|
TuAT1 Regular Session, Auditorium |
|
Intelligent and Flexible Manufacturing 1 |
Chair: Fanti, Maria Pia | Politecnico Di Bari |
Co-Chair: Guidetti, Xavier | ETH Zürich |
|
10:30-10:50, Paper TuAT1.1 | |
>LSAT: Specification and Analysis of Product Logistics in Flexible Manufacturing Systems |
|
van der Sanden, Bram | TNO |
Blankenstein, Yuri | TNO |
Schiffelers, Ramon | ASML |
Voeten, Jeroen | Eindhoven University of Technology |
Keywords: Intelligent and Flexible Manufacturing, Planning, Scheduling and Coordination, Formal Methods in Robotics and Automation
Abstract: LSAT (Logistics Specification and Analysis Tool) is a tool for rapid design-space exploration of supervisory controllers that steer the product logistics and orchestrate the behavior in flexible manufacturing systems. LSAT enables lightweight modeling of system resources, system behavior, and timing characteristics. The tool provides various visualizations to explore the controlled system behavior and analysis and optimization techniques to improve the system performance. Compared to existing approaches, LSAT provides concise modeling using languages tailored towards the application domain, with domain concepts are elements of the language. LSAT provides efficient performance analysis by exploiting the structure of the models. In this paper, we describe the rationale for developing LSAT and position it with respect to other performance modeling and analysis tools. We illustrate the benefits of LSAT with an example system.
|
|
10:50-11:10, Paper TuAT1.2 | |
>Iterative Surface Mapping Using Local Geometry Approximation with Sparse Measurements During Robotic Tooling Tasks |
> Video
|
|
Amersdorfer, Manuel | Christian-Albrechts-Universität Zu Kiel |
Meurer, Thomas | Institute for Automatic Control, Kiel University |
Keywords: Intelligent and Flexible Manufacturing, Manufacturing, Maintenance and Supply Chains, Data fusion
Abstract: We present a cost-efficient and versatile method to map an unknown 3D freeform surface using only sparse measurements while the end-effector of a robotic manipulator moves along the surface. The geometry is locally approximated by a plane, which is defined by measured points on the surface. The method relies on linear Kalman filters, estimating the height of each point on a 2D grid. Therefore, the approximation covariance for each grid point is determined using a radial basis function to consider the measured point positions. We propose different update strategies for the grid points exploiting the locality of the planar approximation in combination with a projection method. The approach is experimentally validated by tracking the surface with a robotic manipulator. Three laser distance sensors mounted on the end-effector continuously measure points on the surface during the motion to determine the approximation plane. It is shown that the surface geometry can be mapped reasonably accurate with a mean absolute error below 1 mm. The mapping error mainly depends on the size of the approximation area and the curvature of the surface.
|
|
11:10-11:30, Paper TuAT1.3 | |
>Reinforcement Learning Based Optimization of Bayesian Networks for Generating Feasible Vehicle Configuration Suggestions |
|
Duerr, Simon | Dr. Ing. H.c. F. Porsche AG |
Lamprecht, Raphael | Fraunhofer Institute for Manufacturing Engineering and Automatio |
Kauffmann, Matthias | Porsche AG |
Huber, Marco F. | University of Stuttgart |
Keywords: Intelligent and Flexible Manufacturing, Probability and Statistical Methods, Reinforcement
Abstract: A promising method in the automotive industry to anticipate future customer demands is the concept of planned orders. Due to multi-variant products, changing customer demands, and dynamic environments the process of generating planned orders is challenging. This paper introduces an approach using graphical models to generate planned order suggestions in a multi-variant order management process. Bayesian networks are modelled by learning the structure from different data sources, which enable the possibility to directly sample configuration suggestions. To find an optimized graph structure, a method using hierarchical correlation clustering and reinforcement learning is applied, taking into account technical and sales-operated feasibility constraints. The method has high potential in practical usage and is evaluated by a real-world use case of the Dr. Ing. h.c. F. Porsche AG.
|
|
11:30-11:50, Paper TuAT1.4 | |
>Industrial Time Series Modeling with Causal Precursors and Separable Temporal Convolutions |
|
Menegozzo, Giovanni | University of Verona |
Dall'Alba, Diego | University of Verona |
Fiorini, Paolo | University of Verona |
Keywords: Intelligent and Flexible Manufacturing, AI-Based Methods, Deep Learning Methods
Abstract: When applied to industrial processes, predictive models (PMs) easily fail to depict the overall complexity of a production plant. Multiple factors can interfere during the production process and the requirements in accuracy, safety and efficiency are very high. To fully achieve the potential of PMs, it is useful to integrate typical characteristics of the manufacturing process and build procedures that account for the domain peculiarities. Cause-effect relationships that entail the flow of an industrial process are rarely considered in PMs for manufacturing despite the relevance that these relationships have in practice. In this work we present a two-step procedure that uses a causal discovery method for time series, eventually validated by domain experts, together with a new neural network architecture named Separable Temporal Convolutional Network (S-TCN) to create a PM for industrial multivariate time series. Causal precursors exploit the sequentiality of the process flow, grouping process machines by their effective temporal activation. The S-TCN architecture is based on separable temporal convolution and enables the efficient forecast of more distant temporal connections than common temporal models. A numerical validation is presented on a large class of synergetic nonlinear stochastic processes. We apply the proposed procedure in an ultra-processed food plant collecting more than 100 days of active production. The presented procedure achieves better results when compared with state-of-the-art algorithms in both conditions.
|
|
11:50-12:10, Paper TuAT1.5 | |
>Plasma Spray Process Parameters Configuration Using Sample-Efficient Batch Bayesian Optimization |
|
Guidetti, Xavier | ETH Zürich |
Rupenyan, Alisa | ETH Zürich |
Fassl, Lutz | Equipment Digitalization Team, Oerlikon Metco |
Nabavi, Majid | Equipment Digitalization Team, Oerlikon Metco |
Lygeros, John | ETH Zurich |
Keywords: Process Control, Probability and Statistical Methods, Intelligent and Flexible Manufacturing
Abstract: Recent work has shown constrained Bayesian optimization to be a powerful technique for the optimization of industrial processes. In complex manufacturing processes, the possibility to run extensive sequences of experiments with the goal of finding good process parameters is severely limited by the time required for quality evaluation of the produced parts. To accelerate the process parameter optimization, we introduce a parallel acquisition procedure tailored on the process characteristics. We further propose an algorithm that adapts to equipment status to improve run-to-run reproducibility. We validate our optimization method numerically and experimentally, and demonstrate that it can efficiently find input parameters that produce the desired outcome and minimize the process cost.
|
|
12:10-12:30, Paper TuAT1.6 | |
>An Architecture for Digital Processes in Manufacturing with Blockchain, Docker and Cloud Storage |
|
Volpe, Gaetano | Politecnico Di Bari |
Mangini, Agostino Marcello | Politecnico Di Bari |
Fanti, Maria Pia | Politecnico Di Bari |
Keywords: Intelligent and Flexible Manufacturing, Factory Automation, Manufacturing, Maintenance and Supply Chains
Abstract: The Blockchain is one of the last decade emerging technologies in software architectures. Its nature of a distributed ledger database allowing verifiable and tamper-proof transactions between untrusted parties makes it suitable for a vast class of domains concerning business processes, including Cloud Manufacturing, a new paradigm for the manufacturing industry based on cloud technologies, for which decentralization and security are key factors. However, existing solutions are still weak in terms of collaboration in providing and consuming heterogeneous services in the Cloud, therefore a standard framework is necessary to overcome this limit. In this paper, we suggest an architecture for consuming digital processes in a manufacturing environment, based on Blockchain and Smart Contracts. Our primary contribution is the integration of Blockchain with two other popular technologies: Docker, a highly portable and scalable container-based platform to run applications, and Cloud Storage. In this system, the logic of a single process is defined by the owner in a self-contained Docker image, whose digest is safely stored in the chain, while input and output files can be stored in a traditional cloud storage service. On each new consumer request, the best runner node is selected through the solution of a simple task assignment problem with a deep learning approach.
|
|
TuAT2 Regular Session, Rhone 1 |
|
Industrial Robots |
Chair: Yi, Jingang | Rutgers University |
Co-Chair: Romanov, Alexey | Moscow Technological University (MIREA) |
|
10:30-10:50, Paper TuAT2.1 | |
>Automatic Generation of Kinematics and Dynamics Model Descriptions for Modular Reconfigurable Robot Manipulators |
|
Nainer, Carlo | Fraunhofer Italia Research |
Feder, Maddalena | Fraunhofer Italia Research |
Giusti, Andrea | Fraunhofer Italia Research |
Keywords: Cellular and Modular Robots, Industrial Robots
Abstract: We propose a unified approach for automatically generating multiple robot-model descriptions for modular robot manipulators. Modular robots need models for kinematics and dynamics to deploy motion control schemes with high performance as it is possible for their fixed structure counterparts. Additionally, these models are also necessary for enabling the optimization of their assembly to optimally meet task and environmental constraints. Manual derivation of models for every different assembly is impractical as the number of possible arrangements can be large. We propose a systematic approach for characterizing single modules and for automatically generating the following widely adopted robot model descriptions for kinematics and dynamics: standard and modified Denavit-Hartenberg, product of exponentials formulation, and the unified robot description format. We numerically verify our approach also considering experimental data from a real modular robot manipulator.
|
|
10:50-11:10, Paper TuAT2.2 | |
>SABER: Modular Reconfigurable Robot for Industrial Applications |
|
Romanov, Alexey | Moscow Technological University (MIREA) |
Yashunskiy, Vladimir | MIREA - Russian Tehcnological University |
Chiu, Wei-Yu | National Tsing Hua University |
Keywords: Industrial Robots
Abstract: This study proposes a heterogeneous modular reconfigurable robot called SABER that is suitable for industrial applications. The SABER is a Step, Assembler, Bridge, Explorer Robot comprising a platform and a reconfigurable rail; it can work in three different configurations: monowheel, rail trolley, and manipulator. The monowheel configuration provides locomotion with a speed up to 10 km/h. The rail trolley configuration allows the robot to overcome gaps larger than its wheel diameter, climb steps as large as the wheel radius, and pass through passages smaller than the wheel diameter. The manipulator configuration enables the robot to manipulate objects using two robotic arms with changeable tools. This design allows several SABERs to work cooperatively, extending capabilities of their mechanics or enlarging the working area. As such, SABERs can address various tasks, including material handling, assembly, diagnostic, and maintenance applications, revealing a concept of universal robots for future factories. Robot performance in terms of the speed, acceleration, oscillation, and gap/step handling was numerically analyzed, showing that SABER is comparable with or better than existing autonomous mobile robots and modular reconfigurable robots.
|
|
11:10-11:30, Paper TuAT2.3 | |
>Suction Point Selection Algorithm Based on Point Cloud for Plastic Waste Sorting |
|
Um, Sangwoo | Korea Advanced Institute of Science and Technology |
Kim, Kyung-Soo | KAIST(Korea Advanced Institute of Science and Technology) |
Kim, Soohyun | KAIST(Korea Advanced Institute of Science and Technology) |
Keywords: Industrial Robots, Computer Vision for Automation
Abstract: This study proposes a novel decision-making algorithm pertaining to picking locations for a suction grasping robot. The algorithm has been developed for plastic waste sorting facilities. Suction grippers are widely used but it becomes difficult for the grippers to pick up randomly crumpled objects. An optimal grasping point selection based on the quantitative evaluation of grasp quality pertaining to estimated contact zone and adjustable cost function has been proposed to solve the aforementioned problem. The geometrical characteristic of the suction cup is used for the donut-shaped evaluation group. The algorithm processes the input point cloud based on candidate generation, candidate evaluation, and the selection of final picking point for robustness. The algorithm structure and underlying logic is explained in detail. The performance change with respect to the cost function, suction cup type, and estimated contact zone has been experimentally investigated. The algorithm achieves a high success rate in terms of suction grasping, and thus, parallel manipulators can rapidly execute pick and place motion for efficient sorting. This study will help expand the use of robotic arms with suction grippers by enabling a wider span of objects to be processed.
|
|
11:30-11:50, Paper TuAT2.4 | |
>Real-Time Estimation of Multiple Potential Contact Locations and Forces |
|
Popov, Dmitry | Innopolis University |
Klimchik, Alexandr | Innopolis University |
Pashkevich, Anatol | Ecole Des Mines De Nantes |
Keywords: Industrial Robots, Human-Centered Automation, Human Factors and Human-in-the-Loop
Abstract: In this work, we propose a contact point localization and external force estimation algorithm for collaborative robots. In comparison with existing approaches, the proposed algorithm can detect and evaluate multiple solutions and more than 3x faster without loss of accuracy. To achieve real-time performance, a hierarchical robot representation and surface mesh preprocessing was used, allowing us to achieve a 50x speedup in a run-time, compared to checking all contact points. Mesh preprocessing includes two-step clustering in the space of vertices normals vectors and vertices positions. The localization method was tested in a simulated Kinova Jaco 2 and real KUKA iiwa LBR 14 collaborative robots. Our solution allows estimating the contact point on the robot surface with 2.3 cm average accuracy in a more than 600 Hz loop.
|
|
11:50-12:10, Paper TuAT2.5 | |
>CoHaptics: Development of Human-Robot Collaborative System with Forearm-Worn Haptic Display to Increase Safety in Future Factories |
> Video
|
|
Altamirano Cabrera, Miguel | Skolkovo Institute of Science and Technology (Skoltech), Moscow, |
Heredia Mena, Juan Esteban | University of Southern Denmark |
Tirado Rosero, Jonathan | Skolkovo Institute of Science and Technology (Skoltech), Moscow, |
Panov, Vladislav | Skolkovo Institute of Science and Technology (Skoltech), Moscow, |
Hagos, Fikre | Skolkovo Institute of Science and Technology Skoltech |
Tsetserukou, Dzmitry | Skolkovo Institute of Science and Technology |
Keywords: Human Factors and Human-in-the-Loop, Industrial Robots, Factory Automation
Abstract: Complex tasks require human collaboration since robots do not have enough dexterity. However, robots are still used as instruments and not as collaborative systems. We are introducing a framework to ensure safety in a human-robot collaborative environment. The system is composed of a haptic feedback display, low-cost wearable mocap, and a new collision avoidance algorithm based on the Artificial Potential Fields (APF). Wearable optical motion capturing system enables tracking the human hand position with high accuracy and low latency on large working areas. This study evaluates whether haptic feedback improves safety in human-robot collaboration. Three experiments were carried out to evaluate the performance of the proposed system. The first one evaluated human responses to the haptic device during interaction with the Robot Tool Center Point (TCP). The second experiment analyzed human-robot behavior during an imminent collision. The third experiment evaluated the system in a collaborative activity in a shared working environment. This study had shown that when haptic feedback in the control loop was included, the safe distance (minimum robot-obstacle distance) increased by 4.1 cm from 12.39 cm to 16.55 cm, and the robot's path, when the collision avoidance algorithm was activated, was reduced by 81%.
|
|
12:10-12:30, Paper TuAT2.6 | |
>Softness-Adaptive Pinch-Grasp Strategy Using Fingertip Tactile Information of Robot Hand |
> Video
|
|
Park, Sungwoo | Korea University, Korea Institute of Science and Technology |
Hwang, Donghyun | Korea Institute of Science and Technology |
Keywords: Industrial Robots
Abstract: We develop a tactile information-based pinch–grasp strategy enabling a robot hand to adaptively grasp easily deformable soft objects. When a robot hand has to perform grasping tasks, the grasp planner develops the grasping strategy based on visual information. However, the intrinsic properties of the target object, such as softness, cannot be detected appropriately using only visual feedback. To overcome this fundamental limitation, we aim to develop a softness-adaptive pinch–grasp strategy using fingertip tactile information. To achieve this, we first categorize soft objects based on the characteristic of resistance to deformation. Moreover, we design a three-dimensional tactile sensor that provides tactile information by measuring and localizing the distributed forces induced on its fingertip. In devising the adaptive grasp strategy, we focus on developing an algorithm that enables a robot hand to grasp soft objects by minimizing object deformation by controlling the pinch force based on the tactile feedback. The experimental results demonstrate that object deformation can be reduced by up to approximately 83.5% by the proposed strategy.
|
|
TuAT3 Regular Session, Rhone 2 |
|
Logistics |
Chair: Absi, Nabil | Ecole Des Mines De Saint Etienne |
Co-Chair: Kokot, Mirko | University of Zagreb |
|
10:30-10:50, Paper TuAT3.1 | |
>Hierarchical Clustering-Based Network Design Algorithm for Many-To-Many Hub Location Routing Problem |
|
Seto, Akane | Hitachi, Ltd |
Uyama, Kazuya | Hitachi.Ltd |
Hosoda, Junko | Hitachi, Ltd |
Miyashita, Naoko | Hitachi, Ltd |
Keywords: Logistics, Optimization and Optimal Control
Abstract: The many-to-many hub location routing problem (MMHLRP) has been attracting interest as a way to improve the efficiency of long-distance delivery. Since more commodities are delivered to urban areas than to rural areas, it is important in actual business to improve the efficiency of deliveries in which the commodities are unevenly distributed. In this work, we propose a network design algorithm utilizing customized hierarchical clustering for MMHLRP. The results of numerical experiments show that the proposed algorithm reduces the total cost compared to the conventional gravity rule based clustering algorithm when the commodities are distributed unevenly.
