ThIPT9.6 | |
>Gradient-Based Simulation Optimization for Economic Design of Control Charts (I) |
|
Zhang, Gongbo | Peking Univesity |
Peng, Yijie | Peking University |
Yang, Shuhuai | Peking Univesity |
Keywords: Simulation and Animation, Optimization and Optimal Control, Process Control
Abstract: We propose a gradient-based simulation optimization approach for economic design of control charts. A generalized likelihood ratio method is applied to estimate the the gradient. Two stochastic approximation algorithms with increasing sample size in iterations and randomized sample size are developed to determine an optimal upper control limit for exponentially weighted moving average control chart. Numerical results show that the proposed method is an effective approach for economic design of control charts.
|
|
ThIPT10 Regular Session, St Clair 3B |
|
Machine Learning 1 |
Chair: Topp, Elin Anna | Lund University - LTH |
Co-Chair: Moretti, Emilio | Politecnico Di Milano |
|
13:30-13:50, Paper ThIPT10.1 | |
>Socially Compliant Navigation in Indoor Corridors Based on Reinforcement Learning |
> Video
|
|
Li, Chih-Hung G. | National Taipei University of Technology |
Chang, Yu-Hsiang | National Taipei University of Technology |
Keywords: Reinforcement, Machine learning, Autonomous Vehicle Navigation
Abstract: In this paper, the policy function of a mobile robot navigating in indoor corridor environments was obtained through reinforcement learning (RL). Assuming the scenario where right-passing rules are enforced, target paths associated with different corridor widths were defined; the robot’s actions of navigating into the target paths were defined as RL rewards to encourage the common social consensus regarding corridor passing. Specifically, the robot was trained to render proper reactions according to the width of the corridor and the robot’s speed, pose, and relative position in the corridor. The RL models were trained in Gazebo and ROS; the effectiveness of the navigation policy was validated by various tests of different conditions. It was found that different speeds need different strategies; the RL models trained for each specific speed category appear to be optimal. Such results were supported by the cross-examinations on the success rate and the number of corrective actions.
|
|
13:50-14:10, Paper ThIPT10.2 | |
>Supervised Learning Based Observer for In-Process Tool Offset Estimation in Robotic Arc Welding Applications |
|
Schmidt, Alexander | University of Stuttgart |
Kotschote, Christian | AUDI AG |
Riedel, Oliver | University of Stuttgart |
Keywords: Machine learning, Model Learning for Control, Sensor Fusion
Abstract: Workpiece tolerances in manufacturing welding applications can lead to a deviation of the welding tool from the workpiece. Such a tool offset leads to reduced welding quality. This problem can be solved by measuring the exact workpiece geometry and orientation in advance of the welding process. However, measuring the workpiece geometry for each workpiece increases the manufacturing time. Therefore, this work presents a novel approach for an in-process tool offset observer. The observer model is retrieved via supervised learning methods based on real experimental welding data. The methods for extracting features from time-series data are described. A benchmark for multiple supervised learning methods and sensor types is presented. The accuracy of the trained models is tested by welding experiments. The significance of this paper is the demonstration of the feasibility of in-process tool offset estimation for robotic arc welding applications.
|
|
14:10-14:30, Paper ThIPT10.3 | |
>Function-On-Function Regression for Trajectory Prediction of Small-Scale Particles towards Next-Generation Neuromorphic Computing |
|
Joseph, Antony | Binghamton University |
Wu, Juan | Binghamton University |
Yu, Kaiyan | Binghamton University |
Jiang, Lan | The State University of New York at Binghamton |
Cady, Nathaniel | SUNY Polytechnic Institute |
Si, Bing | State University of New York at Binghamton |
Keywords: Machine learning, Model Learning for Control, Automation at Micro-Nano Scales
Abstract: Precise and efficient motion prediction and manipulation of micro- and nanoparticles in a complex fluid suspension system under external electric fields has the potential to revolutionize the manufacture of scalable functional nanodevices. However, the physical motion model of the particle based on physical simulation does not consider the effects in the complex fluid suspension system, e.g., boundary conditions, fluid motion, and particle interactions, and often results in imperfect prediction of particle trajectories under the coupled global field. This study proposes a data-driven approach for small-scale particle trajectory prediction by leveraging both physical simulation model and experimental data. Historical function-on-function regression is used to predict experimental trajectories from corresponding simulation trajectories. A gradient boosting algorithm is used for model estimation. Our study is the first-of-its-kind that uses historical function-on-function regression to demonstrate the efficacy of predicting experimental trajectories from simulation trajectories in small-scale particle manipulation under electrical fields, which eventually leads to the design of new automated processes for efficient and smart manufacturing of functional nanodevices towards next-generation neuromorphic computing.
|
|
14:30-14:50, Paper ThIPT10.4 | |
>Selecting Part Feeding Policies with a Combined Optimization-Machine Learning Approach |
|
Moretti, Emilio | Politecnico Di Milano |
Tappia, Elena | Politecnico Di Milano |
Melacini, Marco | Politecnico Di Milano |
Limère, Veronique | Ghent University |
Keywords: Assembly, Machine learning, AI-Based Methods
Abstract: IN SEVERAL INDUSTRIES, INCREASING ATTENTION IS BEING DEVOTED TO THE DESIGN AND MANAGEMENT OF PART FEEDING SYSTEMS. THIS PAPER APPLIES A COMBINED OPTIMIZATION-MACHINE LEARNING (ML) APPROACH FOR PART FEEDING POLICIES SELECTION TO THE CASE OF A TRUCK ASSEMBLY PLANT. ACCORDING TO THIS APPROACH, FEEDING POLICIES ARE SELECTED THROUGH A ML MODEL, TRAINED USING THE OUTPUT OF AN OPTIMIZATION MODEL PREVIOUSLY APPLIED TO A SAMPLE OF PARTS. RESULTS SHOW THAT THIS APPROACH LEADS TO RESULTS CLOSE TO THE OPTIMAL ONES, AS THE DEVELOPED ML MODELS ARE ABLE TO ESTIMATE THE OPTIMAL POLICIES FOR MOST OF THE PARTS.
|
|
14:50-15:10, Paper ThIPT10.5 | |
>Realeasy: Real-Time Capable Simulation to Reality Domain Adaptation |
> Video
|
|
Dürr, Alexander | Lund University |
Neric, Liam | Lund University |
Krueger, Volker | Lund University |
Topp, Elin Anna | Lund University - LTH |
Keywords: Simulation and Animation, Machine learning, Model Learning for Control
Abstract: We address the problem of insufficient quality of robot simulators to produce precise sensor readings for joint positions, velocities and torques. Realistic simulations of sensor readings are particularly important for real time robot control laws and for data intensive Reinforcement Learning of robot movements in simulation. We systematically construct two architectures based on Long Short-Term Memory to model the difference between simulated and real sensor readings for online and offline application. Our solution is easy to integrate into existing Robot Operating System frameworks and its formulation is neither robot nor task specific. The collected data set, the plug-and-play Realeasy model for the Panda robot and a reproducible real-time docker setup are shared alongside the code. We demonstrate robust behavior and transferability of the learned model between individual Franka Emika Panda robots. Our experiments show a reduction in torque mean squared error of at least one order of magnitude.
|
|
ThIPT11 Regular Session, St Clair 4 |
|
Cloud Computing for Automation |
Chair: Roy, Daniel | Ecole Nationale D'ingénieurs De Metz |
Co-Chair: López, Alejandro | University of the Basque Country (UPV/EHU) |
|
13:30-13:50, Paper ThIPT11.1 | |
>Cloud Architecture Based Multi-Agent System for a Resources Sharing Application Platform |
|
Liu, Shiming | University of Lorraine |
Hennequin, Sophie | ENIM |
Roy, Daniel | Ecole Nationale D'ingénieurs De Metz |
Keywords: Agent-Based Systems, Cloud Computing For Automation, Sensor-based Control
Abstract: This paper presents a part of an application platform dedicated to resources sharing between several enterprises. The physical resources are administered by an Industrial Internet of Things platform (IIoT) and a consortium blockchain platform. The interactions between enterprises and physical resources are controlled with the help of a multi-agent system. The blockchain platform allows decentralizing activities and management whereas the cloud architecture-based multi-agent system permits to centralize the management of the resources sharing application platform. In this paper, we describe all tools and more specifically the multi-agent system with the chosen agents and the matching process of resources sharing. We also explain the links between all tools and the functioning of our proposed resources sharing application platform (complete architecture and data exchange).
|
|
13:50-14:10, Paper ThIPT11.2 | |
>Leveraging Edge Computing and Differential Privacy to Securely Enable Industrial Cloud Collaboration Along the Value Chain |
|
Giehl, Alexander | Fraunhofer AISEC |
Heinl, Michael P. | Fraunhofer AISEC |
Busch, Maximilian | Technical University of Munich |
Keywords: Cloud Computing For Automation, Optimization and Optimal Control, Big-Data and Data Mining
Abstract: Big data continues to grow in the manufacturing domain due to increasing interconnectivity on the shop floor in the course of the fourth industrial revolution. The optimization of machines based on either real-time or historical machine data provides benefits to both machine producers and operators. In order to be able to make use of these opportunities, it is necessary to access the machine data, which can include sensitive information such as intellectual property. Employing the use case of machine tools, this papers presents a solution enabling industrial data sharing and cloud collaboration while protecting sensitive information. It employs the edge computing paradigm to apply differential privacy to machine data in order to protect sensitive information and simultaneously allow machine producers to perform the necessary calculations and analyses using this data.
|
|
14:10-14:30, Paper ThIPT11.3 | |
>Predictive Offloading in Fog Manufacturing for Computational Pipelines Using Multi-Task Learning |
|
Nallendran, Vignesh Raja | Virginia Tech |
Wang, Lening | Virginia Tech |
Jin, Ran | Virginia Tech |
Keywords: Machine learning, Cloud Computing For Automation, Data fusion
Abstract: In smart manufacturing, it is significant to integrate the computation service with the manufacturing process to support real-time process controls and data analytics. A suitable computing architecture to handle the influx of data generated from the manufacturing process is Fog manufacturing. In Fog manufacturing, the Fog-cloud collaborative architecture is enabled through a distributed computing platform to facilitate responsive, scalable, and reliable data analysis in manufacturing networks. However, effective utilization of the Fog-cloud computing service requires optimal offloading strategies due to limited computational and bandwidth resources in Fog manufacturing. Therefore, a predictive offloading method that can properly deploy each computation task based on the predicted run-time metrics (e.g., time-latency) is desired. However, the run-time metrics collected in Fog manufacturing are heterogeneous in nature and cannot be modeled through conventional predictive analysis. This is because the computational flow and the data sources vary among different Fog nodes. To overcome this issue, in this paper, a multi-task learning model based predictive offloading method is proposed to assign the computation tasks based on their predicted run-time metrics in Fog manufacturing. The proposed method is evaluated on a Fog manufacturing testbed. The results show that the predictive offloading method can adequately predict the run-time metrics, and further effectively offload the computation tasks to maximize the run-time performance of the computation service.