|
|
10:50-11:10, Paper TuAT3.2 | |
>Sizing of a Heterogeneous Fleet of Robots in a Logistics Warehouse |
|
Rjeb, Achraf | LIMOS UCA |
Gayon, Jean-Philippe | LIMOS, INP Clermont Auvergne |
Norre, Sylvie | LIMOS UCA |
Keywords: Logistics, Autonomous Vehicle Navigation
Abstract: We are interested in a fleet sizing problem with several types of robots. The operations are divided into several phases: loading, loaded travel, unloading and empty travel. The purpose is to determine the size of the robot fleet, that is, to determine the optimal types and number of robots for each type. We first consider the case where loads are all of the same type (homogeneous) before generalizing to several types of loads (heterogeneous). In both cases, we show that the problem can be formulated by an integer linear program. In the case of homogeneous loads, we consider a relaxation of the problem where the number of robots can be a real number. For this approximation, we show that it is optimal to use a single type of robot. It is also near-optimal when the number of robots must be an integer number.
|
|
11:10-11:30, Paper TuAT3.3 | |
>A Mixed Integer Programming Formulation for the Truck Drivers Scheduling Problem Considering the European Union Drivers Rules |
|
Pena Arenas, Ivan Guillermo | Université Clermont Auvergne |
Garaix, Thierry | Ecole Nationale Superieure De Mines De Saint Etienne |
Lacomme, Philippe | LIMOS |
Tchernev, Nikolay | Université Clermont Auvergne |
Keywords: Logistics, Planning, Scheduling and Coordination, Optimization and Optimal Control
Abstract: The legal driver rules in the European Union defines a general framework with restrictions in driving time or working time between breaks and rest periods. These constraints must be addressed for an evaluation of any vehicle trip that met the regulation rules. For many real-world applications, the final trips have to satisfy these rules. The set of EU rules is the most complex design of trips and extension to other rules should be easy following the model we propose here. Compared to previous contributions we provide a new mixed integer linear program, which includes all the weekly rules to schedule when the sequence of visits is fixed. We also provide a new benchmark with detailed optimal solutions. Based on a set of numerical experiments, we discuss the relevance of different simplifications in the model used in the literature.
|
|
11:30-11:50, Paper TuAT3.4 | |
>A Lagrangian Column Generation Approach for the Probabilistic Crowdsourced Logistics Planning |
|
Han, Chung-Kyun | LG Display |
Cheng, Shih-Fen | Singapore Management University |
Keywords: Logistics, Planning, Scheduling and Coordination, Optimization and Optimal Control
Abstract: In recent years we have seen the movement for the retail industry to move their operations online. Along the process, it has created brand new patterns for the fulfillment service, and the logistics service providers serving these retailers have no choice but to adapt. The most challenging issues faced by all logistics service providers are the highly fluctuating demands and the shortening response times. All these challenges imply that maintaining a fixed fleet will either be too costly or insufficient. One potential solution is to tap into the crowdsourced workforce. However, existing industry practices of relying on human planners or worker's self-planning have been shown to be inefficient and laborious. In this paper, we introduce a centralized planning model for the crowdsourced logistics delivery paradigm, considering individual worker's spatio-temporal preferences. Considering worker's spatio-temporal preferences is important for the planner as it could significantly improve crowdsourced worker's productivity. Our major contributions are in the formulation of the problem as a mixed-integer program and the proposal of an efficient algorithm that is based on the column generation and the Lagrangian relaxation frameworks. Such a hybrid approach allows us to overcome the difficulty encountered separately by the classical column generation and Lagrangian relaxation approaches. By using a series of real-world-inspired numerical instances, we have demonstrated the effectiveness of our approach against classical column generation and Lagrangian relaxation approaches, and a decentralized, agent-centric greedy approach. Our proposed hybrid approach is scalable to large problem instances, with reasonable solution quality, and achieves better allocation fairness.
|
|
11:50-12:10, Paper TuAT3.5 | |
>Path Continuity for Multi-Wheeled AGVs |
|
Kokot, Mirko | University of Zagreb |
Miklic, Damjan | Romb Technologies |
Petrovic, Tamara | Univ. of Zagreb |
Keywords: Logistics, Intelligent Transportation Systems, Industrial Robots
Abstract: Notwithstanding the growing presence of AGVs in the industry, there is a lack of research about multi-wheeled AGVs which offer higher maneuverability and space efficiency. In this paper, we present generalized path continuity conditions as a continuation of previous research done for vehicles with more constrained kinematic capabilities. We propose a novel approach for analytically defining various kinematic modes (motion modes), that AGVs with multiple steer and drive wheels can utilize. This approach enables deriving vehicle kinematic equations based on the vehicle configuration and its constraints, path shape, and corresponding motion mode. Finally, we derive general continuity conditions for paths that multi-wheeled AGVs can follow, and show through examples how they can be utilized in layout design methods.
|
|
12:10-12:30, Paper TuAT3.6 | |
>Container Drayage Problem: A Real-World Application in the Region of Marseille-Fos, France |
|
Abi-Nader, Diana | Mines Sait-Etienne, DMS-Logistics |
Absi, Nabil | Ecole Des Mines De Saint Etienne |
Azefack, Cyriac | DMS Logistics |
Korabi, Taki Eddine | DMS-Logistics |
Keywords: Logistics, Planning, Scheduling and Coordination, Manufacturing, Maintenance and Supply Chains
Abstract: This paper aims to present a generalized model of the container drayage problem within an application on industrial instances. The main contribution is the relaxation of some conservative assumptions regarding fleet composition and truck capacities. Moreover, the paper uses a more realistic network concerning business needs. Indeed, the experiments are made using real-world data with a heterogeneous fleet of trucks, a general network instead of a star one, and a significant number of various input requests.
|
|
TuAT4 Regular Session, Rhone 3A |
|
Sensor-Based Control |
Chair: Lamiraux, Florent | CNRS |
Co-Chair: de Farias, Cristiana | University of Birmingham |
|
10:30-10:50, Paper TuAT4.1 | |
>Configuration Estimation of Continuum Robots Using Piecewise Constant Curvature Generalized Epi-Polar Constraint Model |
|
Cheng, Hao | Tsinghua University |
Liu, Houde | Shenzhen Graduate School, Tsinghua University |
Wang, Xueqian | Center for Artificial Intelligence and Robotics, Graduate School |
Liang, Bin | Tsinghua University |
Keywords: Sensor-based Control, Computer Vision in Automation, Sensor Fusion
Abstract: In recent years, continuum robots have attracted more attention for they can work in more severe environments. However, at present, most of the research focuses on mechanical structure innovation, and there are few pieces of research on the control of this kind of robot. Since continuum robots are deformable, their shape is a general curve in space. Therefore, they are not fully defined by actuator positions, which are different from the traditional rigid robots. To achieve more accurate control, a method of sensing robot configuration in real-time is necessary. However, the existing visual-based approaches all adopt external global cameras, which is difficult to adapt to the demand of unknown unstructured environments. This paper presents a system capable of estimating the configuration of continuum robots under piecewise constant curvature (PCC) assumption from cameras mounted on each constant curvature segment. Specifically, we first proposed the PCC 2R model, which is equivalent to each cc-segment of PCC continuum robots by two joints rigid bodies, thereby reducing the problem complexity and improving the numerical stability of the estimation. Then, based on the PCC 2R model, we proposed the PCC generalized epi-polar constraint to completely constrain the four degrees of freedom of each cc-segment in planar, it can be solved through one corresponds, to estimate the configuration of continuum robots under PCC. Finally, the above approach is verified by experiment.
|
|
10:50-11:10, Paper TuAT4.2 | |
>Towards Robotic Metal Scrap Cutting: A Novel Workflow and Pipeline for Cutting Path Generation |
> Video
|
|
Akl, James | Worcester Polytechnic Institute |
Alladkani, Fadi | Worcester Polytechnic Institute |
Calli, Berk | Worcester Polytechnic Institute |
Keywords: Reactive and Sensor-Based Planning, Sustainability and Green Automation, Human-Centered Automation
Abstract: We propose a novel framework for robotic metal scrap cutting in unstructured scrap yards. In this framework the robots and workers collaborate: the worker marks the cutting locations on the scrap metal with spray paint and the robot then generates the cutting trajectories. This leverages worker expertise, while deferring the dull, dirty, dangerous aspects to the robot. For the robot, this requires a 3-D exploration and curve reconstruction stage for path generation. We use a non-uniform rational basis spline (NURBS) model and a topological skeletonization method for path generation, and implement and compare these methods via simulations. These simulations employ a realistic sensor noise model and highly-detailed 3-D scans of complex, real-life scrap pieces. Real-robot experiments with three different shapes are also provided.
|
|
11:10-11:30, Paper TuAT4.3 | |
>Sparse Multi-Sensor Monitoring System Design for Vehicle Application |
|
Khatiry Goharoodi, Saeideh | Ghent University |
Ooijevaar, Ted | Flanders Make |
Bey-Temsamani, Abdellatif | Flanders Make |
Crevecoeur, Guillaume | Ghent University |
Keywords: Sensor Fusion
Abstract: In today’s fast growing vehicle industry, the number of functionalities (comfort features, monitoring features, safety features, etc.) is steadily increasing. Each of these functionalities are developed independently from each other, hence the sensors are not shared among them. Although this design approach results into robust monitoring of these different functionalities, it requires a large number of sensors in different locations resulting in a complex hardware and software architecture (e.g. complex wires). This paper describes our approach where a multi sensor design method is used to optimally select locations of sensors that are shared by different functionalities. This results into a reduced number of sensors that monitor the same amount of functionalities. We demonstrate in this paper, an optimization algorithm based on Multi-Objective Integer Programming (MOIP) for optimal sensor placement for monitoring Motion Sickness Dose Value (MSDV) estimation and Speed Bump Detection (SBD) as part of a driver assistant system. The algorithm is further validated on a numerical data-set captured from an IPG CarMaker vehicle model. The methodology can be further extended to more functionalities with large number of applications in vehicle industry.
|
|
11:30-11:50, Paper TuAT4.4 | |
>Improving the Wheel Odometry Calibration of Self-Driving Vehicles Via Detection of Faulty Segments |
|
Fazekas, Mate | SZTAKI |
Gaspar, Peter | SZTAKI |
Nemeth, Balazs | MTA SZTAKI Institute for Computer Science and Control |
Keywords: Calibration and Identification, Autonomous Vehicle Navigation, Sensor Fusion
Abstract: The motion estimation of a self-driving car has to be as accurate as possible for proper control and safe driving. Therefore, the GNSS, IMU, or perception-based methods should be improved, e.g. with the integration of the wheel motion. This method is robust and cost-effective, but the calibration of the model parameters behind the wheel-based odometry is difficult. It is resulted from the nonlinear dynamics of the system and the requirement of parameter estimation with high precision, which is an open problem in the presence of noises yet. This paper proposes a novel architecture that simultaneously detects the faulty measurement segments, which results in biased parameter estimation. Furthermore, the measurements utilized for the calibration are also corrected to improve the efficiency of the parameter estimation. With the algorithm, the distortion effects of the noises can be eliminated, and accurate calibration of the nonlinear wheel odometry model can be obtained. The effectiveness of the detection and pose correction techniques and the operation of the calibration process are illustrated through vehicle test experiments.
|
|
11:50-12:10, Paper TuAT4.5 | |
>Dual Quaternion-Based Visual Servoing for Grasping Moving Objects |
|
de Farias, Cristiana | University of Birmingham |
Adjigble, Komlan Jean Maxime | University of Birmingham |
Tamadazte, Brahim | Univ. Bourgogne Franche-Comté, CNRS |
Stolkin, Rustam | University of Birmingham |
Marturi, Naresh | University of Birmingham |
Keywords: Sensor-based Control, Motion Control, Motion and Path Planning
Abstract: This paper presents a new dual quaternion-based formulation for pose-based visual servoing. Extending our previous work on local contact moment (LoCoMo) based grasp planning, we demonstrate grasping of arbitrarily moving objects in 3D space. Instead of using the conventional axis-angle parameterization, dual quaternions allow designing the visual servoing task in a more compact manner and provide robustness to manipulator singularities. Given an object point cloud, LoCoMo generates a ranked list of grasp and pre-grasp poses, which are used as desired poses for visual servoing. Whenever the object moves (tracked by visual marker tracking), the desired pose updates automatically. For this, capitalising on the dual quaternion spatial distance error, we propose a dynamic grasp re-ranking metric to select the best feasible grasp for the moving object. This allows the robot to readily track and grasp arbitrarily moving objects. In addition, we also explore the robot null-space with our controller to avoid joint limits so as to achieve smooth trajectories while following moving objects. We evaluate the performance of the proposed visual servoing by conducting simulation experiments of grasping various objects using a 7-axis robot fitted with a 2-finger gripper. Obtained results demonstrate the efficiency of our proposed visual servoing.
|
|
12:10-12:30, Paper TuAT4.6 | |
>Performing Manufacturing Tasks with a Mobile Manipulator: From Motion Planning to Sensor Based Motion Control |
|
Mirabel, Joseph | LAAS-CNRS |
Lamiraux, Florent | CNRS |
Ha, Thuc Long | Laas - Cnrs |
Nicolin, Alexis | Airbus Operations & LAAS-CNRS |
Stasse, Olivier | CNRS |
Boria, Sébastien | AIRBUS Operation SAS |
Keywords: Sensor-based Control, Manipulation Planning, Factory Automation
Abstract: We present a framework combining mobile ma- nipulation planning and sensor based motion control. The result of manipulation planning is a reference robot trajectory composed of segments. For each segment, a hierarchical task based controller is automatically built. Some of the segments correspond to motions where some parts of the robot and of the environment are known to be close to each other. The corresponding controller implements a task of relative pose between these elements. The task error is measured by vision sensors using AprilTags. The higher priority of this task with respect to the reference joint trajectory greatly improves the accuracy of the task affected by the poor accuracy of the robot base localization and other sources of error.
|
|
TuAT5 Regular Session, Rhone 3B |
|
Additive Manufacturing |
Chair: Huang, Qiang | University of Southern California |
Co-Chair: McGovern, Sean | Worcester Polytechnic Institute |
|
10:30-10:50, Paper TuAT5.1 | |
>An Efficient Control-Oriented Modeling Approach for Vibration-Prone Delta 3D Printers Using Receptance Coupling |
|
Edoimioya, Nosakhare | University of Michigan |
Okwudire, Chinedum | University of Michigan |
Keywords: Additive Manufacturing
Abstract: Delta 3D printers have the potential to significantly increase throughput in additive manufacturing because they enable faster and more precise motion when compared to traditional serial-axis 3D printers. Further improvements in motion speed and part quality can be realized through model-based feedforward vibration control, as demonstrated on several serial-axis 3D printers. However, delta 3D printers have not benefited from model-based controllers due to their coupled nonlinear dynamics which vary as a function of position. In this paper, we propose a framework to obtain linear models of delta 3D printers as functions of position. We decompose the dynamics into two sub-models: (1) an experimentally-identified sub-model containing decoupled vibration dynamics; and (2) an analytically-derived sub-model containing coupled rigid-body dynamics. These two sub-models are combined into one using receptance coupling. Employing the proposed approach, experiments are used to demonstrate reasonably accurate predictions of the position-dependent vibration dynamics of a delta 3D printer across its workspace using only two frequency response measurements at one location.
|
|
10:50-11:10, Paper TuAT5.2 | |
>A Drone-Assisted 3D Printing by Crane Structures in Construction Industry |
|
Parisi, Fabio | Polytechnic University of Bari |
Mangini, Agostino Marcello | Politecnico Di Bari |
Fanti, Maria Pia | Politecnico Di Bari |
Parisi, Nicola | Politecnico Di Bari |
Keywords: Automation in Construction, Additive Manufacturing, Cyber-physical Production Systems and Industry 4.0
Abstract: Additive manufacturing is a disruptive technology that is starting to be analyzed and studied also in construction industry. There are different approaches in its usage, and most of them concern the development of specific technologies, that are not always the optimal path to follow if environmental and energy related aspects are considered. In this paper a different paradigm to apply in the additive manufacturing exploitation in the construction industry is given. Starting from well studied and settled technologies like tower crane and drones, a simple, effective and already possible alternative for the usage of 3d printing at the real scale of construction is presented. A preliminary model is studied, its feasibility is investigated and the promising results of this first study are presented. Finally, the baseline and path for future investigation for the real development of this approach are enlightened.
|
|
11:10-11:30, Paper TuAT5.3 | |
>Echo State Network Learning for the Detection of Cyber Attacks in Additive Manufacturing |
|
Zhou, Houliang | The University of Texas at Arlington |
Liu, Chenang | Oklahoma State University |
Tian, Wenmeng | Mississippi State University |
Kan, Chen | University of Texas at Arlington |
Keywords: Additive Manufacturing, Sensor Fusion, Machine learning
Abstract: Owing to its layer-by-layer nature, additive manufacturing (AM) has been leveraged in various industries for the fabrication of parts with complex geometries. In the era of the Industrial Internet of Things (IIoT), AM processes are increasingly casted into both cyber and physical domains. As such, it poses AM under high risks of cyber attacks, leading to altered AM parts with potentially compromised mechanical properties and functionalities. It is imperative to develop new methodologies for the detection of cyber attacks for quality and reliability assurance of products in cyber-physical AM processes. Based on the echo state network, an online monitoring approach is developed in this study to extract features from side channels for the detection of cyber attacks. Two real-world case studies are conducted to evaluate the proposed approach on the fused filament fabrication (FFF) process. Experimental results have shown that the proposed approach is effective in identifying abnormities induced by different types of cyber attacks. The proposed approach has a strong potential to be extended to other AM processes with various sensing and monitoring devices.
|
|
11:30-11:50, Paper TuAT5.4 | |
>Scan-Driven Fully-Automated Pipeline for a Personalized, 3D Printed Low-Cost Prosthetic Hand |
|
Herbst, Yair | Technion - Israel Institute of Technology |
Polinsky, Shunit | Technion - Israel Institute of Technology |
Fischer, Anath | Technion Isreael Institute of Technology, Faculty of Mechanical |
Medan, Yoav | Technion - Israel Institute of Technology, Haifa3D |
Schneor, Ronit | Technion - Israel Institute of Technology |
Kahn, Joshua | Technion - Israel Institute of Technology |
Wolf, Alon | Technion |
Keywords: Prosthetics and Exoskeletons, Intelligent and Flexible Manufacturing, Additive Manufacturing
Abstract: The process of fitting a prosthetic hand that is comfortable, functional, easy to use, has an acceptable appearance and overall improves the amputees' quality of life is a complex, tedious and costly process. The very high price tag due to the time spent on manually fitting the device by a trained specialist makes these devices inaccessible to large portions of the population. We present a concept and preliminary results for a fully automated fitting and manufacturing pipeline for a personalized low-cost prosthetic hand. The hand is personalized in almost every aspect, from appearance to user interface, control and feedback. The pipeline only requires a 3D printer, RealSense cameras, a few basic mechanical components, and basic tools for the model assembly. The user scan-driven data and the user preferences initiate a fully-automated pipeline which culminates in a customized, easy-to-assemble PCB design and ready to print STL files, including the optimized orientation, support and layout, such that the final parts are only one click away. We believe that the proposed pipeline and design can highly impact the accessibility of prosthetic hands and could potentially be expanded to other medical applications.
|
|
11:50-12:10, Paper TuAT5.5 | |
>Efficient Feasibility Checking on Continuous Coverage Motion for Constrained Manipulation |
|
McGovern, Sean | Worcester Polytechnic Institute |
Xiao, Jing | Worcester Polytechnic Institute (WPI) |
Keywords: Additive Manufacturing, Factory Automation, Motion and Path Planning
Abstract: Many industrial robotic applications require a manipulator to move the end-effector in a constrained motion to cover a surface region, including painting, spray coating, abrasive blasting, polishing, shotcreting. etc. The manipulator has to satisfy both task constraints imposed on the end-effector (such as maintaining certain distance and angle with respect to the target surface while traversing it) and manipulator joint constraints. Given a robot manipulator and a target surface patch, an important question is whether there exists a feasible path for the manipulator to move continuously along the surface patch to cover it entirely while satisfying both manipulator and task constraints. This question is largely open as it has not been addressed systematically, even though there is substantial literature on path planning of constrained manipulation motion. In this paper, we introduce a general and efficient method to provide answers to this question.
|
|
TuAT6 Regular Session, St Clair 1 |
|
Probability and Statistical Methods |
Chair: Prigozin, Amit | Technion - Israel Institute of Technology |
Co-Chair: Onal, Cagdas | WPI |
|
10:30-10:50, Paper TuAT6.1 | |
>Model-Based Quasi-Static SLAM in Unstructured Dynamic Environments |
|
Deeb, Amy | Dalhousie University |
Pan, Ya-Jun | Dalhousie University |
Seto, Mae | Defence R&D Canada |
Keywords: Energy and Environment-Aware Automation, Computer Architecture for Robotic and Automation, Probability and Statistical Methods
Abstract: Inspired by the application of unmanned aerial vehicles to ship mapping and navigation in the Canadian Arctic, this paper proposes model-based quasi-static simultaneous localization and mapping (MBQS-SLAM). It extends the authors’ earlier work on model-based dynamic SLAM, to the specific challenges of marine Arctic environments where landmarks move at speeds that cannot be detected in consecutive observations - defining quasi-static motion. After developing model-based quasi-static factors and the MBQS-SLAM algorithm, encouraging performance is shown for both simulations and laboratory experiments in quasi-static environments.
|
|
10:50-11:10, Paper TuAT6.2 | |
>An Estimation Method of Transmission Line Parameters Based on Measurements of Injection Power and Voltage Phasor in Power Grid |
|
Zhou, Yadong | Xi'an Jiaotong University |
Hu, Bowen | Xi'an Jiaotong University |
Wu, Jiang | Xian Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Guan, Xiaohong | Xi'an Jiaotong University |
Chen, Wei | State Grid Shaanxi Electric Power Company |
Liu, Ting | Xi'an Jiaotong University |
Keywords: Power and Energy Systems automation, Modelling, Simulation and Validation of Cyber-physical Energy Systems, Smart Grids
Abstract: Estimation of transmission line parameters is an important problem for state estimation of power grid. In recent years, some studies have shown that the transmission line parameters can be estimated by some specific measurement data of the power grid. However, most of the studies mainly focus on single type of measurement data. In this paper, we propose an estimation method of transmission line parameters based on multi-measurement data of power grid, including measurements of injection power and voltage phasor. And considering the different reliability of the measurements, we make different restrictions on the multi-measurements. In our method, parameters and connection topology of transmission lines are estimated separately, and an ALS optimization model is established to solve the problem. Compared to the previous methods, the objective functions and constraints of our optimization model are improved, with considering different constraints based on phasor measurements, to get better estimation accuracy. The effectiveness and performance of the developed estimation method are demonstrated based on experiments using IEEE 30-bus and 118-bus systems.
|
|
11:10-11:30, Paper TuAT6.3 | |
>Mobile Robotic Radiation Surveying with Recursive Bayesian Estimation and Attenuation Modelling |
|
Anderson, Robert | The University of Texas at Austin |
Pryor, Mitchell | University of Texas |
Adrian, Abeyta, Adrian | University of Texas at Austin |
Landsberger, Sheldon | The University of Texas at Austin |
Keywords: Environment Monitoring and Management, Probability and Statistical Methods, Building Automation
Abstract: Both routine radiation surveys and incident response - as currently performed by human workers - are time consuming and involve significant, potentially unanticipated radiation dose (especially for accident response where levels prohibit human presence entirely). Previous efforts addressed routine surveys using robotic systems, but they must also localize and characterize discrete sources when anomalies occur. To be effective, characterization must account for real-world complications including multiple sources, environmental attenuation, spatial localization, and isotopic identification. Recursive Bayesian Estimation using grid-based estimators and particle filters has been previously investigated, but we eliminate previous open-space assumptions by including attenuation models and sensor position height in the sequential sampling strategy. The method also incorporates autonomous isotopic identification via gamma spectroscopy, which supports the attenuation modelling and improves the computational efficiency in multi-source cases. Additionally, survey points are autonomously optimized using Fisher information. A radioactive decay model is implemented to generalize the method by addressing short-lives nuclides. The developed hardware is evaluated in multiple scenarios where no operator intervention is required, effectively eliminating operator dose uptake from localization/characterization in real-world scenarios. Two different mobile robots are used, demonstrating the portability of the data collection and analysis software. Also, the generalized complexity is documented to support planning in future scenarios.
|
|
11:30-11:50, Paper TuAT6.4 | |
>Task Allocation in Multi-Robot Systems Based on the Suitability Level of the Individual Agents |
> Video
|
|
AlBuraiki, Omar | University of Ottawa |
Payeur, Pierre | University of Ottawa |
Keywords: Probability and Statistical Methods, Swarms, Agent-Based Systems
Abstract: This paper examines task allocation in multi-robot systems in the context where a suitability level of the specialized robots is considered. Based on the assumption that each individual agent possesses specialized functional capabilities and that the expected tasks impose specific requirements, a formulation of the agents’ specialization is defined to estimate individual agents’ task allocation probabilities. The original task allocation process involves a centralized matching scheme to associate each agent’s suitability level with corresponding detected tasks. Then, the task-agent matching scheme is expanded to coordinate the most specialized agent or group of agents while also considering availability factors. Early experimental results are presented and analyzed to demonstrate the effectiveness of the proposed framework.
|
|
11:50-12:10, Paper TuAT6.5 | |
>Fast Probabilistic 3-D Curvature Proprioception with a Magnetic Soft Sensor |
|
Mitchell, Mason | Worcester Polytechnic Institute |
Hurley, Forrest | Georgia Institute of Technology |
Onal, Cagdas | WPI |
Keywords: Probability and Statistical Methods, Compliant Assembly, Sensor-based Control
Abstract: This paper introduces a cost-effective and high speed approach for predicting a 2-DOF bend parameterization for soft bodies through a magnetic and constant curvature system. We propose a design for a probabilistic particle filter that can be paired with magnetic simulations to produce highly accurate and fast pose information for parameter-constrained magnets. We include the design, fabrication, modeling, and experimental results of a physical sensor with the ability to produce both bend directionality and bend magnitude results with a speed of ~60Hz. The proposed design consists of a magnet and tri-axis Hall effect sensor embedded in a soft silicone body. We demonstrate the effectiveness of this system through real-world interaction tests.
|
|
12:10-12:30, Paper TuAT6.6 | |
>Tactile-Based Gripper Localization on 1-D Deformable Objects |
> Video
|
|
Prigozin, Amit | Technion - Israel Institute of Technology |
Degani, Amir | Technion - Israel Institute of Technology |
Keywords: Assembly, Force and Tactile Sensing, Probability and Statistical Methods
Abstract: As part of automation processes, robotic manipulators are occasionally required to assemble deformable objects, e.g., installing an O-ring into a groove. However, deformable objects are characterized by high uncertainty due to shape and length change under external forces. These uncertainties make the assembly process complex and slow and may lead to errors between the actual and desired gripping location. In this paper, we present a localization technique to estimate the actual gripping point by using the grid localization algorithm based on tactile sensing. To reduce the dependency on complex and relatively slow vision sensors, the pose estimation process is based only on tactile feedback, by recognizing features, e.g., corners, along the deformable object. In simulations and experiments, the proposed algorithm converged to the correct gripping point after three detected features with an accuracy of less than 1 mm.
|
|
TuAT7 Regular Session, St Clair 2 |
|
Robot Networks and Medical Robots and Systems |
Chair: Schwaner, Kim Lindberg | University of Southern Denmark |
Co-Chair: Cavusoglu, M. Cenk | Case Western Reserve University |
|
10:30-10:50, Paper TuAT7.1 | |
>A Benchtop Robot and Automation Solution for Prefilled Syringes in Pharmaceutical Manufacturing |
> Video
|
|
DelSpina, Brandon | Clemson University |
Zhang, Yu | Clemson University |
Wang, Yue | Clemson University |
Keywords: Product Design, Development and Prototyping, Automation in Life Science: Biotechnology, Pharmaceutical and Health Care, Factory Automation
Abstract: Pharmaceutical manufacturing has strict requirements on the production environment, procedure, and product quality. Current manual operations face challenges in sterility, efficiency, and ergonomics. The introduction of industrial robotic arms can potentially meet biological manufacturing needs regarding ease to clean, high accuracy, and repeatability. This paper represents a benchtop robot and automation system aiming for the manufacturing of prefilled syringes. The system’s hardware components include an ISO certified robot manipulator (Yaskawa GP8), end-effector tools, cap feeding system, and telescoping transfer platform. A control system for the robot manipulator and Arduino is developed with the Robot Operating System Industrial (ROS-Industrial). An interpolation method is applied in low-level control to realize linear trajectories of the robot’s end-effector. Reliable filling, capping, and sealing are demonstrated to produce prefilled 60 mL syringe in 136 secs. Moreover, during a production run of 33 prefilled syringes, 32 are completed, with only one requiring human intervention due to cap misalignment of the syringe with the capping spout.
|
|
10:50-11:10, Paper TuAT7.2 | |
>Autonomous Needle Manipulation for Robotic Surgical Suturing Based on Skills Learned from Demonstration |
|
Schwaner, Kim Lindberg | University of Southern Denmark |
Dall'Alba, Diego | University of Verona |
Jensen, Pernille Tine | Aarhus University Hospital |
Fiorini, Paolo | University of Verona |
Savarimuthu, Thiusius Rajeeth | University of Southern Denmark |
Keywords: Medical Robots and Systems
Abstract: In the future, surgical robots will grant the option of executing surgical tasks autonomously, supervised by the surgeon. We propose a simple framework for learning surgical action primitives that can be used as building blocks for composing more elaborate surgical tasks. Our method is based on Learning from Demonstration (LfD) as this allows us to exploit existing expert knowledge from recordings of surgical procedures. We demonstrate that we can learn needle manipulation actions from human demonstration, constructing an action library which is used to autonomously execute part of a surgical suturing task. Actions are learned from single demonstrations and we use Dynamic Movement Primitives (DMPs) to encode low-level Cartesian space trajectories. Our method is experimentally validated in a non-clinical setting, where we show that learned actions can be generalized to previously unseen conditions. Experiments show a 81 % task success rate for moderate variations from the initial conditions of the demonstration with a mean needle insertion error of 3.8 mm.
|
|
11:10-11:30, Paper TuAT7.3 | |
>Visually Guided Needle Driving and Pull for Autonomous Suturing |
> Video
|
|
Ozguner, Orhan | Case Western Reserve University |
Shkurti, Tom | Case Western Reserve University |
Lu, Su | Case Western Reserve University |
Newman, Wyatt | Case Western Reserve University |
Cavusoglu, M. Cenk | Case Western Reserve University |
Keywords: Medical Robots and Systems
Abstract: This paper presents a visually-guided autonomous needle driving algorithm for autonomous robotic surgical suturing. Surgical needle tracking, needle path planning, and optimum needle grasp selection algorithms are employed. The procedure is performed in 5 major steps: needle grasp, needle hand-off, needle drive, needle regrasp, and needle pull. The performance of the procedure is experimentally evaluated using the physical da Vinci surgical robotic system and da Vinci Research Kit (dVRK). Initial results suggest that the dVRK can successfully perform needle driving with visual guidance.
|
|
11:30-11:50, Paper TuAT7.4 | |
>Distributed Optimal Control Framework Based on Coordinate Descent Optimization for Multi-Agent Robots |
> Video
|
|
Murtaza, Muhammad Ali | Georgia Institute of Technology |
Wingo, Bruce | Georgia Institute of Technology |
Kilanga, Dan | Georgia Institute of Technology |
Hutchinson, Seth | Georgia Institute of Technology |
Keywords: Robot Networks, Optimization and Optimal Control, Control Architectures and Programming
Abstract: In this paper, we present a distributed optimal control framework for a multi-agent robotics system based on coordinate descent optimization. Our framework exploits the underlying graph topology to compute the optimal control trajectory in a distributed manner. It only requires a modest amount of information exchange among the neighboring robot, and the computation depends on the underlying graph structure connecting the agents. Hence, if the underlying graph topology is sparse, e.g. a line graph, then the framework scales well with the problem's dimension, and any fast convergent algorithm can be used to ensure real-time computation. To show the efficacy of the framework, we apply it to a problem where a team of robots is tasked with establishing a communication link between source and destination while minimizing the overall system's mobility and communication energy. We analyzed its performance in simulation and on actual robots using an experimental robotic testbed, robotarium cite{wilson2020robotarium}, and compare it to the centralized solution of the same problem. The results show that the distributed framework converges and outperforms its centralized version as the problem's dimension increases. While the aforementioned energy-balancing problem serves to motivate the paper, the algorithm is defined and presented in a more general setting, and its potential extensions to other types of systems are pointed out.
|
|
11:50-12:10, Paper TuAT7.5 | |
>Learning Medical Suturing Primitives for Autonomous Suturing |
> Video
|
|
Amirshirzad, Negin | Ozyegin University |
Sunal, Begum | Ozyegin University |
Bebek, Ozkan | Ozyegin University |
Oztop, Erhan | Osaka University / Ozyegin University |
Keywords: Medical Robots and Systems, AI and Machine Learning in Healthcare, Learning and Adaptive Systems
Abstract: This paper focuses on a learning from demonstration approach for autonomous medical suturing. A conditional neural network is used to learn and generate suturing primitives trajectories which were conditioned on desired context points. Using our designed GUI a user could plan and select suturing insertion points. Given the insertion point our model generates joint trajectories on real time satisfying this condition. The generated trajectories combined with a kinematic feedback loop were used to drive an 11-DOF robotic system and shows satisfying abilities to learn and perform suturing primitives autonomously having only a few demonstrations of the movements.