|
|
14:30-14:50, Paper ThIPT11.4 | |
>FogROS: An Adaptive Framework for Automating Fog Robotics Deployment |
|
Chen, Kaiyuan | University of California, Berkeley |
Liang, Yafei | University of California, Berkeley |
Jha, Nikhil | University of California, Berkeley |
Ichnowski, Jeffrey | UC Berkeley |
Danielczuk, Michael | UC Berkeley |
Gonzalez, Joseph E. | UC Berkeley |
Kubiatowicz, John | UC Berkeley |
Goldberg, Ken | UC Berkeley |
Keywords: Software, Middleware and Programming Environments, Cloud Computing For Automation
Abstract: As many robot automation applications increasingly rely on multi-core processing or deep-learning models, cloud computing is becoming an attractive and economically viable resource for systems that do not contain high computing power onboard. Despite its immense computing capacity, it is often underused by the robotics and automation community due to the lack of expertise in cloud computing and cloud-based infrastructure. Fog Robotics balances computing and data between cloud edge devices. We propose a software framework, FogROS, as an extension of the Robot Operating System (ROS), the de-facto standard for creating robot automation applications and components. It allows researchers to deploy components of their software to the cloud with minimal effort, and correspondingly gain access to additional computing cores, GPUs, FPGAs, and TPUs, as well as predeployed software made available by other researchers. FogROS allows a researcher to specify which components of their software will be deployed to the cloud and to what type of computing hardware. We evaluate FogROS on 3 examples: (1) simultaneous localization and mapping (ORB-SLAM2), (2) Dexterity Network (Dex-Net) GPU-based grasp planning, and (3) multi-core motion planning using a 96- core cloud-based server. In all three examples, a component is deployed to the cloud and accelerated with a small change in system launch configuration, while incurring additional latency of 1.2s, 0.6s, and 0.5s due to network communication, the computation speed is improved by 2.6×, 6.0× and 34.2×, respectively. Code, videos, and supplementary material can be found at https://github.com/BerkeleyAutomation/FogROS.
|
|
14:50-15:10, Paper ThIPT11.5 | |
>On the Development of Fog-Edge Feedback Applications |
|
Hurtado, Ekaitz | UPV/EHU |
López, Alejandro | University of the Basque Country (UPV/EHU) |
Armentia, Aintzane | University of the Basque Country (UPV/EHU) |
Sarachaga, Isabel | UPV/EHU |
Casquero, Oskar | Faculty of Engineering in Bilbao, University of the Basque Count |
Estévez, Elisabet | Universidad De Jaén |
Marcos, Marga | EIB, University of the Basque Country |
Keywords: Cloud Computing For Automation
Abstract: This work in progress presents a generic procedure for developing reactive fog applications using container technology and Kubernetes application orchestrator.
|
|
ThBT1 Regular Session, Auditorium |
|
Machine Learning 2 |
Chair: Luo, Xin | Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences |
Co-Chair: Navarro, Laurent | Mines Saint-Etienne |
|
16:00-16:20, Paper ThBT1.1 | |
>A Methodology to Identify Cognition Gaps in Visual Recognition Applications Based on Convolutional Neural Networks |
|
Vietz, Hannes | University of Stuttgart, Institute of Industrial Automation And |
Rauch, Tristan | University of Stuttgart |
Löcklin, Andreas | University of Stuttgart, Institute of Industrial Automation And |
Jazdi, Nasser | University of Stuttgart - Institute of Industrial Automation And |
Weyrich, Michael | Univerity of Stuttgart, IAS |
Keywords: Machine learning, Computer Vision for Transportation, AI-Based Methods
Abstract: Developing consistently well performing visual recognition applications based on convolutional neural networks, e.g. for autonomous driving, is very challenging. One of the obstacles during the development is the opaqueness of their cognitive behaviour. A considerable amount of literature has been published which describes irrational behaviour of trained CNNs showcasing gaps in their cognition. In this paper, a methodology is presented that creates worst-case images using image augmentation techniques. If the CNN’s cognitive performance on such images is weak while the augmentation techniques are supposedly harmless, a potential gap in the cognition has been found. The presented worst-case image generator is using adversarial search approaches to efficiently identify the most challenging (worst) image. This is evaluated with the well-known AlexNet CNN using images depicting a typical driving scenario.
|
|
16:20-16:40, Paper ThBT1.2 | |
>Incorporating Generalized Momentum Method to Accelerate Clustering Analysis of Complex Networks |
|
Hu, Lun | Chinese Academy of Sciences |
Pan, Xiangyu | Wuhan University of Technology |
Luo, Xin | Chongqing Institute of Green and Intelligent Technology, Chinese |
Keywords: Machine learning, Big-Data and Data Mining, AI-Based Methods
Abstract: Many complicated systems can be represented by complex networks. Their accurate clustering analysis plays a critical role in understanding their intrinsic organizations. An effective Fuzzy-based Clustering Algorithm for Networks (FCAN) has thus been developed. However, its major disadvantage is its slow convergence to optimal or near-optimal solutions. To overcome this problem, we make use of a generalized momentum method to accelerate it and accordingly propose a fast fuzzy clustering algorithm, namely F^2CAN. Experimental results on several practical datasets demonstrate that F^2CAN performed better than FCAN in terms of efficiency while maintaining the same-level accuracy. Hence, it is more promising to conduct an accurate and fast clustering analysis for complex networks.
|
|
16:40-17:00, Paper ThBT1.3 | |
>In-Situ Quality Monitoring of Extrusion-Based Additive Manufacturing Via Random Forests and Clustering (I) |
|
Bertoli, Luisa | Politecnico Di Milano |
Caltanissetta, Fabio | Politecnico Di Milano |
Bianca Maria Colosimo, Bianca | Politenico Milano |
Keywords: Machine learning, Probability and Statistical Methods, Additive Manufacturing
Abstract: The attention towards in-situ sensing in Additive Manufacturing has dramatically increased over the last years, paving the way to a paradigm shift for quality monitoring and control via big data analysis of signals, images and videos. In-situ quality monitoring represents an opportunity for waste reduction and costs savings via inline detection of process flaws, which allows early identification of scraps and the possibility to correct process parameters for a first-time-right production. Despite of this great potential, no clear and assessed methodologies exist to automatically detect out-of-control states and defects occurrence via in-situ image analysis. This paper discusses opportunities and challenges of in-situ monitoring of Extrusion-based Additive Manufacturing processes by presenting a methodology for in-line defect detection based on stochastic textured surface modelling via Random Forests and k-means clustering for control charting. Significant advantages are shown thus presenting an interesting direction for future research.
|
|
17:00-17:20, Paper ThBT1.4 | |
>Neural Dynamic Assembly Sequence Planning |
|
Kitz, Kristof | Technical University of Chemnitz |
Thomas, Ulrike | Chemnitz University of Technology |
Keywords: Assembly, Reinforcement, Machine learning
Abstract: The automatic generation of feasible assembly sequences from CAD-data is for several reasons a challenging task. One reason being that with increasing number of parts of an assembly group, the number of possible sequences grows exponentially, making an exhaustive search non-practical. We face this combinatorial problem by using Reinforcement Learning (Deep-Q-Learning) to approximate the cost-function of the assembly with an artificial neural network (ANN) and guide the search for an asymptotically optimal solution of this combinatorial problem. Assembly costs are calculated with a collision-based assembly-by-disassembly approach. The derived method is tested on assemblies of different sizes and types. The presented method provides collision-free assembly sequences very fast, due to its depth-first character and solves small and medium tasks reliably.
|
|
17:20-17:40, Paper ThBT1.5 | |
>Using Random Forest and Bayesian Optimization to Predict Hamstringinjuries in Football (soccer) |
|
Bagheri, Fatemeh | Inter‐university Laboratory of Human Movement Science (LIB |
Navarro, Laurent | Mines Saint-Etienne |
Lahti, Johan | Université Côte d’Azur, LAMHESS, Nice, France |
Nagahara, Ryu | Sports Research and Development Core, University of Tsukuba, Iba |
Mendiguchia, Jurdan | Zentrum Rehabilitation and Performance Center, Department of Phy |
Morin, Jean-Benoit | Sports Performance Research Institute New Zealand, Auckland Univ |
Edouard, Pascal | Department of Clinical and Exercise Physiology, Sports Medicine |
Keywords: Machine learning
Abstract: This paper investigates the ability of Random Forest algorithm to predict the risk of hamstring injuries among soccer players. To this aim, we used Random Forest with Bayesian optimization method to construct the model. A total of 284 male football players underwent a preseason screening evaluation including individual anthropometrical and sprint acceleration mechanical output parameters, then subsequent sprint acceleration mechanical output measurements throughout the season. All hamstring injuries were collected by the medical team throughout the season. A precision of 0.93 and recall score of 0.98 in classifying whether a hamstring injury will occur or not during the season, and a f1-score of 95% was reached. In conclusion, these results suggest that robust machine learning techniques, using data usually collected within the football practice, can appropriately monitor injury risk within a season amongst football players.
|
|
ThBT2 Special Session, Rhone 1 |
|
Scheduling Methods for Advanced Manufacturing Systems |
Chair: Sauvey, Christophe | Université De Lorraine |
Co-Chair: Liji, Shen | WHU |
Organizer: Sauvey, Christophe | Université De Lorraine |
Organizer: Soukhal, Ameur | Université De Tours |
Organizer: Sauer, Nathalie | IUT De Thionville Yutz |
|
16:00-16:20, Paper ThBT2.1 | |
>Application of a Genetic Algorithm to Establish an Optimal Production Plan for a Shipyard |
|
Kim, Ji Yoon | Daewoo Shipbuilding & Marine Engineering Co. LTD., |
Jang, Young Jae | Korea Advanced Institute of Science and Technology |
Keywords: Planning, Scheduling and Coordination, Learning and Adaptive Systems, Intelligent and Flexible Manufacturing
Abstract: The genetic algorithm (GA), an evolutionary computing model, is a computational model that attempts to solve complex real-world problems by simulating evolutionary processes. The GA has a simple and general-purpose structure and is widely used to solve engineering problems by adaptive exploration and optimization. In this paper, we propose a method for applying GAs to real-world problems with complex constraints, such as the establishment of factory-specific production plans for shipyards. We verify experimentally that the proposed method can identify an optimal production plan for real plants.
|
|
16:20-16:40, Paper ThBT2.2 | |
>Joint Optimization of Production and Maintenance for a Serial-Parallel Twostage Production System (I) |
|
Shen, Yilan | Peking University |
Zhang, Xi | College of Engineering, Peking University |
Jiang, Wei | Peking University |
Keywords: Manufacturing, Maintenance and Supply Chains, Planning, Scheduling and Coordination
Abstract: Motivated by the tool-cutting process in a launch vehicles manufacturing plant, we study a serial unrelated parallel two-stage production system with imperfect Preventive Maintenance(PM). A degradation model of machines and a joint stochastic optimization model of production scheduling and maintenance for maximizing the production efficiency are developed. The system has an essential characteristic: considering the interaction of the machine deterioration, the actual processing time and the assignment of jobs. Meanwhile, an efficient dynamic adaptive Opportunistic Maintenance(OM) is also proposed in this paper. The random-key GA approach is applied to this problem and numerical results demonstrate the efficiency of the proposed method.
|
|
16:40-17:00, Paper ThBT2.3 | |
>A Monte Carlo Based Method to Maximize the Service Level on the Makespan in the Stochastic Flexible Job-Shop Scheduling Problem (I) |
|
Flores Gomez, Mario | EMSE |
Borodin, Valeria | Mines Saint-Etienne |
Dauzere-Peres, Stephane | Mines Saint-Etienne |
Keywords: Robust Manufacturing, Probability and Statistical Methods
Abstract: This paper considers the flexible job-shop scheduling problem with stochastic processing times. Contrary to the classical criteria in the literature used to deal with random parameters, the probability of the makespan to be smaller than a predefined value, called makespan service level, is maximized. A tabu search approach combined with a Monte Carlo sampling method is proposed. Computational experiments are conducted on extended versions of benchmark instances, including the probabilistic description of random parameters. The numerical results illustrate the impact of key characteristics of the proposed approach.