|
|
12:10-12:30, Paper TuAT7.6 | |
>Robust Assignment Using Redundant Robots on Transport Networks with Uncertain Travel Time |
|
Prorok, Amanda | University of Cambridge |
|
|
TuAT8 Regular Session, Rhone 4 |
|
Manufacturing, Maintenance and Supply Chains |
Chair: Dotoli, Mariagrazia | Politecnico Di Bari |
Co-Chair: Inui, Masatomo | Ibaraki University |
|
10:30-10:50, Paper TuAT8.1 | |
>Kernel-Based Dynamic Ensemble Technique for Predictive Maintenance |
|
Lu, Hsuan-Wen | National Cheng Kung University |
Lee, Chia-Yen | National Taiwan University |
Keywords: Manufacturing, Maintenance and Supply Chains, AI-Based Methods
Abstract: Prognostic and health management (PHM) has been widely used in manufacturing system, particularly, for predictive maintenance (PdM). The purpose of PdM is to predict whether equipment or parts is in health. Typically, the statistical exponential models with the health index were often applied for the remaining useful life (RUL) estimation. However, due to the limitation of the health index construction, the prediction model may fail to respond in time. In fact, no single prediction model can predict perfectly for RUL. This study proposes a kernel-based dynamic ensemble technique (KDET) embedded with Inference Confidence Index (ICI) to build the weight adjustment of each model and model retraining mechanism. The ICI is built to measure the belief of the prediction by evaluating the similarity of multiple prediction models, and thus guide the concept drift to update the models immediately for the incoming streamline data. Two datasets are applied to validate the proposed KDET, and the results shows that the KDET can dynamically and effectively integrate multiple models for robust RUL prediction over time and thus improve the PdM system.
|
|
10:50-11:10, Paper TuAT8.2 | |
>Time-Gated Remaining Useful Life Prediction with Non-Periodical Inspection Data |
|
Deng, Yingjun | Tianjin University |
Zhang, Mingpeiyu | Microsoft |
Wang, Tian | Beihang University |
Wu, Huaming | Tianjin University |
Keywords: Failure Detection and Recovery, Manufacturing, Maintenance and Supply Chains, Deep Learning Methods
Abstract: This paper addresses the problem of remaining useful lifetime (RUL) prediction with non-periodical inspection data. To construct the RUL predictor, a two-stage solution is presented with the recently proposed time-gated long short-term memory network (TGLSTM) and a surrogate Wiener propagation model. The TGLSTM is used to process the non-periodical time-stamps and the Wiener propagation model aims to control the sequence-wise uncertainty. Also to process the high-frequency sensory data during each inspection, the temporal convolutional network (TCN) is introduced. The modeling rationale comes from the observed fact that prediction uncertainty reduces when time tends to the failure time, and the key insight is to introduce the latent Wiener process to model the joint probability density to observe the RUL prediction from the TGLSTM predictor and the actual RUL record simultaneously. Moreover, the TGLSTM predictor is interactively trained with the uncertainty propagation model. Our model is validated using the non-periodically under-sampled data from a turbofan engine degradation simulation use case.
|
|
11:10-11:30, Paper TuAT8.3 | |
>Fast Computation of Volumetric Thickness of 3D Objects Using GPU |
|
Inui, Masatomo | Ibaraki University |
Naganuma, Shinnosuke | , Ibaraki University |
Oki, Nao | The Department of Intelligent Systems Engineering, Ibaraki Unive |
Umezu, Nobuyuki | Ibaraki University |
Keywords: Product Design, Development and Prototyping, Manufacturing, Maintenance and Supply Chains, Factory Automation
Abstract: We proposed a novel definition for thickness, termed “volumetric thickness” to clearly visualize thickness distribution in solid objects. Because our proposed method analyzes the thickness at internal points of an object, the computational complexity for its implementation is significantly higher than that of the conventional ray and sphere methods, which analyze do so only at points on the object’s surface. This paper describes a fast analysis technique for volumetric thickness of 3D objects. To realize the computation in a realistic timeframe, we utilize a culling technique with hierarchical bounding boxes and a parallel processing function within the graphics processing unit (GPU). In the culling process, the necessary data is collected for appropriate utilization of the fast shared memory of GPU. We also present the numerical experiments conducted to verify the effectiveness of the proposed method.
|
|
11:30-11:50, Paper TuAT8.4 | |
>Industry 4.0 Asset-Based Risk Mitigation for Production Operation |
|
Winkler, Dietmar | Vienna University of Technology |
Novak, Petr | Czech Technical University in Prague - CIIRC |
Vyskocil, Jiri | Czech Technical University in Prague - CIIRC |
Meixner, Kristof | TU Wien |
Biffl, Stefan | TU Wien |
Keywords: Factory Automation, Intelligent and Flexible Manufacturing, Manufacturing, Maintenance and Supply Chains
Abstract: During engineering and operation of flexible robot-based production systems meeting the Industry 4.0 (I40) paradigm, engineers require guidance to analyse and resolve issues that may disturb the production process. Challenging issues stem from causes in several heterogeneous engineering disciplines. Unfortunately, current risk mitigation guidelines frequently focus on isolated components and not on risks of the entire system. This fragmented guidance is hard to apply for engineers who are not aware of existing dependencies between components of different types. Therefore, risks in the engineering and operation of I40 components are hard to identify and mitigate. In this paper, we propose the Industry 4.0 Asset based Risk Mitigation (I4ARM) approach, providing knowledge for efficient root cause analysis to non-experts based on (a) a minimal model for knowledge representation as an I40 asset network with cause-effect annotations and (b) the I4ARM method for model building and risk mitigation with structured guidance. We build on the I40 asset network concept, cause effect analysis, and decision trees to enable efficient and effective risk mitigation with structured guidance. I4ARM facilitates for engineers (a) defining an Industry 4.0 asset network and relationships, (b) identifying risks, and (c) supporting risk mitigation. We conceptually evaluate I4ARM for a real-world I40 use case. The results showed that the I40 Asset Network with cause-effect relationships and decision trees is usable and useful both for experienced and novice engineers to efficiently and systematically mitigate risks in I40 environments.
|
|
11:50-12:10, Paper TuAT8.5 | |
>An MPC-Based Approach for the Feedback Control of the Cold Sheet Metal Forming Process |
|
Bozza, Augusto | Polytechnic of Bari |
Cavone, Graziana | Polytechnic of Bari |
Carli, Raffaele | Politecnico Di Bari |
Mazzoccoli, Luigi | Polytechnic of Bari |
Dotoli, Mariagrazia | Politecnico Di Bari |
Keywords: Control Architectures and Programming, Manufacturing, Maintenance and Supply Chains, Process Control
Abstract: In the automotive sector the cold forming of metal sheets is one of the main production activities. However, it is also one of the main source of production wastes. The generally adopted strategy to reduce the number of abnormal stamped parts is the feedback control of the stamping press (i.e., the machine control), while the feedback control of the stamping process is rarely considered. The process control, differently from the press control, can allow the monitoring of the state of the stamped part during the formation phase and the provision of corrective actions in case of abnormal behaviors of the metal sheet, thus ensuring a more precise control of the process. In this context, this paper presents a novel methodology for the cold metal forming process control based on Model Predictive Control (MPC). Firstly, a dynamical model of the system is defined that describes the draw-in of n critical points of the metal sheet as a function of the Blank Holder Force (BHF) and the punch stroke. Then, two different MPC-based real-time controllers are built for two different types of press configuration: the monolithic and the differential one. In the first case, a mono-MPC control system evaluates the draw-in of n critical points and computes a single couple of control signals (i.e., the BHF and the punch stroke). In the second case, a multi-MPC control system computes n different couples of control signals, i.e., one for each monitored draw-in. Finally, a case study is presented with the aim to test both the architectures, considering several simulation scenarios (with or without external disturbances on the plant), in order to make a control system architectures comparison in terms of tracking errors and workpiece quality.
|
|
TuAT9 Special Session, St Clair 3A |
|
New Methods for Intelligent Manufacturing Operations Management |
Chair: Jiang, Zhibin | Shanghai Jiao Tong University |
Co-Chair: Li, Jingshan | University of Wisconsin - Madison |
Organizer: Jiang, Zhibin | Shanghai Jiao Tong University |
Organizer: Li, Jingshan | University of Wisconsin - Madison |
Organizer: Zhou, Liping | Shanghai Jiao Tong University |
|
10:30-10:50, Paper TuAT9.1 | |
>Cell Formation for Cellular Reconfigurable Manufacturing Systems with Alternative Routing (I) |
|
Guo, Siqi | Shanghai Jiao Tong University |
Cui, Feng | Shanghai Jiaotong University |
Geng, Na | Shanghai Jiao Tong University |
Jiang, Zhibin | Shanghai Jiao Tong University |
Keywords: Intelligent and Flexible Manufacturing, Optimization and Optimal Control, Planning, Scheduling and Coordination
Abstract: Cellular reconfigurable manufacturing systems (CRMS) are designed to deal with dynamic demand. This paper tries to optimize cell formation decision in CRMSs with alternative routing. A mixed integer nonlinear programming model is developed to simultaneously optimize cell formation, equipment configuration, and product mix plan. The objective is to maximize profit. The linearization methods are proposed and then the model is solved by Gurobi. Numerical experiment and sensitive analysis are conducted to show the efficiency and effectiveness of the proposed approach.
|
|
10:50-11:10, Paper TuAT9.2 | |
>The Just-In-Time Job-Shop Rescheduling with Rush Orders by Using a Meta-Heuristic Algorithm (I) |
|
Ren, Xiyue | Shanghai Jiao Tong University |
Wang, Xiuxian | Shanghai Jiao Tong University |
Geng, Na | Shanghai Jiao Tong University |
Jiang, Zhibin | Shanghai Jiao Tong University |
Keywords: Intelligent and Flexible Manufacturing, Planning, Scheduling and Coordination, Optimization and Optimal Control
Abstract: In the real manufacturing system, rescheduling is inevitable because of rush orders. To improve the rush order inserting problem of rescheduling, this paper focuses on the just-in-time job-shop rescheduling problem(JIT JSRP), in which each job has its own due date and any earliness /tardiness leads to the penalty. A mixed integer programming model is established to minimize the weighted penalty cost of earliness/tardiness and the starting time deviations. The paper develops a hybrid tabu-variable neighborhood search (HTVNS) algorithm to solve the problem. Moreover, the adaptive shake operator selection algorithm and two improved N5 neighborhood structures are introduced to improve the efficiency of the algorithm. In numerical experiments, the improved algorithm is testified using 36 cases with different scales and arrival times of rush orders, and compared with classical meta-heuristic algorithms. The computational results show the effectiveness of the proposed improved algorithm.
|
|
11:10-11:30, Paper TuAT9.3 | |
>Integrated Production and Transportation Scheduling in a Make-To-Order Manufacturing Network with Heterogeneous Vehicles (I) |
|
Liu, Kefei | Shanghai Jiaotong University |
Zhou, Liping | Shanghai Jiao Tong University |
Jiang, Zhibin | Shanghai Jiao Tong University |
Keywords: Cyber-physical Production Systems and Industry 4.0, Hybrid Strategy of Intelligent Manufacturing, Intelligent Transportation Systems
Abstract: In the situation of intelligent manufacturing, the demand for individual customization increases. Many enterprises adopt make-to-order production. With the increasing trend of globalization, numerous enterprises are multinational, which have multi-location plants and distributors. The customized orders are split into multiple different jobs and then produced by different plants. Under the requirements of low-cost production and rapid response to demand, the coordination between production and transportation is becoming increasingly important. Considering the features of order splitting production, heterogeneous vehicle transportation, and collaborative delivery in make-to-order production, this paper studies the integrated scheduling of production and transportation of multiple splittable orders in networked plants and distributors with different geographical locations. A mixed-integer linear programming model is developed for deciding the production plant, start production date, the transport vehicles, and departure date of each job from its assigned plant with the objective of minimizing total cost including manufacturing cost, inventory cost, transportation cost, and backorder cost. The commercial solver Gurobi is used to solve this model. Numerical experiments based on the real data of a customized furniture enterprise in China are performed to show the efficiency of the proposed approach and demonstrate that our integrated scheduling approach is more effective than sequential scheduling.
|
|
11:30-11:50, Paper TuAT9.4 | |
>Reconfigurable Timed Extended Reachability Graphs for Scheduling Problems in Uncertain Environments |
|
Hayane, Oussama | GREAH - Université Le Havre Normandie |
Lefebvre, Dimitri | University LE HAVRE |
Keywords: Petri Nets for Automation Control, Intelligent and Flexible Manufacturing, Manufacturing, Maintenance and Supply Chains
Abstract: This paper aims to develop a new method to determine a robust scheduling control for systems evolving in uncertain environments. Time Petri Nets with controllable and uncontrollable transitions are used to model the system. The controllable transitions represent the operations and the uncontrollable transitions represent unexpected events that correspond either to interruption of operations or to unavailability of resources. The developed method computes reconfigurable control sequences based on the determination of series of timed extended reachability graphs (R-TERG). Once an unexpected event is detected, a reconfiguration point is created and the R-TERG is updated. Successive applications of the Dijkstra algorithm allow to reconfigure the control sequence in order to preserve optimality with respect to the faults that affect the system.
|
|
11:50-12:10, Paper TuAT9.5 | |
>Predicting the Distribution of Product Completion Time in Multi-Product Manufacturing Systems |
|
Huang, Jing | University of Virginia |
Chang, Qing | University of Virginia |
Arinez, Jorge | General Motors Research & Development Center |
Keywords: Intelligent and Flexible Manufacturing, Manufacturing, Maintenance and Supply Chains, AI-Based Methods
Abstract: Product completion time is a random variable, which results from the random disturbances in the production systems that delay the processing of products unexpectedly. Existing methods for product completion time prediction mostly predict its mean value. However, mean value only accounts for the first moment of a probability distribution, and is not sufficient for depicting the full spread of the product completion time. In this paper, we propose a novel method for predicting the probability distribution of production completion time by combining system model and deep learning. The original data collected from the plant floor are boosted through a model-based oversampling process, which helps retrieve more distribution information from machine operation history. Supported by theoretical and empirical evidence, the location family of the Tweedie distribution is discovered to well fit the product competition time. A hybrid framework is established to predict distribution parameters given system state as input, so as to predict the completion time distributions in a real-time fashion. The location parameter is analytically evaluated with system model. Other parameters are predicted or determined with data-driven methods, including a long-short term memory network and classic Tweedie prediction techniques.
|
|
12:10-12:30, Paper TuAT9.6 | |
>Data-Enabled Identification and Prediction of Permanent Production Loss for Production System with Variable Cycle Time |
|
Li, Chen | University of Virginia |
Huang, Jing | University of Virginia |
Chang, Qing | University of Virginia |
Keywords: Intelligent and Flexible Manufacturing, Discrete Event Dynamic Automation Systems, Manufacturing, Maintenance and Supply Chains
Abstract: Real time production performance evaluation plays a vital role in diagnosing manufacturing system health status and achieving productivity improvements. However, most existing studies on system performance evaluation are based on steady state analysis and focused on the production system with fixed machine cycle time. The real-time performance evaluation for a manufacturing system with variable machine cycle time, although typical for a large number of realistic scenarios, has been mostly ignored. The development of smart manufacturing and increasingly available sensor data have provided unprecedented opportunities to carry out thorough analysis on the real-time performance of such complex systems. In this paper, we developed a data-enabled methodology to efficiently identify and predict the real-time permanent production loss for a production system with variable machine cycle times. The concept and evaluation method of the opportunity window are introduced to facilitate the permanent production loss estimation. Numerical case studies are presented to demonstrate the effectiveness of the proposed methods for opportunity window evaluation and production loss identification.
|
|
TuAT10 Regular Session, St Clair 3B |
|
Deep Learning in Robotics and Automation 1 |
Chair: Pb, Sujit | IISER Bhopal |
Co-Chair: Chong, Nak Young | Japan Advanced Inst. of Sci. and Tech |
|
10:30-10:50, Paper TuAT10.1 | |
>Automation of Winding Thick Wires for the Electrical Industry with Performance-Indicating Discriminator |
|
Niho, Takayoshi | The University of Tokyo |
Osa, Takayuki | Kyushu Institute of Technology |
Suzuki, Shota | Furukawa Electric Co |
Moriki, Kazuya | Furukawa Electric Co |
Sugita, Naohiko | The University of Tokyo |
Nakao, Masayuki | The University of Tokyo |
Keywords: Factory Automation, Deep Learning in Robotics and Automation, Failure Detection and Recovery
Abstract: In the electronics industry, winding thick wires is one of the most frequent tasks in factories. Winding thick wires is a heavy and dangerous task, therefore, its automation is needed for improving the productivity and safety. However, automatic winding of thick wires is often challenging due to the anisotropic rigidity of wires: Taking appropriate winding actions for each state is important. In this study, using a scale model, an automatic winding framework based on an imitation learning approach was developed. The scale model comprises a robot arm and a line profile sensor that observes the wire wound on a rotating drum. The framework consists of two components. The first is a function for generating winding actions. This function is implemented using neural networks and generates winding actions for the input (winding state). We herein revealed that the winding state should be represented by feature values rather than raw data from the profile sensor. The second is a function for indicating the quality of the winding action. This is similar to the discriminator introduced in previous studies involving generative adversarial networks, but with a difference in the training dataset: When the network generates an action comparable to that of human experts, the data is not added to the training dataset of the discriminator. This keeps the discriminator having the ability to indicate the quality of generated actions. In the experiments, using the proposed discriminator, the human intervention was requested when the action generated by a network was unskilled; otherwise, the winding action is performed automatically.