|
|
17:00-17:20, Paper ThBT2.4 | |
>Minimizing Total Energy Cost in a Flexible Job Shop with Fixed Sequences |
|
Liji, Shen | WHU |
Dauzere-Peres, Stephane | Mines Saint-Etienne |
Maecker, Söhnke | WHU - Otto Beisheim School of Management |
Keywords: Planning, Scheduling and Coordination
Abstract: In this paper, we consider the problem of minimizing the total energy cost of schedules in flexible job-shops subject to fixed operation sequences on the machines and a maximum makespan. A mixed integer programming formulation and heuristic procedures are presented. Computational results show their performances as well as the potential of energy cost savings.
|
|
17:20-17:40, Paper ThBT2.5 | |
>Heuristics Based on MILP for Solving a Flow Shop Rescheduling Problem Simultaneously Considering Efficiency and Stability (I) |
|
Tighazoui, Ayoub | Université De Lorraine, LGIPM |
Sauvey, Christophe | Université De Lorraine |
Sauer, Nathalie | Université De Lorraine, LGIPM |
Keywords: Planning, Scheduling and Coordination
Abstract: To keep a competitive edge, the production firms must quickly react for dealing with uncertainties due to new orders arrivals. In addition, production problems can also disrupt the already established planning. Therefore, the integration of rescheduling process in manufacturing systems is necessary for efficiently revising the initial planning, in preference with little movements. In the literature, many papers consider the efficiency criteria for measuring the schedule performance. However, the stability is also an important aspect for this kind of problems. In this paper, a performance measure considering simultaneously the schedule efficiency and stability is investigated. This one associates the total weighted waiting times and the total weighted completion times deviation. This association of criteria is inspired from real life industrial systems. An iterative predictive-reactive strategy is used for rescheduling the jobs under disruptions caused by new jobs arrival. This problem has first been modeled as a Mixed Integer Linear Programing (MILP) model. Due to its NP-hard complexity, the resolution is only possible for a limited number of jobs. Thereby, two heuristics based on the MILP are designed and discussed in terms of solution quality and computing times.
|
|
ThBT3 Regular Session, Rhone 2 |
|
Automation at Micro-Nano Scales |
Chair: Lutz, Philippe | Femto-St - Umr Cnrs 6174 - Ufc/ensmm/utbm |
Co-Chair: Xu, Qingsong | University of Macau |
|
16:00-16:20, Paper ThBT3.1 | |
>Planning In-Hand Dexterous Micro-Manipulation Using 3-D Rotations Decomposition |
|
Kumar, Pardeep | FEMTO-ST Institute, Univ. Bourgogne Franche-Comte |
Gauthier, Michael | FEMTO-ST Institute |
Dahmouche, Redwan | Université De Franche Comté |
Keywords: Automation at Micro-Nano Scales, Manipulation Planning, Motion and Path Planning
Abstract: This paper aims to contribute to the improvement of dexterity in contact micro-manipulation by performing in-hand dexterous micro-manipulation planning. Previous experimental works on planar micro-manipulation showed that such an approach allows for large rotations of arbitrary shaped objects. Moving from planar to 3-D manipulation significantly increases the complexity of the manipulation planning, especially when considering the rolling of the fingers on the object during the manipulation. We propose in this paper a dexterous manipulation planning algorithm that leverages the complexity of 3-D manipulation planning by decomposing the desired 3-D rotations into three successive rotations within two different planes. Optimal paths of the manipulating fingers are thus obtained in the planar spaces and then combined to form the trajectories in the 3-D space. Besides the relevance of the approach, the simulation results show that exploiting adhesion forces improves the robustness of the manipulation and extends the manipulation capabilities, but at the expense of the computation time.
|
|
16:20-16:40, Paper ThBT3.2 | |
>3D Pose Identification of Micro and Nanowires in Fluid Suspensions |
|
Song, Jiaxu | Binghamton University |
Yu, Kaiyan | Binghamton University |
Keywords: Automation at Micro-Nano Scales
Abstract: This paper presents a passive autofocus-based 3D-pose estimation scheme for micro- and nanowires in fluid suspension under bright-field microscopes. The proposed method integrates classic passive auto-focusing (AF) algorithms, rule-based hill-climb methods, and an automatic and efficient scheme to estimate the positions and orientations of multiple micro- and nanowires in 3D. Experimental results validate the performance of the proposed strategy. This work provides the first foundational step toward the automated control of micro- and nanoagents in 3D microfluidic environments.
|
|
16:40-17:00, Paper ThBT3.3 | |
>Automated Robotic Assembly of 3D Mesostructure Via Guided Mechanical Buckling |
|
Cai, Ying | University of Southern California |
Han, Zhonghao | University of Southern California |
Cranney, Trey | University of Southern California |
Zhao, Hangbo | University of Southern California |
Gupta, Satyandra K. | University of Southern California |
Keywords: Automation at Micro-Nano Scales, Assembly
Abstract: We present an automated assembly approach to forming 3D mesostructures using guided mechanical buckling of patterned thin films. This task requires accurate positioning of mesostructures over large distances. We use an industrial robot with a high degree of repeatability and large reach. We utilize image-guided localization and positioning to enable accurate pick and place of mesoscale thin films, dispensing of nanoliter adhesive in targeted regions, and automatic 2D to 3D shape transformation via mechanical buckling. We achieved the positioning accuracy of 80 microns, as demonstrated in the example of automated mechanical assembly of 3D mesostructures. The positioning accuracy could be further improved by enhancing the positioning accuracy of the robot, increasing the image resolution, and optimizing the assembly process. The use of industrial robots with image-guided localization and positioning provides potential opportunities for high-accuracy, low-cost, and complex robotic manipulation at meso- and microscale.
|
|
17:00-17:20, Paper ThBT3.4 | |
>Design and Development of a Teleoperated Robotic Microinjection System with Haptic Feedback |
|
Feng, Kai | University of Macau |
Xu, Qingsong | University of Macau |
Tam, Lap Mou | University of Macau |
Keywords: Automation at Micro-Nano Scales
Abstract: This paper proposes a new teleoperated robotic system with haptic feedback dedicated to biological microinjection. The system consists of a custom-built master device and a slave micromanipulator device. Microinjection task is conducted by the micromanipulator equipped with a microforce sensor. The slave device is controlled by a human operator via the master device, which includes a touch device and a haptic device. The haptic device is designed based on the working principle of a syringe, which is driven by a voice coil motor. The haptic device enables the operator to feel the microinjection process when the operator holds the handle by a thumb. Through the kinematics analysis and workspace matching, a position mapping method is introduced to convert the position of the master device to that of the slave micromanipulator. Experimental calibration results show that the microforce sensor provides the resolution of 0.101 mN. The haptic device responds quickly to the contact force with an average response time of 0.2 s. Moreover, the position tracking results show that the slave micromanipulator can follow the command of the master device faithfully. With the intuitive haptic device, the operator can feel the contact and determine the injection status for improving the success rate of the microinjection task.
|
|
17:20-17:40, Paper ThBT3.5 | |
>6-DoF Full Robotic Calibration Based-On 1-D Interferometric Measurements for Micro and Nano-Scales Applications |
|
Bettahar, Houari | Aalto University |
Lehmann, Olivier | Universite De Franche-Comté |
Clévy, Cédric | Franche-Comté University |
Courjal, Nadège | Univ. Bourgogne Franche-Comté, Univ. De Franche-Comté/CNRS/ENSMM |
Lutz, Philippe | Femto-St - Umr Cnrs 6174 - Ufc/ensmm/utbm |
Keywords: Automation at Micro-Nano Scales, Calibration and Identification
Abstract: This paper proposes an original approach for robotic calibration that is based on measurements along a single direction (1-D). Among all applications, the field of micro andnano robotics has been chosen as case-study because of the strong needs for high positioning accuracy (10-100 nm typically) while measuring with sufficient resolution along multi-DoF (Degrees-of-Freedom) is still a fully open question. 1-D measurements relying on FP (Fabry-Perot) interferences is used and the proposed modelling of a 6-DoF nanopositioning robot enables to derive the measurement strategy as well as the identification procedure for both extrinsic and intrinsic parameters. Experimental investigations demonstrate that the approach is easy to implement, low cost and enables to understand what are the main influential parameters onto positioning accuracy. They also conduct to very high accuracy in 6 DoF positioning: a positioning accuracy estimate of 50 nm and 0.004_ has notably been obtained for the full pose (position and orientation respectively) and can be held during several hours after the measurements.
|
|
ThBT4 Regular Session, Rhone 3A |
|
Autonomous Agents |
Chair: Nguyen-Cong, Trinh | Carl Zeiss AG |
Co-Chair: Nagi, Rakesh | University of Illinois, Urbana-Champaign |
|
16:00-16:20, Paper ThBT4.1 | |
>Proficiency Constrained Multi-Agent Reinforcement Learning for Environment-Adaptive Multi UAV-UGV Teaming |
|
Yu, Qifei | Cognitive Robotics and AI Lab (CRAI), College of Aeronautics And |
Zhexin, Shen | Cognitive Robotics and AI Lab (CRAI), College of Aeronautics And |
Pang, Yijiang | Cognitive Robotics and AI Lab (CRAI), College of Aeronautics And |
Liu, Rui | Kent State University |
Keywords: Autonomous Agents, Reinforcement, AI-Based Methods
Abstract: A mixed aerial and ground robot team, which includes both unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs), is widely used for disaster rescue, social security, precision agriculture, and military missions. However, team capability and corresponding configuration vary since robots have different motion speeds, perceiving ranges, reaching areas, and resilient capabilities to the dynamic environment. Due to heterogeneous robots inside a team and the resilient capabilities of robots, it is challenging to perform a task with an optimal balance between reasonable task allocations and maximum utilization of robot capability. To address this challenge for effective mixed ground and aerial teaming, this paper developed a novel teaming method, proficiency aware multi-agent deep reinforcement learning (Mix-RL), to guide ground and aerial cooperation by considering the best alignments between robot capabilities, task requirements, and environment conditions. Mix-RL largely exploits robot capabilities while being aware of the adaption of robot capabilities to task requirements and environment conditions. Mix-RL's effectiveness in guiding mixed teaming was validated with the task "social security for criminal vehicle tracking".
|
|
16:20-16:40, Paper ThBT4.2 | |
>Spatial-Temporal Graph Neural Network for Interaction-Aware Vehicle Trajectory Prediction |
|
Chen, Junan | Cornell University |
Wang, Yan | Cornell |
Wu, Ruihan | Cornell University |
Campbell, Mark | Cornell University |
Keywords: Autonomous Agents, Machine learning, Collision Avoidance
Abstract: In this paper, a Spatial Temporal Graph Neural Network (STGNN) model is developed, including a temporal block and Graph Neural Network (GNN) block, to solve the problem of vehicle trajectory prediction in unstructured scenes. Specifically, a temporal block combines a recurrent neural network and non-local operation to extract the features from past trajectories, and a GNN block models the subtle interactions between vehicles. The proposed model is evaluated on two datasets: Unstructured Scene Dataset and Argoverse Dataset. Experiment results show that the STGNN model achieves a better performance in the unstructured scenes and can be applied to common scenes where rules of the road dominate.
|
|
16:40-17:00, Paper ThBT4.3 | |
>Distributed Adaptive Formation Control of Multi-Agent Systems in Three-Dimensional Space |
> Video
|
|
Chen, Chih-Wei | National Taiwan University |
Chiang, Ming-Li | National Taiwan University |
Su, Kuan-Yu | National Taiwan University |
Chen, Yu-Wen | National Taiwan University |
Fu, Li-Chen | National Taiwan University |
Keywords: Autonomous Agents, Robust/Adaptive Control, Formal Methods in Robotics and Automation
Abstract: In this paper, we discuss formation and maneuver control of multi-agent systems (MAS) in the three-dimensional space. The system is controlled in a distributed manner with connected communication links. Our considered agents are mainly focused on the micro unmanned aerial vehicle (UAV) and thus a second-order integrator model is adopted. The control target is to achieve desired formation and reference trajectory tracking. The agents only use information from their neighbors and keep the connectivity to maintain the communication while tracking the reference trajectory. Moreover, our approach allows the agents to form into the desired shape with a given order relation, due to the given formation formulation and the adaptive control structure. Several examples and software in the loop simulation are given to validate the proposed results.
|
|
17:00-17:20, Paper ThBT4.4 | |
>Autonomous Navigation for Quadrupedal Robots with Optimized Jumping through Constrained Obstacles |
> Video
|
|
Gilroy, Scott | University of California, Berkeley |
Lau, Derek | University of California Berkeley |
Yang, Lizhi | University of California, Berkeley |
Izaguirre, Ed | UC Berkeley |
Biermayer, Kristen | UC Berkeley |
Xiao, Anxing | Harbin Institute of Technology, Shenzhen |
Sun, Mengti | University of California, Berkeley |
Agrawal, Ayush | University of California at Berkeley |
Zeng, Jun | University of California, Berkeley |
Li, Zhongyu | University of California, Berkeley |
Sreenath, Koushil | University of California, Berkeley |
Keywords: Autonomous Agents, Collision Avoidance, Task Planning
Abstract: Quadrupeds are strong candidates for navigating challenging environments because of their agile and dynamic designs. This paper presents a methodology that extends the range of exploration for quadrupedal robots by creating an end-to-end navigation framework that exploits walking and jumping modes. To obtain a dynamic jumping maneuver while avoiding obstacles, dynamically-feasible trajectories are optimized offline through collocation-based optimization where safety constraints are imposed. Such optimization schematic allows the robot to jump through window-shaped obstacles by considering both obstacles in the air and on the ground. The resulted jumping mode is utilized in an autonomous navigation pipeline that leverages a search-based global planner and a local planner to enable the robot to reach the goal location by walking. A state machine together with a decision making strategy allows the system to switch behaviors between walking around obstacles or jumping through them. The proposed framework is experimentally deployed and validated on a quadrupedal robot, a Mini Cheetah, to enable the robot to autonomously navigate through an environment while avoiding obstacles and jumping over a maximum height of 13 cm to pass through a window-shaped opening in order to reach its goal.
|
|
17:20-17:40, Paper ThBT4.5 | |
>Decentralized Makespan Minimization for Uniformly Related Agents |
|
Sengupta, Raunak | University of Illinois, Urbana-Champaign |
Nagi, Rakesh | University of Illinois, Urbana-Champaign |
Keywords: Agent-Based Systems, Autonomous Agents, Planning, Scheduling and Coordination
Abstract: We consider a set of indivisible operations and a set of uniformly related agents, i.e., agents with different speeds. Our aim is to develop a task allocation algorithm that minimizes the makespan in a decentralized manner. To achieve this, we first present the Operation Trading Algorithm. We show that the algorithm guarantees a worst case approximation factor of 1.618 for the 2 agent case and frac{1+sqrt{4n-3}}{2} for the general n agent case. Further, we prove that the algorithm guarantees a near-optimal makespan for real-life scenarios with large number of operations under the assumption of a fully connected network of agents. The algorithm also guarantees an approximation factor less than 2 for any number of identical agents. Following this, we present a Decentralized random Group Formation protocol which enables the agents to implement OTA(n) in a decentralized manner in presence of communication failures. Finally, using numerical results, we show that the algorithm generates near optimal allocations even in the presence of communication failures. Additionally, the algorithm is parameter free and allows fast re-planning, making it robust to machine failures and changes in the environment.
|
|
17:40-18:00, Paper ThBT4.6 | |
>Fast Adaptable 6D Object Pose Estimation for Autonomous Robotics |
|
Nguyen-Cong, Trinh | Carl Zeiss AG |
Zhu, Jinyao | Carl Zeiss AG |
Glasenapp, Carsten | Carl Zeiss AG |
Karl, Matthias | Carl Zeiss AG |
Keywords: Autonomous Agents, Deep Learning in Robotics and Automation, Computer Vision in Automation
Abstract: We propose and demonstrate a robust 6D object pose estimation which is based on a deep-learning method. Robustness was confirmed by examining impacts such as illumination, occlusion, or object truncation making it a viable tool for autonomous robotic applications. This work belongs to a multi-stage (from coarse to very fine) pose estimation pipeline which is currently under development.
|
|
ThBT5 Regular Session, Rhone 3B |
|
Building Automation |
Chair: Ferrarini, Luca | Politecnico Di Milano |
Co-Chair: Böhm, Michael | University of Stuttgart |
|
16:00-16:20, Paper ThBT5.1 | |
>UltraBot: Autonomous Mobile Robot for Indoor UV-C Disinfection |
> Video
|
|
Perminov, Stepan | Skolkovo Institute of Science and Technology |
Mikhailovskiy, Nikita | Skolkovo Institute of Science and Technology |
Sedunin, Alexander | Skolkovo Institute of Science and Technologies |
Okunevich, Iaroslav | Skolkovo Institute of Science and Technology |
Kalinov, Ivan | Skolkovo Institute of Science and Technology |
Kurenkov, Mikhail | Skolkovo Institute of Science and Technology |
Tsetserukou, Dzmitry | Skolkovo Institute of Science and Technology |
Keywords: Automation Technologies for Smart Cities, Product Design, Development and Prototyping
Abstract: The paper focuses on the development of the autonomous robot UltraBot to reduce COVID-19 transmission and other harmful bacteria and viruses. The motivation behind the research is to develop such a robot that is capable of performing disinfection tasks without the use of harmful sprays and chemicals that can leave residues, require airing the room afterward for a long time, and can cause the corrosion of the metal structures. UltraBot technology has the potential to offer the most optimal autonomous disinfection performance along with taking care of people, keeping them from getting under UV-C radiation. The paper highlights UltraBot's mechanical and electrical structures as well as low-level and high-level control systems. The conducted experiments demonstrate the effectiveness of the robot localization module and optimal trajectories for UV-C disinfection. The results of UV-C disinfection performance revealed a decrease of the total bacterial count (TBC) by 94% on the distance of 2.8 meters from the robot after 10 minutes of UV-C irradiation.
|
|
16:20-16:40, Paper ThBT5.2 | |
>Practical Full Automation of Excavation and Loading for Hydraulic Excavators in Indoor Environments |
> Video
|
|
Yoshida, Hiroshi | NEC Corporation |
Yoshimoto, Tatsuya | NEC Corporation |
Umino, Daiki | Obayashi Corporation |
Mori, Naoki | Obayashi Corporation |
Keywords: Automation in Construction, Sensor Networks
Abstract: We propose an autonomous operation system for hydraulic excavators that fully automatizes the sequence of excavation and loading in an indoor environment at actual tunneling construction sites. This system possesses all the components and functions that are necessary for excavation and loading. It also consists of an architecture composed of a networked control system. This architecture enables the system to mitigate the physical limitations of sensors and computers such as installation locations and computing power. We also developed practical engineering solutions that make our system suitable for construction sites, such as prediction control concerning hydraulic and mechanical delays and an external actuator that requires no special modifications to construction machinery.
|
|
16:40-17:00, Paper ThBT5.3 | |
>Optimal Dynamic Duct Static Pressure Method in a Multi-Zone Variable Air Volume System |
|
Wang, Xuetao | Tsinghua University |
Zhao, Qianchuan | Tsinghua University |
Wang, Yifan | Tsinghua University |
Xing, Tian | TsingHua University |
Keywords: Building Automation, Energy and Environment-Aware Automation
Abstract: Reducing the energy consumption of a variable air volume (VAV) system in heating, ventilation, and air-conditioning (HVAC) systems attracts many attentions. In this paper, a novel method, namely, optimal dynamic duct static pressure (ODSP) method is proposed to find the globally optimal solutions of the original non-convex multi-zone VAV system energy optimization problem. Our method is based on the primal dual interior point method to solve an equivalent convex optimization problem of the multi-zone VAV system energy optimization problem. Different from existing methods, our method is a real-time optimization method and explicitly considers the duct network model to find the energy saving strategy. Numerical simulations illustrate the effectiveness of our method compared with current methods, which show that our method provides better energy efficiency and nearly 20% energy is saved. Our method also has the guarantee to find the optimal solutions.
|
|
17:00-17:20, Paper ThBT5.4 | |
>Reconfiguration Strategy for Fault-Tolerant Control of High-Rise Adaptive Structures |
|
Dakova, Spasena | University of Stuttgart |
Wagner, Julia Laura | University of Stuttgart |
Gienger, Andreas | University of Stuttgart |
Tarín, Cristina | University of Stuttgart |
Böhm, Michael | University of Stuttgart |
Sawodny, Oliver | University of Stuttgart |
Keywords: Building Automation, Failure Detection and Recovery
Abstract: Adaptive structures comprise sensors, a control unit as well as actuators and are able to actively counteract to external disturbances. However, an actuator failure can lead to a significant performance degradation, which can be reduced by designing a reconfigurable control law. In this paper, a fault-tolerant controller for an adaptive high-rise structure is designed, which is able to adapt to multiple actuator failures and thus is an important step for the automation of adaptive structures. The proposed control law consists of two parts – a static offset compensation, which counteracts the constant force applied by the faulty actuators on the mechanical structure, and a reconfigurable linear quadratic regulator, which optimally minimizes the vibrations of the structure using the remaining functioning actuators. The introduced approach is evaluated in a simulation study considering a wind disturbance. As a result the fault tolerant control scheme ensures a more efficient energy operation and yields a performance improvement of up to 33% compared to a nominal controller. Consequently, the reliability of adaptive structures is increased.
|
|
17:20-17:40, Paper ThBT5.5 | |
>Hierarchical Nonlinear MPC for Large Buildings HVAC Optimization |
|
Rastegarpour, Soroush | Politecnico Di Milano |
Ferrarini, Luca | Politecnico Di Milano |
Keywords: Building Automation, Energy and Environment-Aware Automation
Abstract: This paper studies the problem of performance improvement and energy consumption reduction of the heating, ventilation and air conditioning system of a large-scale university building through the application of nonlinear predictive control strategies concerning also practical and implementation issues. The system consists of two heat pumps, a water-to-water and an air-to-water type, and two different air handling units, which regulate and circulate air in all thermal zones. In such applications, prediction of the future dynamical behavior of the heat pumps is extremely important to enforce efficiency, but it is also very challenging due to the load dependency and nonlinearity of the coefficient of performances of those heat pumps. On the other hand, another source of potential model mismatch is the nonlinear characterization of the heat transfer coefficients of the AHU induced by variable air and water velocity, which gives rise to a non-trivial nonlinear system. To do so, two nonlinear model predictive control strategies are investigated to deal with many physical constraints and nonlinear problems. Finally, a sensitivity and robustness analysis are performed to highlight the merits, defects and impacts of those control algorithms on the energy performance of the building.