|
|
10:50-11:10, Paper TuAT10.2 | |
>Automated Masonry Crack Detection with Faster R-CNN |
|
Marin, Borja | Heriot-Watt University |
Erden, Mustafa Suphi | Heriot-Watt University |
Brown, Keith E. | Heriot-Watt University |
Keywords: Deep Learning in Robotics and Automation, Computer Vision in Automation, Automation in Construction
Abstract: Inspection of masonry buildings, typically railway bridges, for crack detection is currently performed by humans under tedious and sometimes dangerous working conditions. Over the past years, computer vision based techniques have been developed to automate structure visual inspections. These techniques could be integrated with (semi) autonomous drone surveillance to collect images of assets for full automation of simultaneous inspection and crack detection in railway bridges. In this study we have adopted the architecture of Faster R-CNN object detectors to provide crack detection in images. In this architecture, we have tested three networks (Mobilenetv2, Resnet50 and ZF512) to be utilised as feature extractors in a limited resource system for crack detection. We propose a new way of performing detection that we call Progressive Detection to increase the robustness of detection, considering otherwise only partially detected cracks. Since one of the main goals of visual inspection is checking the health of every single defect, we have revisited binary classification of images with and without cracks from a detection point of view, with the objective of minimising crack missing rates. Results show that Mobilenetv2 performs both successfully and fast enough to be applied in a drone application as a feature extractor network, achieving a close level of performance to the more sophisticated network Resnet50 with half its inference time. Regarding classification, Mobilenetv2 achieves its best performance in the early stages of its training process, showcasing 93% accuracy and a crack miss rate ranging from 1% to 15%. These results are comparable to Resnet's and better than ZF512's.
|
|
11:10-11:30, Paper TuAT10.3 | |
>Planar Pushing of Unknown Objects Using a Large-Scale Simulation Dataset and Few-Shot Learning |
|
Gao, Ziyan | Japan Advanced Institute of Science and Technology |
Elibol, Armagan | Japan Advanced Institute of Science and Technology |
Chong, Nak Young | Japan Advanced Inst. of Sci. and Tech |
Keywords: Deep Learning in Robotics and Automation, Model Learning for Control, Learning and Adaptive Systems
Abstract: Contact-rich object manipulation skills challenge the recent success of learning-based methods. It is even more difficult to predict the state of motion of novel objects due to the unknown physical properties and generalization issues of the learning-based model. In this work, we aim to predict the dynamics of novel objects in order to facilitate model-based control methods in planar pushing. We deal with this problem in two aspects. First, we present a large-scale planar pushing simulation dataset called SimPush. It is characterized by a large number of pushes and a variety of object physical properties, providing a wide avenue for exploring the object responses to the pusher action. Secondly, we propose a novel task-aware representation for pushes. This method keeps the spatial relation between the object and pusher and emphasizes the local contact features. Finally, we propose an encoder-decoder structured model possessing a cascaded residual attention mechanism to integrate prior knowledge to infer novel object motions. We experimentally show that the proposed model purely trained by SimPush attains good performance and robust prediction of novel object motions.
|
|
11:30-11:50, Paper TuAT10.4 | |
>Enhancement of Visual Place Recognition for Robot Localization Subject to Pedestrian Occlusion |
|
Li, Yu-Jen | National Taipei University of Technology |
Chen, Hsin-Hung | National Taipei University of Technology |
Li, Chih-Hung G. | National Taipei University of Technology |
Keywords: Deep Learning in Robotics and Automation, Computer Vision in Automation, Computer Vision for Transportation
Abstract: Visual detection methods have been vastly applied for place recognition and localization of autonomous mobile robots (AMRs). As most of the AMRs are deployed in human-centric environments, encountering dynamic changes such as passing-by pedestrians is inevitable. Pedestrian occlusion may greatly reduce the performance of autonomous localization; however, so far there has not been a method developed to address this problem. In this article, we proposed an online image inpainting process to reduce the adverse influence of the pedestrian on visual localization. Specifically, the proposed scheme integrates a deep neural network (DNN)-based pedestrian detector to find and remove the pedestrian pixels on the image. We then repair the image using another DNN that has exhibited excellent image inpainting performance. To verify the proposed scheme, series of field tests were carried out on indoor corridors of an office building. The results showed that the pedestrian appearing in the surveillance camera of the AMR may reduce the accuracy of the topological localization system by 10%. The proposed scheme successfully corrected more than 50% of the predictions that were erroneously made previously.
|
|
11:50-12:10, Paper TuAT10.5 | |
>OFFSEG: A Semantic Segmentation Framework for Off-Road Driving |
|
Viswanath, Kasi | Indian Institute of Science Education and Research Bhopal |
Singh, Kartikeya | IISER Bhopal |
Jiang, Peng | Texas A&M University |
Pb, Sujit | IISER Bhopal |
Saripalli, Srikanth | Texas A&M |
Keywords: Deep Learning in Robotics and Automation, Computer Vision for Transportation
Abstract: Off-road image semantic segmentation is challenging due to the presence of uneven terrain, unstructured class boundaries, irregular features and strong textures. These aspects affect the vehicle perception. Current off-road datasets exhibit difficulties like class imbalance and understanding of varying environmental topography. To overcome these issues, we propose a framework for off-road semantic segmentation (OFFSEG) that involves (i) a pooled class semantic segmentation with four classes (sky, traversable region, non-traversable region and obstacle) using state-of-the-art deep learning architectures (ii) a color segmentation methodology to segment out specific sub-classes (grass, puddle, dirt, gravel, etc.) from the traversable region for better scene understanding. The evaluation of the framework is carried out on two off-road driving datasets, namely, RELLIS-3D and RUGD. We have also tested the proposed framework on IISERB campus data. The results show that OFFSEG achieves good performance and also provides detailed information on the traversable region.
|
|
TuAT11 Regular Session, St Clair 4 |
|
AI-Based Methods 1 |
Chair: Dalmas, Benjamin | Ecole Des Mines De Saint-Etienne |
Co-Chair: Li, Xiaopeng | Xian Jiaotong University |
|
10:30-10:50, Paper TuAT11.1 | |
>Performance Analysis for the Forecast of Electric Energy Consumption in a Cooperative House Based on LSTM |
|
Solis, Jorge | Karlstad University / Waseda University |
Sjöberg, David | Karlstad University |
Nilsson, Magnus | Glava Energy Center |
Ericson, Johan | Glava Energy Center |
Keywords: Smart Grids, AI-Based Methods, Energy and Environment-aware Automation
Abstract: Our research aims to develop an adaptive control system for photovoltaic systems with energy storage that adapts after changing different kinds of conditions. In particular, for efficient controlling of battery storage, the precise prediction of electricity consumption is required. Due to the complexity of the proposed research; in a previous research, the authors proposed a simplified LSTM model based on 48 hours data in order to predict two hours ahead the electricity consumption. In this paper, we rather focused to investigate a suitable sampling time resolution to produce LSTM-model based on a specific house in order to predict the energy consumption in another house having similar characteristics for performance analysis purposes. Based on the experimental results, it is suitable to sample the input data for training purposes every 15 minutes and to select a best suitable LSTM-model.
|
|
10:50-11:10, Paper TuAT11.2 | |
>Uncertainty Set Prediction of Aggregated Wind Power Generation Based on Bayesian LSTM and Spatio-Temporal Analysis |
|
Li, Xiaopeng | Xian Jiaotong University |
Wu, Jiang | Xian Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Liu, Kun | Xi'an Jiaotong University |
Yu, Jun | Xi'an Jiaotong Univerwsity |
Guan, Xiaohong | Xi'an Jiaotong University |
Keywords: Big data Analytics for Large-scale Energy Systems, Renewable Energy Sources, AI-Based Methods
Abstract: Aggregated stochastic characteristics of geographically distributed wind generation will provide valuable information for secured and economical system operation in electricity markets. This paper focuses on the uncertainty set prediction of the aggregated generation of geographically distributed wind farms. A Spatio-temporal model is proposed to learn the dynamic features from partial observation in near-surface wind fields of neighboring wind farms. We use Bayesian LSTM, a probabilistic prediction model, to obtain the uncertainty set of the generation in individual wind farms. Then, spatial correlation between different wind farms is presented to correct the output results. Numerical testing results based on the actual data with 6 wind farms in northwest China show that the uncertainty set of aggregated wind generation of distributed wind farms is less volatile than that of a single wind farm.
|
|
11:10-11:30, Paper TuAT11.3 | |
>An Unsupervised-Learning Based Method for Detecting Groups of Malicious Web Crawlers in Internet |
|
Yue, Tianyi | Xi'an Jiaotong University |
Zhou, Yadong | Xi'an Jiaotong University |
Hu, Bowen | Xi'an Jiaotong University |
Xu, Zhanbo | Xi'an Jiaotong University |
Guan, Xiaohong | Xi'an Jiaotong University |
Zhou, Hao | River Security Information Technology (Shanghai) Co., Ltd |
Liu, Ting | Xi'an Jiaotong University |
Keywords: AI-Based Methods, Learning and Adaptive Systems, Machine learning
Abstract: Malicious web crawler has been a serious threat to the security and performance of web servers in Internet. Generally, malicious web crawler systematically obtains massive web pages without approval, and may involve the theft of data assets. In this paper, we propose an unsupervised learning based method for detecting malicious web crawler. The method can be divided into three phases. Firstly, the method generates a representative vector for each client by combining the information of its visiting statistic behaviors and page request stream. Secondly, a new subspace clustering algorithm is developed to cluster the clients into groups. Finally, four metrics are designed to detect the groups of malicious web crawlers. The proposed method is validated based on a real data set consisting of 580 thousand accessing requests. Experimental results show that the proposed method can accurately detect malicious web crawlers with a high TPR (true positive rate) of 91.0% and a low FPR (false positive rate) of 1.3%.
|
|
11:30-11:50, Paper TuAT11.4 | |
>Iterative Backpropagation Disturbance Observer with Forward Dynamics Model |
> Video
|
|
Murooka, Takayuki | The University of Tokyo |
Hamaya, Masashi | OMRON SINIC X Corporation |
von Drigalski, Felix Wolf Hans Erich | OMRON SINIC X Corporation |
Tanaka, Kazutoshi | OMRON SINIC X Corporation |
Ijiri, Yoshihisa | OMRON Corp |
Keywords: AI-Based Methods, Model Learning for Control, Robust/Adaptive Control
Abstract: Disturbance Observer (DOB) has been widely used for robotic applications to eliminate various kinds of disturbances. Recently, learning-based DOB has attracted significant attention as it can deal with complex robotic systems. In this study, we propose the Iterative Backpropagation Disturbance Observer (IB-DOB) method. IB-DOB learns the forward model with a neural network and calculates disturbances via iterative backpropagations, which behaves like the inverse model. Our method can not only improve estimation performances owing to the iterative calculation but also be applied to both model-free and -based learning control. We conducted experiments for two manipulation tasks: the cart pole with Deep Deterministic Policy Gradient (DDPG) and the pushing object task with Deep Model Predictive Control (DeepMPC). Our method demonstrated better task performances than the baselines without DOB and with DOB using a learned inverse model even though disturbances of external forces and model errors were provided.
|
|
11:50-12:10, Paper TuAT11.5 | |
>AutoSS: A Deep Learning-Based Soft Sensor for Handling Time-Series Input Data |
|
Bargellesi, Nicolň | University of Padova |
Beghi, Alessandro | Universita` Di Padova |
Rampazzo, Mirco | Universitŕ Di Padova |
Susto, Gian Antonio | University of Padova |
Keywords: AI-Based Methods, Deep Learning Methods, Big Data in Robotics and Automation
Abstract: Soft Sensors are data-driven technologies that allow to have estimations of quantities that are impossible or costly to be measured. Unfortunately, the design of effective soft sensors is heavily impacted by time-consuming feature engineering steps that may lead to sub-optimal information, especially when dealing with time-series input data. While domain knowledge may come into help when handling feature extraction in soft sensing applications, the feature extraction typically limit the adoption such technologies: in this work, we propose AutoSS, a Deep-Learning based approach that allow to overcome such issue. By exploiting autoencoders, dilated convolutions and an ad-hoc defined architecture, AutoSS allow to develop effective soft sensing modules even with time-series input data. The effectiveness of AutoSS is demonstrated on a real world case study related to Internet of Things equipment.
|
|
12:10-12:30, Paper TuAT11.6 | |
>A Hierarchical Model to Enable Plan Reuse and Repair in Assembly Domains |
|
Parashar, Priyam | University of California - San Diego |
Naik, Aayush | University of California - San Diego |
Hu, Jiaming | University of California San Diego |
Christensen, Henrik Iskov | UC San Diego |
Keywords: AI-Based Methods, Failure Detection and Recovery, Intelligent and Flexible Manufacturing
Abstract: The manufacturing world is moving towards a setup with high mix and medium volume. This requires new production paradigms. The World Robotics Challenge (2018 & 2020) was designed to challenge teams to design systems that are easy to adapt to new tasks and to ensure robust operation in a semi-structured environment. We present a layered strategy to transform missions into tasks and actions and provide a set of strategies to address simple and complex failures. We propose a model for characterizing failures using this model and discuss repairs. Simple failures are by far the most common in our WRC system and we also present how we repaired them.
|
|
TuBT1 Special Session, Auditorium |
|
Machine Learning and Data Analytics for Failure Analysis in Manufacturing
Industry |
Chair: Boucher, Xavier | Ecole Nationale Supérieure Des Mines De Saint-Etienne |
Co-Chair: Sharma, Kanuj | University of Stuttgart |
Organizer: Boucher, Xavier | Ecole Nationale Supérieure Des Mines De Saint-Etienne |
|
15:30-15:50, Paper TuBT1.1 | |
>Thermal Source Separation for 3D Defect Localization Using Independent Component Analysis (ICA) from Time-Resolved Temperature Response (TRTR) (I) |
|
Kögel, Michael | Fraunhofer Institute for Microstructure of Materials and Systems |
Brand, Sebastian | Fraunhofer Institute for Microstructure of Materials and Systems |
Große, Christian | Fraunhofer Institute for Microstructure of Materials and Systems |
Jacobs, Kristof JP | Imec, 3000 Leuven, Belgium & Fac. Engineering, Dept. Material Sc |
De Wolf, Ingrid | Imec, 3000 Leuven, Belgium & Fac. Engineering, Dept. Material Sc |
Altmann, Frank | Fraunhofer Institute for Microstructure of Materials and Systems |
Keywords: Machine learning, Probability and Statistical Methods
Abstract: The current paper presents the application of independent component analysis (ICA) for blind source separation of lock-in thermography (LIT) measurement data aiming the localization and identification of thermally active spots that are related to multiple subsurface defects in packaged devices. Recently 3D defect localization by LIT was extended acquiring the time-resolved temperature response (TRTR) [1]. The TRTR represents the temporal sequence of the radiated thermal signal which is induced by periodic electrical excitation of the sample under test and is recorded for each pixel in the thermogram. The signal emitted by a thermal source propagates through a sample by unidirectional conduction and radiation until it reaches the surface from where it radiates off towards the camera. Conduction simultaneously occurs in all three spatial directions resulting in latera spreading and thus a broadened hot-spot at the surface. The time a thermal signal requires to reach the surface via conduction depends on the thermal properties of the crossed materials but also the depth inside the sample resulting in specifically altered shapes of the TRTRs. When the surficial hot-spots of thermal sources originated at different depths superimpose, the resulting interference of the thermal trace leads to falsified phase values and thus an erroneous depth localization. The identification of the independent spatial components and their time dependent coefficients in the TRTR sequence, using ICA allows the extraction of the spatiotemporal evolution of the thermal spreads of each thermal source by a nonlinear mixing model. This enables an accurate separation of phase shifts of the thermal signals of the individual sources.
|
|
15:50-16:10, Paper TuBT1.2 | |
>On Welding Defect Detection and Causalities between Welding Signals (I) |
|
Melakhsou, Abdallah Amine | Ecole Des Mines De Saint-Etienne |
Batton Hubert, Mireille | Institut Fayol Ecole Nationale Supérieure Des Mines De Saint Et |
Keywords: Probability and Statistical Methods, Failure Detection and Recovery, Machine learning
Abstract: In the manufacturing of hot water tanks, welding quality evaluation and fault detection is a critical operation that is still frequently conducted visually or with the help of none destructive tests, which can generate a high consumption of time and resources. To overcome this problem, many methods have been proposed for defect detection based on the classification of the welding signals. However, most of the proposed methods do not consider the problems of defect localization and the generalization from small sample size. Moreover, studies on the interactions between welding signals seem to be absent in the literature despite its importance in the formulation of the defect detection problem. In this paper, we aim to address these gaps by presenting a study on the causalities between welding signals in the circular welding of hot water tanks. Based on the findings from the causality study, we propose a method that detects and localize welding defect with high accuracy and handles the problem of small sample size. Furthermore, we present a study on the defects root cause and show the possibility of early defect prediction.