|
|
ThBT6 Special Session, St Clair 1 |
|
Distributed Path and Task Planning of Multiagent Systems |
Chair: Mahulea, Cristian | Universidad De Zaragoza |
Co-Chair: Baran, Robin Loïc | KTH Royal Institute of Technology |
Organizer: Mahulea, Cristian | Universidad De Zaragoza |
Organizer: Montijano, Eduardo | Universidad De Zaragoza |
Organizer: Dimarogonas, Dimos V. | KTH Royal Institute of Technology |
|
16:00-16:20, Paper ThBT6.1 | |
>Assessing and Restoring ``traffic-State Order'' in Open, Irreversible, Dynamically Routed, Zone-Controlled Guidepath-Based Transport Systems (I) |
|
Reveliotis, Spiridon | Georgia Institute of Technology |
|
|
16:20-16:40, Paper ThBT6.2 | |
>A ROS Package for Human-In-The-Loop Planning and Control under Linear Temporal Logic Tasks (I) |
> Video
|
|
Baran, Robin Loïc | KTH Royal Institute of Technology |
Tan, Xiao | KTH Royal Institute of Technology, Sweden |
Varnai, Peter | KTH Royal Institute of Technology |
Yu, Pian | KTH Royal Institute of Technology |
Ahlberg, Sofie | KTH Royal Institute of Technology |
Guo, Meng | Bosch Group |
Shaw Cortez, Wenceslao | Royal Institute of Technology (KTH) |
Dimarogonas, Dimos V. | KTH Royal Institute of Technology |
Keywords: Formal Methods in Robotics and Automation, Human Factors and Human-in-the-Loop, Task Planning
Abstract: In this paper, we propose a ROS software package for planning and control of robotic systems with a human-in-the-loop focus. The software uses temporal logic specifications, specifically Linear Temporal Logic, for a language-based method to develop correct-by-design high level robot plans. The approach is structured to allow a human to adjust the high-level plan online. A human may also take control of the robot (in a low-level control fashion), but the software prevents the human from implementing dangerous behaviour that would violate the high-level task specification. Finally, the planner is able to learn human-preferred high-level tasks by tracking human low-level control inputs in an inverse learning framework. The proposed approach is demonstrated in a warehouse setting with multiple robot agents to showcase the efficacy of the proposed solution.
|
|
16:40-17:00, Paper ThBT6.3 | |
>Probabilistic Multi-Robot Path Planning with High-Level Specifications Using Petri Net Models (I) |
|
Montijano, Eduardo | Universidad De Zaragoza |
Mahulea, Cristian | Universidad De Zaragoza |
Keywords: Petri Nets for Automation Control, Motion and Path Planning, Robot Networks
Abstract: This paper considers the path planning problem of a multi-robot system in an uncertain environment with high-level specifications. The specification that the team of robots has to satisfy is given in form of a Boolean formula over the set of regions of interest. Opposed to the majority of solutions for this type of problem, in this paper we consider that the robots do not know with precision the location of these regions, having a probability distribution over the labeling instead. We present a planning algorithm that leverages the properties of Petri net models to efficiently solve an optimization problem that maximizes the probability of satisfying the formula. In addition, the solution of the optimizer avoids regions with high probability of being obstacles and tries to reduce the chances of collisions of the team. We also discuss the application of the algorithm in a distributed scenario, where the robots have different probability distributions and even outdated locations of the rest of the team. Simulations are presented to validate the proposed methodology.
|
|
17:00-17:20, Paper ThBT6.4 | |
>A Network-Flow Reduction for the Multi-Robot Goal Allocation and Motion Planning Problem (I) |
|
Salvado, João | Orebro University |
Mansouri, Masoumeh | Birmingham University |
Pecora, Federico | Örebro University |
Keywords: Motion and Path Planning, Planning, Scheduling and Coordination, Optimization and Optimal Control
Abstract: This paper deals with the problem of allocating goals to multiple robots with complex dynamics while computing obstacle-free motions to reach those goals. The spectrum of existing methods ranges from complete and optimal approaches with poor scalability, to highly scalable methods which make unrealistic assumptions on the robots and/or environment. We overcome these limitations by using an efficient graph-based method for decomposing the problem into sub-problems. In particular, we reduce the problem to a Minimum-Cost Max-Flow problem whose solution is used by a multi-robot motion planner that does not impose restrictive assumptions on robot kinodynamics or on the environment. We show empirically that our approach scales to tens of robots in environments composed of hundreds of polygons.
|
|
17:20-17:40, Paper ThBT6.5 | |
>A Distributed Optimization and Control Framework for a Network of Constraint Coupled Residential BESSs |
|
Kaheni, Mojtaba | University of Cagliari |
Usai, Elio | University of Cagliari |
Franceschelli, Mauro | University of Cagliari, Italy |
Keywords: Agent-Based Systems, Demand Side Management, Distributed Generation and Storage
Abstract: In this paper we propose a distributed optimization and control protocol to optimize the behavior of a network of domestic Battery Energy Storage Systems (BESS) which offers several advantages with respect to privacy protection due to its control and information sharing architecture. We use the Lagrange multipliers in our optimization protocol. The model of the network of BESS includes the possibility of local power generation and power transfer to the grid. The proposed method enables autonomous decision making for each BESS. Information regarding the state of charge and related constraints of each BESS is not shared, thus increasing the protection of the user privacy. Numerical results are provided to validate the proposed approach.
|
|
17:40-18:00, Paper ThBT6.6 | |
>A Decentralized B&B Algorithm for Motion Planning of Robot Swarms with Temporal Logic Specifications |
|
Yan, Ruixuan | Rensselaer Polytechnic Institute |
Julius, Agung | Rensselaer Polytechnic Institute |
Keywords: Agent-Based Systems, Planning, Scheduling and Coordination
Abstract: In this paper, we study the problem of decentralized motion planning of robot swarms under high-level temporal logic specifications with a top-down approach. We use Swarm Signal Temporal Logic (SwarmSTL) to express swarm-level specifications. By encoding SwarmSTL formulas as mixed binary-integer constraints on the swarm features, the motion planning problem is formulated as a mixed-integer quadratic programming (MIQP) problem. We develop a decentralized Branch and Bound (B&B) algorithm with a node decentralization scheme such that the nodes in the B&B tree can be processed in parallel with communication among the agents and the agents can achieve consensus on the solution. Also, several search strategies to accelerate the decentralized B&B algorithm are proposed, and the performance improvements are presented. We evaluate the proposed algorithm using a supply transportation example with different formulas.
|
|
ThBT7 Special Session, St Clair 2 |
|
Collaborative Ontologies for Semiconductor Supply Chains |
Chair: Ehm, Hans | Infineon Technologies AG |
Co-Chair: Yugma, Claude | Ecole Des Mines De Saint-Etienne |
Organizer: Ehm, Hans | Infineon Technologies AG |
Organizer: Yugma, Claude | Ecole Des Mines De Saint-Etienne |
|
16:00-16:20, Paper ThBT7.1 | |
>Collaborative Ontology Development -- Bridging the Gap between Knowledge Engineers and Domain Experts (I) |
|
Wiens, Vitalis | Tib, L3s, Luh |
Dimitrakopoulos, George | Harokopio University of Athen |
Keywords: Human Factors and Human-in-the-Loop, Domain-specific Software and Software Engineering
Abstract: Semantic Web technologies receive growing attention in industrial and academic contexts. Thus, ontology development typically involves joint efforts of domain experts and knowledge engineers. Domain experts find it often hard to follow logical notations in OWL representation, and knowledge engineers often lack the depth knowledge of a specific domain to create ontologies of sufficient quality. To more directly involve domain experts in ontology modeling and foster communication with knowledge engineers, approaches require mechanisms to reduce OWL formalization complexity for domain experts but maintain the full OWL modeling features for knowledge engineers. This work presents a user-centered approach for collaborative ontology development, addressing different user groups, their needs and backgrounds in Semantic Web contexts.
|
|
16:20-16:40, Paper ThBT7.2 | |
>A Hardware / Software Allocation Core Ontology for Collaboration Along the Value Chain (I) |
|
Wawrzik, Frank | TU Kaiserslautern |
Keywords: Big-Data and Data Mining, Manufacturing, Maintenance and Supply Chains
Abstract: This abstract introduces a hardware core ontology that supports allocation from hardware to software and is designed to facilitate innovation development throughout the automotive value chain. The ontology consistently describes various processor parts with their properties, functions and dependencies.
|
|
16:40-17:00, Paper ThBT7.3 | |
>The Digital Reference: Semantically Connecting Semiconductor Supply Chains to Customers - the Open Online Sales and Marketing Vision (I) |
|
Ramzy, Nour | Leibniz Universität Hannover , Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Wiens, Vitalis | Tib, L3s, Luh |
Laura, Kohnen | Infineon Technologies AG |
Keywords: Manufacturing, Maintenance and Supply Chains, Semiconductor Manufacturing
Abstract: The semiconductor industry is a competitive market where supply chains are characterized by high innovation. This entails the importance of highly connecting customers' demands with supply chains and offering solutions to their needs. The Digital Reference (DR)is a Semantic Web representation for semiconductor supply chains and supply chains containing semiconductors. This vocabulary enables heterogeneous data integration from various supply chain parts. Within funded projects e.g. SC3, a sustainable Industrial Reference platform suitable for extending, maintaining, and using the DR is created. Data infrastructure for DR is established allowing incremental extension of DR with domain knowledge and data to further expand its scale to more areas of semiconductor supply chain processes. In order to semantically connect Supply Chain parties, we present a system architecture that relies on an existing solution finder platform (OOSMP) with a knowledge graph based on the Digital Reference. The ontology-based product exploration contributes to establishing relations between concepts and entities of the platform and link products with solutions. We evaluate the solution, the integration of the OOSMP ontology to the Digital Reference enables semantically connecting customers, products, and applications of products.
|
|
17:00-17:20, Paper ThBT7.4 | |
>Investigating Semantic Web As Enabler for Semiconductor Supply Chain Collaboration (I) |
|
Ti, Yun | RWTH Aachen |
Moder, Patrick | Infineon Technologies AG |
Ramzy, Nour | Leibniz Universität Hannover , Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Keywords: Cyber-physical Production Systems and Industry 4.0, Manufacturing, Maintenance and Supply Chains, Semiconductor Manufacturing
Abstract: Supply Chain Collaboration is a promising approach for network members to exchange information in a trustworthy environment in order to obtain better supply chain performance. Driven by technology development, there are many technology candidates related to information sharing for supply chain management to obtain better supply chain performance. Semantic Web technologies have received increased attention from industry and academic domains. However, there is a lack of research that focuses on possible application of Semantic Web technologies. This work presents a practical approach investigating impact of Semantic Web as an enabler for the semiconductor supply chain. In this study, we show how Semantic Web could improve the supply chain collaboration to increase Supply Chain Performance in Semiconductor chain.
|
|
17:20-17:40, Paper ThBT7.5 | |
>Ontology-Based Analysis of Potential of CO2 Savings in a Semiconductor Industry (I) |
|
Irene Eliza Thomas, Irene | Infineon Technologies |
Ismail, Abdelgafar | Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Keywords: Big data Analytics for Large-scale Energy Systems, Semiconductor Manufacturing, Manufacturing, Maintenance and Supply Chains
Abstract: The significant growth in the global atmospheric CO2 concentration from over a century has been associated with the recent notable global climate change. The importance of carbon footprint analysis has been increasingly discussed as a crucial step to tackle the climate change. Despite technical advancements in the past decade, there are still misconceptions about an efficient technique to calculate and analyze the same. This paper presents a holistic approach based on ontology (knowledge processing) to help make a smarter decision at an early stage to analyze the impact along with safety considerations. A case study of an electric car from a semiconductor industry point of view is used to explain the novelty of the approach. The paper also points out the contribution towards the direction of a holistic knowledge base called the digital reference which links through different knowledge across sectors to enable engineers to come up with more suitable decisions.
|
|
17:40-18:00, Paper ThBT7.6 | |
>How to Test a Human-Based Consensus for Trust Installation in Ontologies? (I) |
|
Felix Neizert, Felix | Infineon Technologies AG, RWTH Aachen University |
Summerer, Christoph | Technical University of Munich |
Ehm, Hans | Infineon Technologies AG |
Regnath, Emanuel | Technical University of Munich |
Ramzy, Nour | Leibniz Universität Hannover , Infineon Technologies AG |
Steinhorst, Sebastian | Technical University of Munich |
Keywords: Human Factors and Human-in-the-Loop, Behavior-Based Systems, Domain-specific Software and Software Engineering
Abstract: Ontologies based on Semantic Web technology are readable by humans and machines alike. This attribute makes them a possible technology enabler, meaning at the same time that one single-source of truth is needed in order to ensure the applicability. For frequently updated ontologies it is therefore essential to have trust installed in the newest version. [1] Blockchain technology seems to be a promising solution for installing the trust, offering additional potential benefits like transparency, immutability and efficiency. [2] Central part of a blockchain technology is the consensus-finding mechanism that can be based on human voting principles. Such a consensus approach that suits to the use of ontology validation was already developed: Named “Human-based Consensus for Trust Installation in Ontologies” [3] it is in the poster track of the 2021 IEEE International Conference on Blockchain and Cryptocurrency (ICBC 2021). In a next step we need to go further and find a way of applying and testing this approach, since the hardly predictable human participation and behavior are a central part of the blockchain approach. In the following the developed human-based consensus approach and its analytical evaluation is shortly described with reference to the respective paper [3], followed by our approach of creating a platform and the right environment in order to apply, test and evaluate the consensus mechanism.
|
|
ThBT8 Regular Session, Rhone 4 |
|
Agricultural Automation and Biomimetics |
Chair: Godage, Isuru S. | Depaul University |
Co-Chair: Papanikolopoulos, Nikos | University of Minnesota |
|
16:00-16:20, Paper ThBT8.1 | |
>A Novel Variable Stiffness Soft Robotic Gripper |
> Video
|
|
Kodippili Arachchige, Dimuthu Dharshana | DePaul University |
Chen, Yue | University of Arkansas |
Walker, Ian | Clemson University |
Godage, Isuru S. | Depaul University |
Keywords: Compliant Assembly, Biomimetics, Actuation and Joint Mechanisms
Abstract: Compliant grasping is crucial for secure handling objects not only vary in shapes but also in mechanical properties. We propose a novel soft robotic gripper with decoupled stiffness and shape control capability for performing adaptive grasping with minimum system complexity. The proposed soft fingers conform to object shapes facilitating the handling of objects of different types, shapes, and sizes. Each soft gripper finger has a length constraining mechanism (an articulable rigid backbone) and is powered by pneumatic muscle actuators. We derive the kinematic model of the gripper and use an empirical approach to simultaneously map input pressures to stiffness control and bending deformation of fingers. We use these mappings to demonstrate decoupled stiffness and shape (bending) control of various grasping configurations. We conduct tests to quantify the grip quality when holding objects as the gripper changes orientation, the ability to maintain the grip as the gripper is subjected to translational and rotational movements, and the external force perturbations required to release the object from the gripper under various stiffness and shape (bending) settings. The results validate the proposed gripper’s performance and show how the decoupled stiffness and shape control can improve the grasping quality in soft robotic grippers.
|
|
16:20-16:40, Paper ThBT8.2 | |
>A Portable Agricultural Robot for Continuous Apparent Soil Electrical Conductivity Measurements to Improve Irrigation Practices |
|
Campbell, Merrick | University of California, Riverside |
Ye, Keran | University of California, Riverside |
Scudiero, Elia | University of California, Riverside |
Karydis, Konstantinos | University of California, Riverside |
Keywords: Agricultural Automation, Environment Monitoring and Management
Abstract: Near-ground sensing data, such as geospatial measurements of soil apparent electrical conductivity (ECa), are used in precision agriculture to improve farming practices and increase crop yield. Near-ground sensors provide valuable information, yet, the process of collecting, assessing, and interpreting measurements requires significant human labor. Automating parts of this process via the use of mobile robots can help decrease labor burden, and increase the accuracy and frequency of data collections, and overall increase the adoption and use of ECa measurement technology. This paper introduces a roboticized means to autonomously perform geospatial ECa measurements and map soil moisture content in micro-irrigated orchard systems. We retrofit a small wheeled mobile robot with a small electromagnetic induction sensor by studying and taking into consideration the effect of the robot body to the sensor's readings, and develop a software stack to enable autonomous logging of geo-referenced measurements. The proposed roboticized ECa measurement method is evaluated by mapping a 50m x 30m field against the baseline of human-conducted measurements obtained by walking the sensor in the same field and following the same path. Experimental testing reveals that our approach yields roboticized measurements comparable to human-conducted ones, despite the robot's small form factor.
|
|
16:40-17:00, Paper ThBT8.3 | |
>Resolution-Optimal, Energy-Constrained Mission Planning for Unmanned Aerial/Ground Crop Inspections |
|
Edmonds, Merrill | Rutgers, the State University of New Jersey |
Yigit, Tarik | Rutgers University |
Yi, Jingang | Rutgers University |
Keywords: Agricultural Automation, Planning, Scheduling and Coordination
Abstract: Precision agriculture relies on large-scale visual inspections for accurate crop monitoring and yield maximization. For many farms, the scales of production preclude manual inspections, and it is therefore desirable for larger producers to employ unmanned ground and aerial vehicles (UGV/UAV) to automate the necessary proximal and remote sensing tasks, respectively. This paper presents a new problem formulation for cooperative crop inspection missions under fuel and pathing constraints. We propose an a priori optimization method that leverages knowledge of the energy constraints and plot topology to determine resolution-optimal walks on a graph representing the union of reachable sets for each robot. We show that approximating the reachable sets guarantees energy efficiency. We further show that UGV-UAV interactions such as set-hopping can increase the effective continuous monitoring range. Simulation studies show that our method accounts for charge-recharge cycles that are typical of long inspection missions, while also optimizing capture time and sensing resolution.
|
|
17:00-17:20, Paper ThBT8.4 | |
>ReQuBiS - Reconfigurable Quadrupedal-Bipedal Snake Robots |
> Video
|
|
Zade, Harshad | Visvesvaraya National Institute of Technology |
Varude, Aadesh | Visvesvaraya National Institute of Technology |
Pandya, Karan | Visvesvaraya National Institute of Technology, Nagpur |
Kamat, Ajinkya | VNIT |
Chiddarwar, Shital | Visvesvaraya National Institute of Technology, Nagpur |
Thakker, Rohan | Nasa's Jet Propulsion Laboratory, Caltech |
Keywords: Biomimetics, Actuation and Joint Mechanisms, Compliant Joints and Mechanisms
Abstract: The selection of mobility modes for robot navigation consists of various trade-offs. Snake robots are ideal for traversing through constrained environments such as pipes, cluttered and rough terrain, whereas bipedal robots are more suited for structured environments such as stairs. Finally, quadruped robots are more stable than bipeds and can carry larger payloads than snakes and bipeds but struggle to navigate soft soil, sand, ice, and constrained environments. A reconfigurable robot can achieve the best of all worlds. Unfortunately, state-of-the-art reconfigurable robots rely on the rearrangement of modules through complicated mechanisms to dissemble and assemble at different places, increasing the size, weight, and power (SWaP) requirements. We propose Reconfigurable Quadrupedal-Bipedal Snake Robots (ReQuBiS), which can transform between mobility modes without rearranging modules. Hence, requiring just a single modification mechanism. Furthermore, our design allows the robot to split into two agents to perform tasks in parallel for biped and snake mobility. Experimental results demonstrate these mobility capabilities in snake, quadruped, and biped modes and transitions between them.
|
|
17:20-17:40, Paper ThBT8.5 | |
>A Methodology for the Detection of Nitrogen Deficiency in Corn Fields Using High Resolution RGB Imagery |
|
Zermas, Dimitris | Sentera, INC |
Nelson, Henry | University of Minnesota |
Stanitsas, Panagiotis | University of Minnesota |
Morellas, Vassilios | U. of Minnesota |
Mulla, David | University of Minnesota |
Papanikolopoulos, Nikos | University of Minnesota |
Keywords: Agricultural Automation
Abstract: A major component of an efficient farming strategy is the precise detection and characterization of plant deficiencies followed by the proper deployment of fertilizers. Through the thoughtful utilization of modern computer vision techniques, it is possible to achieve positive financial and environmental results for these tasks. This work introduces an automation framework that attempts to address the three main drawbacks of existing approaches: (i) Lack of generality (methods are tuned for specific datasets), (ii) Difficulty to apply in variable field conditions, and (iii) Lack of tool sophistication that limits their applicability. The proposed methodology utilizes drone collected images to detect nitrogen (N) deficiencies in maize fields and assess their severity using low-cost RGB sensors. Results on data from experimental fields support the merits of the proposed methodology with mean average precision for the detection of N-deficient leaves reaching 82.3%.
|
|
17:40-18:00, Paper ThBT8.6 | |
>AIoT-Cloud-Integrated Smart Livestock Surveillance Via Assembling Deep Networks with Considering Robustness and Semantics Availability |
|
Su, Wei-Tsung | Aletheia University |
Jiang, Lin-Yi | National Cheng Kung University |
O, Tang-Hsuan | National Cheng Kung University |
Lin, Yu-Chuan | National Cheng Kung University |
Hung, Min-Hsiung | Chinese Culture University |
Chen, Chao-Chun | National Cheng Kung University |
Keywords: Agricultural Automation, Deep Learning Methods, Software Architecture for Robotic and Automation
Abstract: In this paper, we propose a novel smart livestock surveillance system through cooperation of AIoT (artificial intelligence of things) devices and the cloud computing platform, aiming at providing semantic information via assembling deep networks with AIoT devices of limited resource. The key of the proposed system includes two designs: deep-net assembling as a semantic surveillance service and the expandable-convolutional-block neural network (ECB-Net). The first is a development architecture of the divide-and-conquer philosophy for establishing semantic surveillance systems, and this work provides a concrete instance for promoting deep-net assembling to livestock industries. The second is an AIoT device-friendly neural network for filtering insignificant camera images to achieve high robustness of smart surveillance systems. The technical details from the architecture design to optimal ECB-Net model creation are presented in related sections. Finally, we develop the prototype of the smart livestock surveillance system and deploy it by swine rooms for conducting real-world integrated tests. Testing results reveal the superior performance of our proposed smart livestock surveillance scheme.
|
|
ThBT9 Special Session, St Clair 3A |
|
Simulation Optimization in New Information Age 3 |
Chair: Pedrielli, Giulia | Arizona State University |
Co-Chair: Luo, Jun | Shanghai Jiao Tong University Antai College of Economics & Management |
Organizer: Jia, Qing-Shan | Tsinghua University |
Organizer: Luo, Jun | Shanghai Jiao Tong University Antai College of Economics & Management |
Organizer: Pedrielli, Giulia | Arizona State University |
Organizer: Peng, Yijie | Peking University |
|
16:00-16:20, Paper ThBT9.1 | |
>Exploration Via Distributional Reinforcement Learning with Epistemic and Aleatoric Uncertainty Estimation (I) |
|
Liu, Qi | Harbin Institute of Technology |
Li, Yanjie | Harbin Institute of Technology (Shenzhen) |
Liu, Yuecheng | Harbin Institute of Technology, Shenzhen |
Chen, Meiling | Harbin Institute of Technology |
Lv, Shaohua | Harbin Institute of Technology |
Xu, Yunhong | Harbin Institute of Technology |
Keywords: AI-Based Methods, Machine learning
Abstract: The problem of exploration remains one of the major challenges in deep reinforcement learning (RL). This paper proposes an approach to improve the exploration efficiency for distributional RL. First, this paper proposes a novel method to estimate the epistemic and aleatoric uncertainty for distributional RL using deep ensembles, which is inspired by Bayesian Deep Learning. Second, This paper presents a method to improve the exploration efficiency for deep distributional RL by using estimated epistemic uncertainty. Experimental results show that the proposed approach outperforms the baseline in Atari games.