|
|
16:10-16:30, Paper TuBT1.3 | |
>Anomaly Detection Based on Selection and Weighting in Latent Space (I) |
|
Liao, Yiwen | University of Stuttgart |
Bartler, Alexander | University of Stuttgart |
Yang, Bin | University of Stuttgart |
Keywords: Failure Detection and Recovery, AI-Based Methods, Machine learning
Abstract: With the high requirements of automation in the era of Industry 4.0, anomaly detection plays an increasingly important role in high safety and reliability in the production and manufacturing industry. Recently, autoencoders have been widely used as a backend algorithm for anomaly detection. Different techniques have been developed to improve the anomaly detection performance of autoencoders. Nonetheless, little attention has been paid to the latent representations learned by autoencoders. In this paper, we propose a novel selection-and-weighting-based anomaly detection framework called SWAD. In particular, the learned latent representations are individually selected and weighted. Experiments on both benchmark and real-world datasets have shown the effectiveness and superiority of SWAD. On the benchmark datasets, the SWAD framework has reached comparable or even better performance than the state-of-the-art approaches.
|
|
16:30-16:50, Paper TuBT1.4 | |
>Defect Detection with Small Samples Based on Meta Learning (I) |
|
Yu, Ge | Peking University |
Zhang, Xi | College of Engineering, Peking University |
Keywords: Failure Detection and Recovery, Diagnosis and Prognostics, Cognitive Automation
Abstract: Defects widely exist in industrial data, and the identification affects the system diagnosis performance and state stability. Existing intelligent models are limited due to the defect data sparsity. In this study, we propose a meta learning-based defect detection method to learn a generative paradigm from historical tasks experience when the available adaptation knowledge of related domain is scarce. First, in specific-task learning, considering the distribution discrepancy challenge of the defect datasets, we reduce the Maximum Mean Discrepancy distance in Hilbert space to narrow the sample distribution to share domain-invariant and discriminative feature, enhancing available sample size. Second, in generalized-task learning, we construct an experience set with multiple such specific-task in the form of triples, in which the accuracy improvement ratio before and after adaptation is considered as pseudo labels. Specifically, we integrate a bi-level optimization view into meta learning-based detection framework to balance the specific- task and generalized-task, for generating across-task defect detection model automatically from experience. Furthermore, we validate our research methodology by considering oil pipeline detection processes, which is analyzed by Magnetic Flux Leakage signals. The experiment results shows effectiveness and superiority of our method.
|
|
16:50-17:10, Paper TuBT1.5 | |
>A Hybrid Modelling Approach for Parameter Estimation of Analytical Reflection Models in the Failure Analysis Process of Semiconductors (I) |
|
Kamm, Simon | University of Stuttgart |
Sharma, Kanuj | University of Stuttgart |
Jazdi, Nasser | University of Stuttgart - Institute of Industrial Automation And |
Weyrich, Michael | Univerity of Stuttgart, IAS |
Keywords: Machine learning, Deep Learning in Robotics and Automation, Failure Detection and Recovery
Abstract: Electronic devices are one of the key factors for recent advances in smart production systems or automotive. Reliability and robustness are key issues. To further increase this reliability, occurring failures in an electronic device has to be investigated in post-production failure analysis processes. One recent technique to detect and locate failures in electronic components is Time-Domain Reflectometry. This method offers the chance to detect several kinds of failures (e.g. a hard or soft failure) and localize the failure nondestructively. In theory, this can be determined following defined physical formulas. Nevertheless, the received signals are not perfect and mixed with noise from the measurement device or disturbed by nonoptimal material properties. In addition, complex architectures of devices are hard to model based on analytical models. Thus, these models solely are not sufficient for the failure analysis process. For this reason, a hybrid modeling approach is proposed, using a Machine Learning model in combination with physical models to detect and characterize the failure and its exact position. The Machine Learning model will be trained with simulated Time-Domain Reflectometry data.
|
|
17:10-17:30, Paper TuBT1.6 | |
>Non-Destructive Failure Analysis of Power Devices Via Time-Domain Reflectometry (I) |
|
Sharma, Kanuj | University of Stuttgart |
Kamm, Simon | University of Stuttgart |
Afanasenko, Valentyna | University of Stuttgart |
Muńoz Barón, Kevin | University of Stuttgart |
Kallfass, Ingmar | University of Stuttgart |
Keywords: Failure Detection and Recovery, Behavior-Based Systems, Calibration and Identification
Abstract: In power electronic applications, transistors are a vital component. They are, however, susceptible to failures due to degradation of the interconnections and the chip itself. This paper presents a non-destructive approach for failure detection and location in power electronic devices using time-domain reflectometry. The proposed measurement and data generation method is applied to a silicon-carbide power transistor where several characteristics (R, L, C, open, short) and the location of the failure is simulated and characterized. Moreover, the method is also used to find the intrinsic properties of the transistor such as parasitic inductance and capacitance. The data generated is mapped to physical equations, however, the reflected signal of the time-domain reflectometry can be noisy due to multiple discontinuities in the transmission path. Therefore, the simulation and measurement data can be used to train hybrid machine learning models for parameter extraction which automates the failure analysis in Industry4.0 processes to ensure a smart and reliable manufacturing process.
|
|
17:10-17:30, Paper TuBT1.7 | |
>Intelligent Fault Analysis Decision Flow in Semiconductor Industry 4.0 Using Natural Language Processing with Deep Clustering (I) |
|
Ezukwoke, Kenneth Ifeanyi | École Nationale Supérieure Des Mines De Saint-Étienne |
Toubakh, Houari | Ecole Des Mines De Saint-Étienne |
Hoayek, Anis | Ecole Des Mines De Saint Etienne |
Batton Hubert, Mireille | Institut Fayol Ecole Nationale Supérieure Des Mines De Saint Et |
Boucher, Xavier | Ecole Nationale Supérieure Des Mines De Saint-Etienne |
Gounet, Pascal | STMicroelectronics |
Keywords: Failure Detection and Recovery, AI-Based Methods, Diagnosis and Prognostics
Abstract: Microelectronics production failure analysis is a time-consuming and complicated task involving successive steps of analysis of complex process chains. The analysis is triggered to find the root cause of a failure and its findings, recorded in a reporting system using natural language. Fault analysis, physical analysis, sample preparation and package construction analysis are arguably the most used analysis activity for determining the root-cause of a failure. Intelligent automation of this analysis decision process using artificial intelligence is the objective of the FA 4.0 consortium; creating a reliable and efficient semiconductor industry. This research presents natural language processing (NLP) techniques to find a coherent representation of the expert decisions during fault analysis. The adopted methodology is a Deep learning algorithm based on β-variational autoencoder (β-VAE) for latent space disentanglement and Gaussian Mixture Model for clustering of the latent space for class identification.
|
|
TuBT2 Regular Session, Rhone 1 |
|
Industrial and Service Robotics |
Chair: Landgraf, Christian | Fraunhofer IPA |
Co-Chair: Puck, Lennart | FZI Forschungszentrum Informatik |
|
15:30-15:50, Paper TuBT2.1 | |
>A Target Person Locating Framework Based on Distributed Odometries |
|
Pang, Lei | State Key Laboratory of Management and Control for Complex Syste |
Cao, Zhiqiang | Institute of Automation, Chinese Academy of Sciences |
Yu, Junzhi | Chinese Academy of Sciences |
Li, Zhonghui | Beijing Engo Technology Co., Ltd |
Zhao, Lei | UNISOC Technologies Co., Ltd |
Lv, Pei | UNISOC Technologies Co., Ltd |
Keywords: Industrial and Service Robotics, Sensor-based Control
Abstract: The ability to follow a specified target person by a mobile robot plays an important role. The continuity of person following is a key factor to the performance and practicability. In the practical person following, the failure easily occurs when the target person is out of the sensing range of the robot, which destroys the continuity of the person following and still remains a challenge. To solve the problem of following recovery, a locating framework based on distributed odometries is proposed, which can locate the target effectively with a robot-end odometry and a target-end odometry. In this framework, the robot-end odometry and the target-end odometry are used to estimate poses of the robot and positions of the target person, respectively. These two odometries are fused with the combination of locations of the target person in the robot coordinate system. On this basis, the robot can obtain the estimated target location by receiving the position information from the target-end odometry with a wireless network, which is then transformed to the coordinate in the robot-end odometry. When a missing case occurs, the mobile robot can continue to retrieve the target person by moving along the sampling transformed positions. An implement with the robot-end LiDAR odometry and target-end wearable visual-inertial odometry is given to evaluate the proposed locating framework. Experiments reveal that a mobile robot can recover the normal person following even if the target person is far away from the robot.
|
|
15:50-16:10, Paper TuBT2.2 | |
>Mobility Improvement on the Two-Wheeled Dynamically Balanced Robot – J4.beta |
|
Hsu, Yucheng | National Taipei University of Technology |
Lin, Ming-Chang | Department of Mechanical Engineering, Chung Yuan Christian Unive |
Li, Chih-Hung G. | National Taipei University of Technology |
Keywords: Motion Control, Industrial and Service Robotics, Autonomous Vehicle Navigation
Abstract: Dynamically self-balancing wheeled robots possess the potentials of having a small footprint, low base-to-height ratios, high accelerations and speeds, and low costs. They are suitable for autonomous mobile robots (AMRs) that work in human-centric environments. As part of our ongoing effort in creating self-balancing wheeled robots, in this article, we reported the development of our latest model – J4.beta. In contrast to the previous model – J4.alpha, the new model has a greater dynamic mass-to-total ratio; thus, the acceleration and the ultimate speed are both increased. Here, the maximum speed of 4.4 m/s of the motion platform is achievable by J4.beta. We analyzed the system dynamics and had confirmations from measurements; a speed servo system was developed based on PID control. For simplifying the dynamics of the AMR, a stepper motor instead of a DC motor was adopted for the actuation of the dynamic mass; the overall controlled plant could be approximated as a second-order system. To acquire the PID coefficients, a series of road tests were performed on cement floors in a common office building. A set of suitable PID coefficients was obtained and verified by three speeds: 0.5 m/s, 1 m/s, and 2 m/s. The speed curves exhibited fast ramp-up, low overshoot, setpoint matching, and low oscillation. For regulation testing, a zero speed was set and external disturbance was applied. The robot was witnessed to slow down rapidly and remain stationary without intensive oscillation. In addition that autonomous navigation and remote control were developed, the sampling rate of the control system was largely upgraded to 4k Hz to achieve a better tracking and regulation ability.
|
|
16:10-16:30, Paper TuBT2.3 | |
>VDB-Mapping: A High Resolution and Real-Time Capable 3D Mapping Framework for Versatile Mobile Robots |
|
Grosse Besselmann, Marvin | FZI Forschungszentrum Informatik |
Puck, Lennart | FZI Forschungszentrum Informatik |
Steffen, Lea | FZI Research Center for Information Technology, 76131 Karlsruhe, |
Roennau, Arne | FZI Forschungszentrum Informatik, Karlsruhe |
Dillmann, Rüdiger | FZI - Forschungszentrum Informatik - Karlsruhe |
Keywords: Industrial and Service Robotics, Autonomous Vehicle Navigation, Motion and Path Planning
Abstract: Recent developments of versatile mobile robots demand a precise and complete volumetric representation of space for path and mission planning. However, the immense amount of accumulated raw data points from 3D sensors quickly becomes unmanageable. Thereby, becoming infeasible regarding memory footprint and navigation itself. To abstract the necessary information into a usable representation for navigation, usually volumetric grid maps are applied. Although these approaches solve the memory and handling issues, current implementations tend to require high computational time for the insertion of new data. As a result the generated maps are often not up to date and incomplete due to dropped sensor data. This greatly impairs their usefulness for navigating robot systems in dynamically changing environments. In order to solve these issues we propose a novel probabilistic mapping framework based on OpenVDB, which is a hierarchical tree structure with efficient access methods to discretized volumetric data. By utilizing the fast direct access to the bitmasks of the underlying OpenVDB data structure, a significant performance boost for data insertion is achieved. Thus, enabling real-time processing of the incoming raw 3D sensor data. An in-depth evaluation of the proposed framework is provided, including a performance and memory footprint comparison against the often employed OctoMap framework. The evaluation reveals, that the presented VDB-Mapping is able to efficiently process long range data on high resolution grids.
|
|
16:30-16:50, Paper TuBT2.4 | |
>Design of a Bio-Inspired Quadruped Robot with Scalable Torso |
|
Liu, Yixiang | Shandong University |
Bi, Qing | Shandong Institute of Advanced Technology, Chinese Academy of Sc |
Li, Yibin | Shandong University |
Keywords: Biomimetics, Industrial and Service Robotics
Abstract: Quadrupeds like cats are able to pass through vary narrow gaps or holes which are much smaller than their body. This is owing to the inherent softness and flexibility of their body structures especially the torso. However, very few of current quadruped robots allow their bodies to shrink down when facing small holes. Therefore, this paper presents the design of a bio-inspired quadruped robot with scalable torso. A foldable mechanism with high deformability is firstly designed by integrating several spatial multi-bar linkages. The scalable torso and the quadruped robot are further designed based on the foldable mechanism. The proposed torso enables the quadruped robot to better deal with the obstacles including gaps, gullies, and stairs. Simulation experiments demonstrate that the quadruped robot with scalable torso has better performance of high-speed locomotion compared with rigid torso, such as increased stride length and lower cost of transport.
|
|
16:50-17:10, Paper TuBT2.5 | |
>A Topological Solution of Entanglement for Complex-Shaped Parts in Robotic Bin-Picking |
> Video
|
|
Zhang, Xinyi | Osaka University |
Koyama, Keisuke | Osaka University |
Domae, Yukiyasu | The National Institute of Advanced Industrial Science and Techno |
Wan, Weiwei | Osaka University |
Harada, Kensuke | Osaka University |
Keywords: Industrial and Service Robotics, Hybrid Strategy of Intelligent Manufacturing, Robust Manufacturing
Abstract: This paper addresses the problem of picking up only one object at a time avoiding any entanglement in bin-picking. To cope with a difficult case where the complex-shaped objects are heavily entangled together, we propose a topology-based method that can generate non-tangle grasp positions on a single depth image. The core technique is entanglement map, which is a feature map to measure the entanglement possibilities obtained from the input image. We use entanglement map to select probable regions containing graspable objects. The optimum grasping pose is detected from the selected regions considering the collision between robot hand and objects. Experimental results show that our analytic method provides a more comprehensive and intuitive observation of entanglement and exceeds previous learning-based work in success rates. Especially, our topology-based method does not rely on any object models or time-consuming training process, so that it can be easily adapted to more complex bin-picking scenes.
|
|
17:10-17:30, Paper TuBT2.6 | |
>A Hybrid Neural Network Approach for Increasing the Absolute Accuracy of Industrial Robots |
> Video
|
|
Landgraf, Christian | Fraunhofer IPA |
Ernst, Kilian | Fraunhofer IPA |
Schleth, Gesine | Fraunhofer Institute for Manufacturing Engineering and Automatio |
Fabritius, Marc | IPA Frauenhofer |
Huber, Marco F. | University of Stuttgart |
Keywords: Calibration and Identification, Deep Learning in Robotics and Automation, Industrial and Service Robotics
Abstract: The comparatively poor positioning accuracy of industrial robots limits or even prevents their use in many industrial applications. In contrast to other fields of robotic research, robot accuracy improvement has not been significantly boosted by machine learning-based methods yet. For this reason, we carried out four comprehensive series of measurements using a high-precision laser tracker together with a widely used 6-axis articulated robot. The data will be made publicly available to serve as benchmark data for different techniques. Along with the dataset, this paper introduces a hybrid neural network-based approach to compensate both geometric and non-geometric error sources and increase robot positioning accuracy. We compare our method to previous works and demonstrate advanced results.
|
|
TuBT3 Regular Session, Rhone 2 |
|
Deep Learning in Robotics and Automation 2 |
Chair: Orsag, Matko | University of Zagreb, Faculty of Electrical Engineering and Computing |
Co-Chair: Roberge, Jean-Philippe | École De Technologie Supérieure |
|
15:30-15:50, Paper TuBT3.1 | |
>Motion Prediction Based on Multiple Futures for Dynamic Obstacle Avoidance of Mobile Robots |
|
Zhang, Ze | Chalmers University of Technology |
Dean-Leon, Emmanuel | Technischen Universitaet Muenchen |
Karayiannidis, Yiannis | Chalmers University of Technology & KTH Royal Institute of Techn |
Akesson, Knut | Chalmers University of Technology |
Keywords: Deep Learning in Robotics and Automation, Collision Avoidance, Optimization and Optimal Control
Abstract: The ability to decide and adjust actions according to motion prediction of dynamic obstacles offers a more flexible planning scheme and ampler reaction time to avoid potential impact. Prediction-based collision avoidance implies a two-stage decision-making process from motion prediction to action planning. One of the challenges in motion prediction is the movements of other objects are usually non-deterministic and governed by multimodal probabilistic models. Many studies have been made on motion prediction of dynamic obstacles and action planning for mobile robots separately. The objective of this work is to explore the coherence of them in terms of multiple future prediction by combining a data-driven motion prediction approach with a model-based control strategy. More specifically, we integrate motion prediction from deep learning models, Mixture Density Networks (MDNs) with a Non-linear Model Predictive Control (NMPC) framework. The deep learning models produce the multimodal probability distribution of future positions of dynamic obstacles, which is utilized by the MPC controller as constraints. We show through simulation that the selected deep learning model provides valid predictions of motion in a complex dynamic environment. The prediction result then endows the model predictive controller with the capability to avoid dynamic obstacles in advance.
|
|
15:50-16:10, Paper TuBT3.2 | |
>Soft Robotics Approach to Autonomous Plastering |
|
Polic, Marsela | University of Zagreb |
Maric, Bruno | Univeristy of Zagreb, Faculty of Electrical Engineering and Comp |
Orsag, Matko | University of Zagreb, Faculty of Electrical Engineering and Comp |
Keywords: Deep Learning in Robotics and Automation, Cyber-physical Production Systems and Industry 4.0, AI-Based Methods
Abstract: This paper presents an industrial soft robotics application for the autonomous plastering of complex shaped surfaces, using a collaborative industrial manipulator. In the core of the proposed system is the deep learning based soft body modeling, i.e. deformation estimation of the flexible plastering knife tool. The estimation relies on visual feedback and a deep convolution neural network (CNN). The transfer learning approach and specially designed dataset generation procedures were developed in the learning phase. The estimated deformation of the plastering knife is then used to control the knife inclination with respect to the treated surface, as one of the essential control variables in the plastering procedure. The developed system is experimentally validated, including both the CNN based deformation estimation, as well as its performance in the knife inclination control.
|
|
16:10-16:30, Paper TuBT3.3 | |
>SynPick: A Dataset for Dynamic Bin Picking Scene Understanding |
> Video
|
|
Periyasamy, Arul Selvam | University of Bonn |
Schwarz, Max | University Bonn |
Behnke, Sven | University of Bonn |
Keywords: Deep Learning in Robotics and Automation, Computer Vision in Automation, Simulation and Animation
Abstract: We present SynPick, a synthetic dataset for dynamic scene understanding in bin-picking scenarios. In contrast to existing datasets, our dataset is both situated in a realistic industrial application domain---inspired by the well-known Amazon Robotics Challenge (ARC)---and features dynamic scenes with authentic picking actions as chosen by our picking heuristic developed for the ARC 2017. The dataset is compatible with the popular BOP dataset format. We describe the dataset generation process in detail, including object arrangement generation and manipulation simulation using the NVIDIA PhysX physics engine. To cover a large action space, we perform untargeted and targeted picking actions, as well as random moving actions. To establish a baseline for object perception, a state-of-the-art pose estimation approach is evaluated on the dataset. We demonstrate the usefulness of tracking poses during manipulation instead of single-shot estimation even with a naive filtering approach. The generator source code and dataset are publicly available.
|
|
16:30-16:50, Paper TuBT3.4 | |
>Tactile-Based Object Recognition Using a Grasp-Centric Exploration |
> Video
|
|
Roberge, Jean-Philippe | École De Technologie Supérieure |
L'Écuyer-Lapierre, Louis | École De Technologie Supérieure |
Kwiatkowski, Jennifer | École De Technologie Supérieure |
Nadeau, Philippe | University of Toronto |
Duchaine, Vincent | Ecole De Technologie Superieure |
Keywords: Force and Tactile Sensing, Deep Learning in Robotics and Automation
Abstract: As humans, our grasping and manipulation skills are highly dependent on our ability to perceive tactile properties. Conversely, most of today's robotic operations still relies predominantly on visual feedback for identifying the objects that need to be grasped and handled. In this work, we study the problem of recognizing everyday objects based solely on their tactile attributes. This has a significant practical value, as it could allow object identification even when visual sensing is impossible, or assist vision in difficult contexts. Our method consists of acquiring multi-modal tactile sensing data during a quick grasp-centric exploration phase, with minimal operational cost. Our algorithm was able to recognize objects from a considerably-large set of 50 general purpose items with an accuracy of 98.1%. Moreover, we show that it is possible to reliably identify a large proportion of these objects by only analyzing the deformation pattern that they undergo during compression. Further, we study our method's ability to learn relevant tactile properties to classify new objects. We also share our tactile sensing database that contains various sensor data acquired from more than 1600 experiments, which was used for this work. Finally, we discuss the relative performance and role of each tactile modality for differentiating objects.
|
|
16:50-17:10, Paper TuBT3.5 | |
>LazyDAgger: Reducing Context Switching in Interactive Imitation Learning |
> Video
|
|
Hoque, Ryan | University of California, Berkeley |
Balakrishna, Ashwin | University of California, Berkeley |
Putterman, Carl | University of California, Berkeley |
Luo, Michael | UC Berkeley |
Brown, Daniel | UC Berkeley |
Seita, Daniel | University of California, Berkeley |
Thananjeyan, Brijen | UC Berkeley |
Novoseller, Ellen | University of California, Berkeley |
Goldberg, Ken | UC Berkeley |
Keywords: Deep Learning in Robotics and Automation, Failure Detection and Recovery, Human Factors and Human-in-the-Loop
Abstract: Corrective interventions while a robot is learning to automate a task provide an intuitive method for a human supervisor to assist the robot and convey information about desired behavior. However, these interventions can impose significant burden on a human supervisor, as each intervention interrupts other work the human is doing, incurs latency with each context switch between supervisor and autonomous control, and requires time to perform. We present LazyDAgger, which extends the interactive imitation learning (IL) algorithm SafeDAgger to reduce context switches between supervisor and autonomous control. We find that LazyDAgger improves the performance and robustness of the learned policy during both learning and execution while limiting burden on the supervisor. Simulation experiments suggest that LazyDAgger can reduce context switches by an average of 60% over SafeDAgger on 3 continuous control tasks while maintaining state-of-the-art policy performance. In physical fabric manipulation experiments with an ABB YuMi robot, LazyDAgger reduces context switches by 60% while achieving a 60% higher success rate than SafeDAgger at execution time.
|
|
17:10-17:30, Paper TuBT3.6 | |
>Bimanual Shelf Picking Planner Based on Collapse Prediction |
> Video
|
|
Motoda, Tomohiro | Osaka University |
Petit, Damien | Osaka University |
Wan, Weiwei | Osaka University |
Harada, Kensuke | Osaka University |
Keywords: Deep Learning in Robotics and Automation, Logistics, Machine learning
Abstract: In logistics warehouse, since many objects are randomly stacked on shelves, it becomes difficult for a robot to safely extract one of the objects without other objects falling from the shelf. In previous works, a robot needed to extract the target object after rearranging the neighboring objects. In contrast, humans extract an object from a shelf while supporting other neighboring objects. In this paper, we propose a bimanual manipulation planner based on collapse prediction trained with data generated from a physics simulator, which can safely extract a single object while supporting the other object. We confirmed that the proposed method achieves more than 80% success rate for safe extraction by real-world experiments using a dual-arm manipulator.
|
|
TuBT4 Special Session, Rhone 3A |
|
Simulation and Modelling for the Semiconductor Manufacturing and Supply
Chain Networks |
Chair: Ehm, Hans | Infineon Technologies AG |
Co-Chair: Moench, Lars | University of Hagen |
Organizer: Ismail, Abdelgafar | Infineon Technologies AG |
Organizer: Ehm, Hans | Infineon Technologies AG |
|
15:30-15:50, Paper TuBT4.1 | |
>Transparent Reporting in Semiconductor Supply Chain: Qualitative Analysis of Planning Deviations and Related KPIs (I) |
|
Laura De Roget Aysa, Laura De | Universitat Politčcnica De Catalunya |
Moder, Patrick | Infineon Technologies AG |
Keywords: Cyber-physical Production Systems and Industry 4.0, Manufacturing, Maintenance and Supply Chains, Semiconductor Manufacturing
Abstract: Semiconductor supply chains, or networks respectively, consist of globally distributed sites with production processes spanning multiple countries and several trips around the world [1]. Hence, information sharing and communication between sites follows complex processes, and in case problems occur, finding a solution or the root-cause is a challenge. For a better communication and transparency in supply chain planning, reports are created. In these reports, data regarding customers, order promises, factory commits, deliveries, and status updates are present. Some reports are used only occasionally. The predominant cause of their insufficient usage is incorrect, intransparent or unreliable data that leads experts to not trust them [2]. Some reports could have more value add if trustworthiness is increased. In this project, we tackle this opportunity by increasing the report’s usefulness through improvements in transparency. We approach this goal by qualitatively analyzing correlations of reported deviations to major KPIs and provide a graphical representation in a use case from the semiconductor supply chain domain.
|
|
15:50-16:10, Paper TuBT4.2 | |
>From Global Planning towards Local Execution – Benefits of a Vertical Supply Chain Integration (I) |
|
Schiller, Christian | Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Wanner, Marc | Infineon Technologies AG |
Keywords: Factory Automation, Process Control, Manufacturing, Maintenance and Supply Chains
Abstract: The Semiconductor industry with its long lead times and volatile demand has always been in the forefront of global flexible manufacturing. The full potential of a global production network requires a seamless interaction from Order Management Systems to supply provision Systems to Global Production solvers and to Detailed Scheduling solvers in an integrated control loop system. The vertical supply chain Integration between the global production Program solver and the factories detailed scheduling solvers is showcased in this article. A solution is shown how factories detailed scheduling solvers interact with the production program solvers to keep commits, allow a fast production and utilize the available capacities.
|
|
16:10-16:30, Paper TuBT4.3 | |
>Simulating the COVID19-Pandemic and Its Impact on the Semiconductor Supply Chain (I) |
|
Butgereit, Lea | Infineon Technologies |
Lopera, Manuel Fernando | Infineon Technologies |
Ismail, Abdelgafar | Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Keywords: Simulation and Animation, Manufacturing, Maintenance and Supply Chains, Semiconductor Manufacturing
Abstract: During the COVID-19 pandemic, Infineon's supply chain (SC) impacts were defined by governmental decisions that affected transit times as well as changes in demand. To understand these impacts, an epidemiological model and a simplified aggregated supply chain network are developed and simulated using System Dynamics (SD) and Agent-Based (AB) modelling. This simulation assesses the effect of COVID-19 on different KPI´s by creating multiple scenarios assessing three reactive measures. As a quantification for government restrictions, the Oxford COVID-19 Government Response Tracker (OxCGRT) is used. Factors like the vaccination and variants of concern (VOC) are considered, allowing for good estimation of the current and the future number of infections and of the government response. The model presents variations in inventory levels, cycle time, demand fulfillment and backlog levels for the customer as KPIs. It has been found that with a redundant and a flexible SC, the impacts of the pandemic can be absorbed.
|
|
16:30-16:50, Paper TuBT4.4 | |
>Plan Stability and the Challenge of Value-Added versus Non-Value-Added Plan Changes in the Semiconductor Industry (I) |
|
Eisenmann, Leon | Infineon Technologies AG |
Welling, Tobias Leander | Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Keywords: Planning, Scheduling and Coordination, Intelligent and Flexible Manufacturing, Semiconductor Manufacturing
Abstract: Challenges of long lead times and complex manufacturing processes in the semiconductor industry emphasize the necessity of stable supply chain planning. However, in order to cope with highly volatile, fast paced and competitive market environments, supply chain processes need to react flexible and dynamically. The resulting changes in production plans can be described by plan instability. Generally, plan instability is shaped by different causes and effects on the supply chain and can feature value added and non-value added characteristics. To finally differentiate between positive and negative plan instability, an approach aimed at monitoring and analyzing the effects of plan instability is required. In the semiconductor industry, re-planning processes are typically performed on a daily basis which lead to the necessity to adjust the order confirmations accordingly. Moreover, these changes in order confirmations could serve as an expressive evaluation for plan instabilities in the ATP picture and their value-added. This paper proposes two different approaches, which apply the described observation of confirmation changes to assess plan instability.
|
|
16:50-17:10, Paper TuBT4.5 | |
>Comparison of Different Production Planning Formulations with Workload-Dependent Lead Times (I) |
|
Bierbuesse, Jan | University of Hagen |
Moench, Lars | University of Hagen |
Keywords: Semiconductor Manufacturing, Planning, Scheduling and Coordination
Abstract: We study production planning formulations for wafer fabs based on different types of clearing functions. The first clearing function (CF) type frequently used in the literature is based on an aggregated work in process (WIP) variable while the second CF type is based on disaggregated WIP levels for different products and their processing steps. These so-called multi-dimensional CFs (MDCFs) seem to be a promising way for making better production planning decisions. However, the resulting optimization formulation for MDCFs is non-linear and computationally more challenging compared to formulations based on the conventional CF type. We compare a planning formulation based on the operation-machinebased (OM) MDCF with the conventional allocated CF (ACF) formulation based on a scaled-down wafer fab model. This model contains several engineering products in addition to salable products. Simulation experiments demonstrate that the higher computational effort for the formulations with MDCFs does not necessarily lead to better results if the CFs for the ACF formulation are correctly parameterized.
|
|
17:10-17:30, Paper TuBT4.6 | |
>Mitigating the Bullwhip Effect in the End-To-End Semiconductor Supply Chains Using End-To-End Supply Chain Simulation (I) |
|
Ismail, Abdelgafar | Infineon Technologies AG |
Ehm, Hans | Infineon Technologies AG |
Keywords: Simulation and Animation, Manufacturing, Maintenance and Supply Chains, Semiconductor Manufacturing
Abstract: The fast technological developments in the world are putting high pressure on semiconductor manufacturers and the users of semiconductors to achieve an agile, adaptable and aligned global network of manufacturing on the end-to-end supply chain level throughout all business cycles. Due to the high complexity of the semiconductor manufacturing processes combined with the extreme demand volatility of the semiconductor market and the upstream position in the supply chain, semiconductor manufacturers are exposed to these amplifications of demand fluctuations, commonly known as the bullwhip effect, to a much greater extent. As results of the bullwhip effect, strong demand dynamics in the semiconductor industry could lead to substantial operational consequences. In this abstract paper we address the challenge of the bullwhip effect along the end-to-end semiconductor supply chains and the operational consequences resulting from it. Additionally, we focus on simulation as a decision support system for the mitigation of the bullwhip effect to leverage the collaboration among the supply chain partners.
|
|
TuBT5 Special Session, Rhone 3B |
|
Advance in Sustainable Automation |
Chair: Li, Congbo | Chongqing University |
Co-Chair: Robba, Michela | University of Genoa |
Organizer: Li, Congbo | Chongqing University |
Organizer: Tang, Ying | Rowan University |
Organizer: Jiang, Zhigang | Wuhan University of Science and Technology |
Organizer: Chen, Xingzheng | Chongqing University |
|
15:30-15:50, Paper TuBT5.1 | |
>A Timing Decision-Making Method for Active Remanufacturing Considering Reliability and Environmental Impact (I) |
|
Ouyang, Jun | Wuhan University of Science and Technology |
Jiang, Zhigang | Wuhan University of Science and Technology |
Zhu, Shuo | Wuhan University of Science and Technology |
Keywords: Remanufacturing, Sustainability and Green Automation, Environment Monitoring and Management
Abstract: For the uncertainty problem of remanufacturing blanks, an active remanufacturing timing decision method that considers reliability and environmental impact is proposed in this paper. In this method, the reliability of the product in the service stage is firstly used to characterize the change in its quality. In addition, an improved average rank method is proposed to improve the accuracy of reliability prediction, so as to preliminarily determine the time range of active remanufacturing. Then, the environmental impact of the whole life cycle of used products is quantitatively analyzed, the function of average annual energy consumption and annual waste discharge are applied as indicators. The multi-objective optimization problem is solved with genetic algorithm (GA), and the best time for active remanufacturing is determined. A case study on remanufacturing a used engine is demonstrated to validate the proposed method.
|
|
15:50-16:10, Paper TuBT5.2 | |
>Data-Driven Method for Predicting Energy Consumption of Machine Tool Spindle Acceleration (I) |
|
Huang, Binbin | Wuhan University of Science and Technology |
Jiang, Guozhang | College of Machinery and Automation, Wuhan University of Science |
Yan, Wei | School of Automobile and Traffic Engineering, Wuhan University O |
Jiang, Zhigang | Wuhan University of Science and Technology |
Lu, Chenxun | Wuhan University of Science and Technology |
Zhang, Hua | Wuhan University of Science and Technology |
Keywords: Sustainability and Green Automation, Power and Energy Systems automation, Energy and Environment-aware Automation
Abstract: As an essential operation, spindle acceleration occurs frequently in the machining process, the energy consumption of which has an important impact on the energy efficiency of machine tools, cannot be ignored. However, due to its energy characteristics of short duration, high power peak and complex electromechanical operating of the spindle motor, the energy consumption of the spindle acceleration process is difficult to calculate accurately. To fill this gap, a data-driven method for machine tool spindle acceleration energy prediction is proposed in this paper. Firstly, the energy characteristics of spindle acceleration are studied, and a dataset for the energy prediction is determined. Secondly, an automatic extraction algorithm is developed to extract the time data of power peak, and then a framework for data collection and preprocessing is proposed. Thirdly, a spindle acceleration energy prediction model is established with Back-propagation Neural Network based on the Genetic Algorithm (GA-BP), and the network structure and the operation process are also studied. Finally, a case study of spindle acceleration is given to verify the validity of the proposed approach and model, and the accuracy is also verified with other algorithms.