|
|
16:20-16:40, Paper ThBT9.2 | |
>An Intelligent Control Approach for Heavy Haul Trains Using Deep Reinforcement Learning (I) |
|
Liu, Wentao | Beijing Jiaotong University |
Su, Shuai | Beijing Jiaotong University |
Tang, Tao | Beijing Jiaotong University |
Keywords: Intelligent Transportation Systems, AI-Based Methods, Agent-Based Systems
Abstract: One of the main challenges for the control of the heavy haul train of China is the cyclic air braking strategy on the long steep downward slopes. To address this problem, this paper proposes an intelligent control approach using a deep reinforcement learning algorithm to achieve safe operation, low maintenance costs and high running efficiency. The train control problem is firstly described considering the characteristics of the heavy haul railways of China. Then the cyclic air braking strategy is defined as a Markov decision process (MDP) and the key elements in the reinforcement learning framework are designed. To reduce the overestimation of action values in the Deep-Q-Network (DQN) based method, the Double DQN (DDQN) algorithm is used to solve the train control problem in the paper. The simulation experiments are conducted based on the real-word data of Shuozhou-Huanghua Line and the effectiveness of the DDQN-based approach is illustrated by comparing the performances of the proposed approach with those of the DQN-based method.
|
|
16:40-17:00, Paper ThBT9.3 | |
>Efficient Simulation Optimization with Simulation Learning (I) |
|
Goodwin, Travis | George Mason University |
Xu, Jie | George Mason University |
Chen, Chun-Hung | George Mason University |
Celik, Nurcin | University of Miami |
Keywords: Simulation and Animation, Optimization and Optimal Control, Machine learning
Abstract: Simulation optimization has found great success in automation science and engineering, such as the optimization of manufacturing systems, thanks to its capability to fully account for the complexity and uncertainty in systems. However, it remains a challenge to use simulation optimization in applications where decision time window is very short because of computational efficiency challenge. In this paper, a new framework known as Sequential Allocation using Machine learning Predictions as Light-weight Estimates (SAMPLE) is proposed to address this challenge. SAMPLE utilizes an offline simulation learning phase to train machine learning models using simulation data. When a decision needs to be made, SAMPLE utilizes machine learning predictions under a Bayesian framework to determine optimal allocation of simulation sampling budget. The proposed approach enables fast-time simulation-based decision making for automation systems. SAMPLE is able to work with lightweight machine learning models that may only provide crude approximations but still achieve considerable computational efficiency gain. Numerical experiments with both benchmark test functions and a case study demonstrate the viability of the proposed SAMPLE framework, with significant performance improvement over decision making using only machine learning predictions, or simulations alone.
|
|
17:00-17:20, Paper ThBT9.4 | |
>Season-Dependent Parameter Calibration in Building Energy Simulation (I) |
|
Jeong, Cheoljoon | University of Michigan |
Xu, Ziang | University of Michigan |
Byon, Eunshin | University of Michigan |
Cetin, Kristen | Michigan State University |
Keywords: Calibration and Identification, Building Automation, Optimization and Optimal Control
Abstract: As the energy consumption from residential and commercial buildings makes up approximately three-quarters of the U.S. electricity loads, analyzing building energy consumption behavior becomes essential for effective power grid operation. An accurate physics-based building energy simulator that is built on first principles can predict an individual building's energy response, such as energy consumption and indoor environmental conditions under different weather and operational control scenarios. In the building energy simulator, several parameters that specify building characteristics need to be set a priori. Among those parameters, some parameters are season-dependent, whereas other parameters should be globally employed throughout a year. Existing studies in parameter calibration ignore such heterogeneity, which causes suboptimal calibration results. This study presents a new calibration approach that considers seasonal dependency.
|
|
17:20-17:40, Paper ThBT9.5 | |
>Replica Exchange for Nonconvex Optimization (I) |
|
Dong, Jing | Columbia University |
Tong, Xin | National University of Singapore |
Keywords: Optimization and Optimal Control
Abstract: Gradient descent (GD) is known to converge quickly for convex objective functions, but it can be trapped at local minima. On the other hand, Langevin dynamics (LD) can explore the state space and find global minima, but in order to give accurate estimates, LD needs to run with a small discretization step size and weak stochastic force, which in general slow down its convergence. This paper shows that these two algorithms can ``collaborate" through a simple exchange mechanism, in which they swap their current positions if LD yields a lower objective function. This idea can be seen as the singular limit of the replica-exchange technique from the sampling literature. We show that this new algorithm converges to the global minimum linearly with high probability, assuming the objective function is strongly convex in a neighborhood of the unique global minimum. By replacing gradients with stochastic gradients, and adding a proper threshold to the exchange mechanism, our algorithm can also be used in online settings. We also study non-swapping variants of the algorithm, which achieve similar performance. %We further verify our theoretical results through some numerical experiments, and observe superior performance of the proposed algorithm over running GD or LD alone.
|
|
17:40-18:00, Paper ThBT9.6 | |
>Using Gaussian Processes to Automate Probabilistic Branch & Bound for Global Optimization (I) |
|
Pedrielli, Giulia | Arizona State University |
Huang, Hao | Yuan Ze University |
Zabinsky, Zelda | University of Washington |
Keywords: Optimization and Optimal Control, Machine learning, Foundations of Automation
Abstract: Manufacturing, aerospace, energy and several other industries have witnessed a steep growth of increasingly complex, information rich, devices and systems of devices requiring simulation-based approaches. In fact, most modern systems have such complex behavior that their performance can only be evaluated through, usually computationally expensive, simulations. In such settings, it is of paramount importance to provide solutions with quality guarantees. In this manuscript, we focus on algorithms capable of identifying a level set of solutions in proximity of the global optimum, and specifically on the Probabilistic Branch and Bound (PBnB) method. We propose a new way to automate branching decisions by coupling this method with Gaussian process (GP) estimation. The result is PBnB-GP, where, at each iteration a collection of GPs is used to decide how to branch the input space. PBnB-GP not only returns an estimate of the regions with near-optimal reward (using the power of PBnB), but also a "collection of Gaussian processes" that can produce point estimations for any location in the input space, thus harnessing the power of model-based black-box optimization. We present PBnB-GP for the first time together with preliminary numerical results.
|
|
ThBT10 Special Session, St Clair 3B |
|
Data-Driven Models for Healthcare Operations |
Chair: Zhang, Zheng | Zhejiang University |
Co-Chair: Geng, Na | Shanghai Jiao Tong University |
Organizer: Zhang, Zheng | Zhejiang University |
Organizer: Geng, Na | Shanghai Jiao Tong University |
|
16:00-16:20, Paper ThBT10.1 | |
>Simulation-Based Metaheuristic for Bed Allocation under Clustered Overflow Configuration (I) |
|
Gong, Xuran | Shanghai Jiao Tong University |
Geng, Na | Shanghai Jiao Tong University |
Keywords: Health Care Management, Modelling, Simulation and Optimization in Healthcare, Clinical and Operational Decision Support
Abstract: Inpatient beds are typically one of the most fundamental and significant resources in a hospital. Two different ways to organize inpatient beds, i.e. dedicated and flexible configurations, are adopted in hospitals to reduce operational cost and provide high-quality service. This work considers a generalized configuration named clustered overflow configuration, i.e. all the specialties are partitioned into several clusters; each specialty has its dedicated beds and the specialties in the same cluster share a number of overflow beds. A mixed-integer stochastic programming model is constructed to optimize the partition of specialties and bed allocation among clusters. The objective is to minimize the weighted cost of rejecting and holding patients waiting and nursing cost. A simulation-based metaheuristic approach is proposed to solve the model. A niching genetic algorithm framework is adopted to identify good partitions of specialties. A simulation optimization algorithm known as adaptive hyperbox algorithm is used to optimize the corresponding bed allocation problem given a partition of specialties. A set covering model is constructed to screen the cluster pool and exploit the missed information during the search process. Using the real data collected from a public hospital in Shanghai, a case study is conducted. The numerical results show the efficiency of the proposed method.
|
|
16:20-16:40, Paper ThBT10.2 | |
>Decomposed Simulation Optimization of Cancer Screening Strategies with Random Observations (I) |
|
Zhang, Zheng | Zhejiang University |
Peng, Yijie | Peking University |
Keywords: Modelling, Simulation and Optimization in Healthcare, Health Care Management, AI and Machine Learning in Healthcare
Abstract: This paper considers a two-phase cancer screening strategy optimization problem with multiple latent states and random observations. We propose a threshold policy to determine whether or not to perform a confirmative test in each period of screening based on random observations received from the initial tests. We formulate this problem as a tailored simulation optimization model based on decomposition and reformulation techniques, and we further show that this model has several nice properties that can be utilized to improve computational performance.
|
|
16:40-17:00, Paper ThBT10.3 | |
>Joint Optimization of Patient Delay Announcement and Adaptive Admission Policies Based on Reinforcement Learning (I) |
|
Ma, Yinghua | Zhejiang University |
Zhang, Zheng | Zhejiang University |
Lu, Yuwei | Guangxi University of Science and Technology |
|
|
17:00-17:20, Paper ThBT10.4 | |
>Allocation of Telemedicine Capacity in a Regional Hierarchical System for Community-Based Care (I) |
|
Li, Nan | Tsinghua University |
Yuan, Fang | Department of Information, the First People's Hospital of Yinchu |
Ma, XiaoFei | Department of Information, the First People's Hospital of Yinchu |
Liu, Xiang | Tsinghua University |
Keywords: Modelling, Simulation and Optimization in Healthcare, Health Care Management, Scheduling in Healthcare
Abstract: The development of telemedicine has provided new service modes, which enables providers to deliver care more efficiently. Telemedicine visits can be hosted or non-hosted, depending on the level of care and resources required. A hosted visit is intended for more sophisticated care, where an upper-level provider together with a lower-level provider jointly deliver care, whereas a non-hosted visit involves one provider and one patient, which is intended to resolve minor conditions. To date, limited studies have incorporated both types of telemedicine. This paper studies the resource allocation problem in a hierarchical health system by leveraging two types of telemedicine. We study how providers shall balance their resources between office visits and telemedicine visits across multiple communities in the system. We formulate a Partially Observable Markov Decision Process (POMDP) and develop a genetic algorithm that allocates appointment slots among heterogeneous communities. The genetic algorithm consistently outperforms naive strategies in our numerical studies. This study sheds light for policy makers and providers on how to balance their resources between online and offline in the new telemedicine era.
|
|
17:20-17:40, Paper ThBT10.5 | |
>Discovery of Mental Wellness Via Social Analytics for Liveability in an Urban City (I) |
|
Tan, Kar Way | Singapore Management University |
Keywords: Human Factors in Healthcare, Machine learning, Health Care Management
Abstract: Smart cities, are often perceived as urban areas that use technologies to manage resources, improve economy and enhance community livelihood. In this paper, we share an approach which uses multiple sources of data for evidence-based analysis of the public's views, concerns and sentiments on the topic related to mental wellness. We hope to bring forth a better understanding of the existing concerns of the citizens and available social support. Our study leverages on social sensing via text mining and social network analysis to listen to the voices of the citizens through revealed content from web data sources, such as social media and public forums. By using hybrid data sources, we present the important considerations for mining inherent mental wellness concerns faced by the citizens. The outcome of the analysis includes, both the positive and negative sentiments towards mental wellness and draws relations to national level performance indicators relating to mental wellness. We hope our research could help authorities derive actionable plans for designing health services or public events that bring positive social mixing and happiness by addressing the mental wellness of the residents.
|
|
17:40-18:00, Paper ThBT10.6 | |
>Does Deep Learning REALLY Outperform Non-Deep Machine Learning for Clinical Prediction on Physiological Time Series? |
|
Liao, Ke | Japan Advanced Institute of Science and Technology / RICOH |
Wang, Wei | Ricoh Software Research Center (Beijing) |
Elibol, Armagan | Japan Advanced Institute of Science and Technology |
Meng, Lingzhong | Yale University School of Medicine |
Zhao, Xu | Central South University |
Chong, Nak Young | Japan Advanced Inst. of Sci. and Tech |
Keywords: AI and Machine Learning in Healthcare, Diagnosis and Prognostics
Abstract: Machine learning has been widely used in healthcare applications to approximate complex models, for clinical diagnosis, prognosis, and treatment. As deep learning has the outstanding ability to extract information from time series, its true capabilities on sparse, irregularly sampled, multivariate, and imbalanced physiological data are not yet fully explored. In this paper, we systematically examine the performance of machine learning models for the clinical prediction task based on the EHR, especially physiological time series. We choose Physionet 2019 challenge public dataset cite{reyna2019early} to predict Sepsis outcomes in ICU units. Ten baseline machine learning models are compared, including 3 deep learning methods and 7 non-deep learning methods, commonly used in the clinical prediction domain. Nine evaluation metrics with specific clinical implications are used to assess the performance of models. Besides, we sub-sample training dataset sizes and use learning curve fit to investigate the impact of the training dataset size on the performance of the machine learning models. We also propose the general pre-processing method for the physiology time-series data and use Dice Loss cite{vicar2019sepsis} to deal with the dataset imbalanced problem. The results show that deep learning indeed outperforms non-deep learning, but with certain conditions: firstly, evaluating with some particular evaluation metrics (AUROC, AUPRC, Sensitivity, and FNR), but not others; secondly, the training dataset size is large enough (with an estimation of a magnitude of thousands).
|
|
ThBT11 Regular Session, St Clair 4 |
|
Human-Centered Automation |
Chair: Das, Jnaneshwar | Arizona State University |
Co-Chair: Liu, Rui | Kent State University |
|
16:00-16:20, Paper ThBT11.1 | |
>An Attention Transfer Model for Human-Assisted Failure Avoidance in Robot Manipulations |
|
Song, Boyi | Cognitive Robotics and AI Lab (CRAI), College of Aeronautics And |
Peng, Yuntao | Cognitive Robotics and AI Lab (CRAI), Kent State University |
Luo, Ruijiao | Cognitive Robotics and AI Lab (CRAI), Kent State University |
Liu, Rui | Kent State University |
Keywords: Human-Centered Automation, Cognitive Automation, Task Planning
Abstract: Due to real-world dynamics and hardware uncertainty, robots inevitably fail in task executions, resulting in undesired or even dangerous executions. In order to avoid failures and improve robot performance, it is critical to identify and correct abnormal robot executions at an early stage. However, due to limited reasoning capability and knowledge storage, it is challenging for robots to self-diagnose and -correct their own abnormality in both planning and executing. To improve robot self-diagnosis capability, in this research a novel human-to-robot attention transfer (H2R-AT) method was developed to identify robot manipulation errors by leveraging human instructions. H2R-AT was developed by fusing attention mapping mechanism into a novel stacked neural networks model, transferring human verbal attention into the robot visual attention. With the attention transfer, a robot understands what and where human concerns are to identify and correct abnormal manipulations. Two representative task scenarios: ``serve water for a human in a kitchen" and ``pick up a defective gear in a factory" were designed in a simulation framework CRAIhri with abnormal robot manipulations; and 252 volunteers were recruited to provide about 12000 verbal reminders to learn and test H2R-AT. The method effectiveness was validated by the high accuracy of 73.68% in transferring attention, and the high accuracy of 66.86% in avoiding grasping failures.
|
|
16:20-16:40, Paper ThBT11.2 | |
>Variable Admittance Control with Robust Adaptive Velocity Control for Dynamic Physical Interaction between Robot, Human and Environment |
> Video
|
|
Li, Hsieh-Yu | Singapore University of Technology and Design |
Paranawithana, Ishara | Singapore University of Technology and Design |
Yang, Liangjing | Zhejiang University |
Tan, U-Xuan | Singapore University of Techonlogy and Design |
Keywords: Human-Centered Automation, Human Factors and Human-in-the-Loop, Collaborative Robots in Manufacturing
Abstract: In the field of collaborative robot whose end-effector is complaint in response to human force, the human operator is able to execute tasks based on their judgement. The advantages include human power augmentation and utilization of human's judgement. A force/torque sensor is usually mounted on the end -effector to sense the human force, which is sent to an interaction controller to yield velocity command for the robot to follow. However, when human guides the end-effector to contact an extra dynamic environment that moves in multiple directions, it poses challenges on controller design. The interaction controller needs to provide the stable interaction for human operator to contact such a dynamic contact force (due to the moving environment) while to provide compliant cooperation for the human to follow the motion of the environment. To address these issues, we propose variable admittance control based on robust adaptive velocity control. The human and environmental forces are considered in the variable admittance controller to obtain the suitable interaction between robot, human, and environment. Additionally, robust adaptive control is employed to enhance the tracking performance. Experiments are conducted and the results show the proposed controller enables the human operator to stably contact a dynamic environment and compliantly track the motion of it.
|
|
16:40-17:00, Paper ThBT11.3 | |
>Terrain-Relative Diver Following with Autonomous Underwater Vehicle for Coral Reef Mapping |
|
Antervedi, Lakshmi Gana Prasad | Arizona State University |
Chen, Zhiang | Arizona State University |
Anand, Harish | Arizona State University |
Martin, Roberta | Arizona State University |
Arrowsmith, Ramon | Arizona State University |
Das, Jnaneshwar | Arizona State University |
Keywords: Human-Centered Automation, Motion and Path Planning, Deep Learning in Robotics and Automation
Abstract: Coral reef mapping is an indispensable step in coral conservation efforts across the globe. Monitoring reefs at regular intervals helps conservationists understand and address the problems causing coral reef degradation. Autonomous Underwater Vehicles (AUVs) have a tremendous potential to assist humans in these efforts. Delegating mapping and measurement acquisition tasks to AUVs would not only limit the number of human divers required for the missions but could also improve the quality of the maps developed. Consistency in imagery and spectroscopic measurements could be significantly improved by keeping the imagery payload at a fixed distance from the reefs to reduce heteroscedasticity in the measurements. To this end, we present a terrain-relative diver following system for an AUV that can follow a human diver while maintaining a fixed distance from the terrain. Our proposed system consists of separate modules for diver detection, tracking, and terrain following. We extensively tested our system in Gazebo simulation environment with three different terrain models, including a terrain model of a coral reef in Honaunau Bay, Hawaii. To the best of our knowledge, this is the first diver following system that also carries out terrain-relative navigation, ensuring minimal variation of distance to the terrain. We have released the code for our system, and the datasets used in the detection module.
|
|
17:00-17:20, Paper ThBT11.4 | |
>Development of a Human-Centered CPPS Framework for Robotic Micro-Assembly |
|
Mulet Alberola, Jose Antonio | STIIMA-CNR |
Liyanapathirana, Buddhi Charitha | University of Brescia |
Fassi, Irene | CNR |
Keywords: Human-Centered Automation, Cyber-physical Production Systems and Industry 4.0, Automation at Micro-Nano Scales
Abstract: This study aims at discussing a robotic framework that allows the implementation of a Cyber-Physical Production System (CPPS) for the automated micro-/nano-manipulation, considering the human role from its very first conception. For the electronic manufacturing domain, the CPPS deployment represents a powerful environment for achieving a more efficient production with higher throughputs while reaching sustainable objectives. The specific use case concerns the PCB and FPCB assembly operations, performing precise positioning - both, rigid micro components and flexible macro ones. The manufacturing line is defined by a) precise robot manipulators for a complete automated operation, and b) a collaborative robot able to work hand-to-hand with human operators. We also propose a model to set different metrics to monitor the performance of the system. Challenges faced during the development of the framework in this specific case arise from i) the high accuracy required along a large range of motion, ii) the ability for the manipulation of the components, iii) the involvement of humans during manufacturing tasks, and iv) the manufacturing necessity for a smarter, more adaptable and scalable system.
|
|
17:20-17:40, Paper ThBT11.5 | |
>Adaptive Mechanomyogram Hand Gesture Recognition in Online and Repeatable Environment |
|
Wattanasiri, Panipat | Imperial College London |
Wilson, Samuel | Imperial College London |
Huo, Weiguang | Imperial College London |
Lewis, Alex | SERG Technologies Ltd, United Kingdom |
Christos, Kapatos | Serg Technologies |
Vaidyanathan, Ravi | Imperial College London |
Keywords: Embedded Systems for Robotic and Automation, Human-Centered Automation, Human Factors and Human-in-the-Loop
Abstract: We introduce a complete architecture for real-time hand gesture recognition for human-computer interface (HCI) and robotic control. The system addresses ease of use, calibration, and robustness issues which have inhibited gesture recognition wearables in the field. Our system is packaged as a generic (non-customized) arm wearable that integrates: 1) a novel mechanomyogram (MMG) sensing suite; 2) an integrated inertial measurement unit (IMU); 3) accompanying data acquisition and transmission hardware; and 4) real-time signal recognition algorithms to run on the receiving peripheral (e.g. computer, robot, etc.). We implement a rapid training routine capable of grasp pattern identification from small samples (20 per gesture) with less than 5-minute calibration time, which yields immediate real-time accuracies of 84% in amputees (3 gestures) and 89% in non-amputees (5 gestures), with the capacity to scale as users become more comfortable (accurate) with generated gestures. In repeated (5-day) usage with regular donning and doffing of the armband, 89% - 91% accuracy is achieved with non-amputees using data over the previous days for reparameterization. Findings demonstrate the capacity to adapt to new able-bodied and amputee subjects with a generic armband and small training datasets, adapt as user proficiency increases, and provide consistent prediction for regular long-term use.
|
|
17:40-18:00, Paper ThBT11.6 | |
>Human Joint Torque Modelling with MMG and EMG During Lower Limb Human-Exoskeleton Interaction |
|
Caulcrick, Christopher | Imperial College London |
Huo, Weiguang | Imperial College London |
Hoult, Will | McLaren Applied, McLaren Technology Centre |
Vaidyanathan, Ravi | Imperial College London |
Keywords: Human-Centered Automation
Abstract: Human-robot cooperation is vital for optimising powered assist of lower limb exoskeletons (LLEs). Robotic capacity to intelligently adapt to human force, however, demands a fusion of data from exoskeleton and user state for smooth human-robot synergy. Muscle activity, mapped through electromyography (EMG) or mechanomyography (MMG) is widely acknowledged as usable sensor input that precedes the onset of human joint torque. However, competing and complementary information between such physiological feedback is yet to be exploited, or even assessed, for predictive LLE control. We investigate complementary and competing benefits of EMG and MMG sensing modalities as a means of calculating human torque input for assist-as-needed (AAN) LLE control. Three biomechanically agnostic machine learning approaches, linear regression, polynomial regression, and neural networks, are implemented for joint torque prediction during human-exoskeleton interaction experiments. Results demonstrate MMG predicts human joint torque with slightly lower accuracy than EMG for isometric human-exoskeleton interaction. Performance is comparable for dynamic exercise.} Neural network models achieve the best performance for both MMG and EMG (94.8±0.7% with MMG and 97.6±0.8% with EMG (Mean±SD)) at the expense of training time and implementation complexity. This investigation represents the first MMG human joint torque models for LLEs and their first comparison with EMG. We provide our implementations for future investigations (https://github.com/cic12/ieee_appx).
|