|
|
16:10-16:30, Paper TuBT5.3 | |
>A Novel Integer Linear Programming Formulation for Job-Shop Scheduling Problems (I) |
|
Liu, Anbang | Tsinghua University |
Luh, Peter | University of Connecticut |
Yan, Bing | Rochester Institute of Technology |
Bragin, Mikhail | University of Connecticut |
Keywords: Manufacturing, Maintenance and Supply Chains, Planning, Scheduling and Coordination, Intelligent and Flexible Manufacturing
Abstract: Job-shop scheduling is an important but difficult problem arising in low-volume high-variety manufacturing. It is usually solved at the beginning of each shift with strict computational time requirements. To obtain near-optimal solutions with quantifiable quality within strict time limits, a direction is to formulate them in an Integer Linear Programming (ILP) form so as to take advantages of widely available ILP methods such as Branch-and-Cut (B&C). Nevertheless, computational requirements for ILP methods on existing ILP formulations are high. In this paper, a novel ILP formulation for minimizing total weighted tardiness is presented. The new formulation has much fewer decision variables and constraints, and is proven to be tighter as compared to our previous formulation. For fast resolution of large problems, our recent decomposition-and-coordination method “Surrogate Absolute-Value Lagrangian Relaxation” (SAVLR) is enhanced by using a 3-segment piecewise linear penalty function, which more accurately approximates a quadratic penalty function as compared to an absolute-value function. Testing results demonstrate that our new formulation drastically reduces the computational requirements of B&C as compared to our previous formulation. For large problems where B&C has difficulties, near-optimal solutions are efficiently obtained by using the enhanced SAVLR under the new formulation.
|
|
16:30-16:50, Paper TuBT5.4 | |
>An Energy Management System for Demand Response, Energy Efficiency and Risk Management in Sustainable Districts (I) |
|
Robba, Michela | University of Genoa |
Ferro, Giulio | University of Genoa |
Delfino, Federico | Universitŕ Degli Studi Di Genova |
Rossi, Mansueto | University of Genoa |
Bianco, Giovanni | Universitŕ Degli Studi Di Genova |
Parodi, Luca | University of Genoa |
Keywords: Smart Grids, Building Automation, Distributed Generation and Storage
Abstract: An overall energy management platform will be presented for the management of sustainable districts that has been developed within three main Italian projects: Living Grid (funded by Italian Ministry of Research for the Innovation Technology), for what concerns the module related islanding operation in demand response programs; Pick Up (funded by Liguria Region, Italy) for energy efficiency and demand response; Restabilise 4.0 (funded by the Italian Competence Center for Industry 4.0, Start 4.0) for reliability and risk management in smart grids. In particular, the Energy Management System (EMS) has been tested on the Savona Campus Smart Polygeneration Microgrid (SPM) and Smart Energy Building (SEB), which include production plants from renewable resources (photovoltaics, geothermal, solar panels), tri-generative gas microturbines, absortion chillers, thermal and electrical storage systems, heat pumps, thermal and electrical loads. The presentation will show the EMS, its testing and the experimental campaigns in the different projects.
|
|
16:50-17:10, Paper TuBT5.5 | |
>Reinforcement Learning-Based Selective Disassembly Sequence Planning for the End-Of-Life Products with Structure Uncertainty |
|
Zhao, Xikun | Chongqing University |
Li, Congbo | Chongqing University |
Tang, Ying | Rowan University |
Cui, Jiabin | Chongqing University |
Keywords: Sustainable Production and Service Automation
Abstract: Selective disassembly sequence planning (SDSP) is regarded as an efficient strategy to determine optimal disassembly sequences for extracting target parts (TP) from complex end-of-life (EOL) products. Previous research assumes that all EOL products have the same structure and the optimal selective disassembly sequences are given before the EOL products are removed. However, the products have different operation states during their use stage, which results in high structure uncertainty of EOL products. The structure uncertainty of EOL products often makes the predetermined selective disassembly sequences impractical for minimizing disassembly time and maximizing disassembly profit. This paper undertakes this challenge by integrated reinforcement learning (RL) to determine the optimal disassembly sequences adaptive to the structure uncertainty of the EOL products. Firstly, a multi-level selective disassembly hybrid graph model (MSDHGM) is developed to illustrate the contact, precedence, and level relationships among parts. Then, the SDSP is formulated as a finite Markov Decision Process and a deep Q-network based selective disassembly sequence planning (DQN-SDSP) is proposed. Finally, extensive comparative experiments are conducted to verify the proposed method compared with NSGA-II and ABC algorithms.
|
|
17:10-17:30, Paper TuBT5.6 | |
>Production Rate Control in Offshore Wind Maintenance Planning (I) |
|
Pentinga, Luuk | University of Groningen |
Kilic, Onur Alper | University of Groningen |
Teunter, Ruud | Rijksuniversiteit Groningen |
Veldman, Jasper | University of Groningen |
Keywords: Optimization and Optimal Control, Manufacturing, Maintenance and Supply Chains, Power and Energy Systems automation
Abstract: Equipment availability and reliability is essential to achieve an efficient and continuous production flow. However, the impact of production decisions on system degradation is often neglected. In this paper, we consider an energy system for which we can use condition information to determine a joint condition-based maintenance and production control policy. The system is modeled as a continuous-time renewal-reward system with continuous degradation states. We transform the model to a higher-dimensional auxiliary problem, which is used to determine structural properties of optimal policies, and to efficiently determine such policies. We use this methodology to determine optimal joint condition-based maintenance and production control policies with both continuous and discrete production rates. The numerical results show that expected profits can be significantly increased through a dynamic production control policy, while simultaneously resulting in far less maintenance interventions. Furthermore, we show the efficiency of our methology by benchmarking it against typical alternative methods.
|
|
TuBT6 Regular Session, St Clair 1 |
|
Formal Methods in Robotics and Automation |
Chair: Le Corronc, Euriell | LAAS-CNRS |
Co-Chair: Lennartson, Bengt | Chalmers University of Technology |
|
15:30-15:50, Paper TuBT6.1 | |
>Observer-Based Detection and Localization of Time Shift Failures in (max, +)-Linear Systems |
|
Paya, Claire | Université Paul Sabatier, LAAS-CNRS, STMicroelectronics |
Le Corronc, Euriell | LAAS-CNRS |
Pencolé, Yannick | LAAS/CNRS |
Vialletelle, Philippe | STMicroelectronics |
Keywords: Discrete Event Dynamic Automation Systems, Petri Nets for Automation Control
Abstract: This paper addresses a problem of failure detection and localization in production lines modeled, through Timed Event Graphs (TEG), as (max,+) linear systems with disturbances, over which observers can be developed. The state of the observed system is estimated and an indicator that returns true if a time shift failure is detected is defined. The localization step is proposed for elementary structures of TEG through the results of the detection
|
|
15:50-16:10, Paper TuBT6.2 | |
>Stochastic Image-Based Visual Predictive Control |
|
Sajjadi, Sina | University of Regina |
Fallah, Mostafa M.H. | Ryerson University |
Mehrandezh, Mehran | University of Regina |
Janabi-Sharifi, Farrokh | Ryerson University |
Keywords: Formal Methods in Robotics and Automation, Computer Vision in Automation
Abstract: Image-based visual predictive controllers have gained attention due to their optimality and constraint-handling capabilities. However, their performance deteriorates in presence of the modelling and measurement uncertainties. This paper presents a stochastic image-based visual predictive control method to overcome some shortcomings of the previous schemes cited in literature. In particular, the proposed approach provides a systematic solution to address the image-based constraint compliance in presence of the measurement and modelling uncertainties. The proposed method was implemented on a 6-DOF Denso robot via simulation.
|
|
16:10-16:30, Paper TuBT6.3 | |
>Incremental Abstraction - a Formal Perspective on Petri Net Reduction |
|
Lennartson, Bengt | Chalmers University of Technology |
Keywords: Petri Nets for Automation Control, Formal Methods in Robotics and Automation, Discrete Event Dynamic Automation Systems
Abstract: Bounded Petri nets are in this paper reduced by an incremental abstraction method based on visible bisimulation. An arbitrary bounded Petri net is decomposed into subsystems that are easily transformed to a modular transition system. The basic principle is that places in a Petri net can be interpreted as the synchronous composition of bounded buffers, and a sequence of places can be reduced analytically to a place with extended capacity. Additional restrictions, such as mutual exclusion among shared resources, are formulated as predicates that are easily translated to ordinary transition systems. Since the reduction preserves CTL*-X expressions, it can be used as a stand-alone model checking tool, where temporal properties of the reduced model are easily evaluated. This approach is shown to be very efficient compared to the best-known model checking algorithms available in the software tool nuXmv.
|
|
16:30-16:50, Paper TuBT6.4 | |
>Modelling and Analysis of Product Platforms and Assembly Sequences with Respect to Variability |
|
Ebrahimi, Amir Hossein | Chalmers |
Akesson, Knut | Chalmers University of Technology |
Keywords: Formal Methods in Robotics and Automation, Domain-specific Software and Software Engineering
Abstract: A challenge for highly configurable products is that the manufacturing system has to support all possible variants that can be configured. The production system is often highly automated and the link between the product and the assembly system can be expressed through operations where each operation models how a part in the bill-of-material is assembled to the final product. Typically, operations have precedence constraints that express that certain parts have to be assembled before other parts. However, it is important to make sure that all possible variants can be successfully assembled while satisfying all precedence constraints. In this paper we present fully automated analysis method that is able to analyze the existence of product configurations of that cannot be successfully assembled.
|
|
16:50-17:10, Paper TuBT6.5 | |
>Cartesian Inertia Optimisation Via Redundancy Resolution for Physical Human Robot Interaction |
|
Sutjipto, Sheila | University of Technology, Sydney |
Woolfrey, Jon | University of Technology Sydney |
Carmichael, Marc | Centre for Autonomous Systems |
Paul, Gavin | University of Technology, Sydney |
Keywords: Formal Methods in Robotics and Automation, Optimization and Optimal Control, Compliance and Impedance Control
Abstract: The objective of introducing robotic manipulators into human-centric domains is to improve the efficacy of tasks in a safe and practical manner. The shift toward collaborative manipulator platforms has facilitated physical human-robot interaction (pHRI) in such environments. Often, these platforms are kinematically redundant and possess more degrees of freedom than needed to complete a desired task. When no additional task is defined, it is possible for the manipulator to converge upon joint configurations that are unfavourable for the collaborative task. Consequently, there is potential for the posture of the manipulator to affect the interaction experienced. This paper investigates an inertia-based optimization control method for redundant manipulators interacting with active agents. The inertia-based reconfiguration is evaluated through simulations and quantified with real-life experiments conducted with a robot-robot dyad. It was found that resolving redundancy to reconfigure the Cartesian inertia reduced the energy expenditure of the active agent during the interaction.
|
|
17:10-17:30, Paper TuBT6.6 | |
>Online Partial Conditional Plan Synthesis for POMDPs with Safe-Reachability Objectives: Methods and Experiments |
|
Wang, Yue | Facbook Inc |
Redwan Newaz, Abullah Al | North Carolina Agricultural and Technical State University |
Hernández, Juan David | Cardiff University |
Chaudhuri, Swarat | Rice University |
Kavraki, Lydia | Rice University |
Keywords: Formal Methods in Robotics and Automation, Task Planning, Motion and Path Planning
Abstract: POMDPs offer a standard approach to model uncertainty in many robot tasks. Traditionally, POMDPs are formulated with optimality objectives. Here we study a different POMDP formulation with Boolean objectives. For robotic domains that require a correctness guarantee, Boolean objectives are natural formulations. We investigate POMDPs with a common Boolean objective: safe-reachability, requiring that the robot eventually reaches a goal state with a probability above a threshold while keeping the probability of visiting unsafe states below a different threshold. Our approach builds upon the previous work that represents POMDPs using symbolic constraints. We employ a satisfiability modulo theories (SMTs) solver to efficiently search for solutions, i.e., policies or conditional plans that specify the action to take. A full policy or conditional plan is generally expensive to compute. To improve efficiency, we introduce partial conditional plans that cover sampled events to approximate full conditional plans. Our approach constructs a partial conditional plan parameterized by a replanning probability. We prove that the failure rate of the partial conditional plan is bounded by the replanning probability. Our approach allows users to specify an appropriate bound on the replanning probability to balance efficiency and correctness. Moreover, we update this bound properly to quickly detect whether the current partial conditional plan meets the bound and avoid unnecessary computation. To further improve the efficiency, we cache partial conditional plans for sampled belief states and reuse these cached plans if possible. We validate our approach in several robotic domains. The results show that our approach outperforms a previous policy synthesis approach in these domains.
|
|
TuBT7 Special Session, St Clair 2 |
|
Robotic Control and Robotization of Tasks within Industry 4.0 |
Chair: Noureddine, Farid | Ecole Nationale d'Ingénieurs De Tarbes - F |
Co-Chair: Eymüller, Christian | University of Augsburg |
Organizer: Noureddine, Farid | Ecole Nationale d'Ingénieurs De Tarbes - F |
Organizer: Raharijaona, Thibaut | ENIM University of Lorraine |
Organizer: Camarillo-Gómez, Karla Anhel | Instituto Tecnológico De Celaya |
Organizer: Byiringiro, Jean Bosco | Dedan Kimathi University of Technology-DeKUT |
Organizer: Rakotondrabe, Micky | Laboratoire Génie De Production (LGP) |
|
15:30-15:50, Paper TuBT7.1 | |
>Intelligent Robotic System for Solving Dissection Puzzle Combining K-Nearest Neighbors, Decision Tree and Deep Q Network (I) |
|
Müller, Rainer | ZeMA GGmbh |
Kanso, Ali | ZeMA GGmbH |
Xu, Xiaomei | University Saarland, ZeMA - Center for Mechatronics and Automati |
Keywords: Deep Learning in Robotics and Automation, Machine learning, Reinforcement
Abstract: The autonomous learning of an intelligent robotic system has enormous potential to solve assembly tasks. Deep reinforcement learning algorithms generally apply to robotics and their application level is still limited to relatively simple tasks due to the complexity and variability of the industrial environment. This paper presents an intelligent robotic system for composing a tangram puzzle with a set of polygons through an industrial robot, a suction device and a camera. A camera system is used to detect and estimate the 3D position and orientation of the polygons. An intelligent robotic system combines K-nearest neighbors, decision tree and deep Q network to learn different target photo puzzles and aims at assembling in industry.
|
|
15:50-16:10, Paper TuBT7.2 | |
>RealCaPP: Real-Time Capable Plug & Produce Communication Platform with OPC UA Over TSN for Distributed Industrial Robot Control (I) |
|
Eymüller, Christian | University of Augsburg |
Hanke, Julian | University of Augsburg |
Hoffmann, Alwin | University of Augsburg |
Reif, Wolfgang | University of Augsburg |
Kugelmann, Markus | KUKA GmbH Augsburg |
Grätz, Florian | KUKA GmbH Augsburg |
Keywords: Cyber-physical Production Systems and Industry 4.0, Intelligent and Flexible Manufacturing, Robot Networks
Abstract: The industry of tomorrow is changing from central hierarchical industrial and robot controls to distributed controls on the industrial shop floor. These fundamental changes in network structure make it possible to implement technologies such as Plug & Produce. In other words, to integrate, change and remove devices without much effort at runtime. In order to achieve this goal, a uniform architecture with defined interfaces is necessary to establish real-time communication between the varying devices. Therefore, we propose an approach to use the combination of OPC UA and TSN to automatically configure real-time capable communication paths between robots and other cyber-physical components and execute real-time critical tasks in the distributed control system.
|
|
16:10-16:30, Paper TuBT7.3 | |
>A Literature Review on Control Techniques for Collaborative Robotics in Industrial Applications (I) |
|
Proia, Silvia | Politecnico Di Bari |
Carli, Raffaele | Politecnico Di Bari |
Cavone, Graziana | Polytechnic of Bari |
Dotoli, Mariagrazia | Politecnico Di Bari |
Keywords: Collaborative Robots in Manufacturing, Human-Centered Automation, Control Architectures and Programming
Abstract: One of the key enabling technologies that has made Industry 4.0 a concrete reality is without doubt collaborative robotics, which is also evolving as a fundamental pillar of the next revolution, the so-called Industry 5.0. The improvement of safety and employees' well-being, together with the increment of profitability and productivity, are indeed the main goals of human-robot collaboration (HRC) in the industrial setting. The robotic controller design and the analysis of existing decision and control techniques are crucially needed to develop innovative models and state-of-the-art methodologies for a safe, ergonomic, and efficient HRC. To this aim, this article presents an accurate review of the most recent and relevant papers in the related field, focusing on the control perspective. All the surveyed works are carefully selected and categorized by target (i.e., safety, ergonomics, and efficiency), and then by problem and type of control, in presence or absence of optimization. Finally, the discussion of the achieved results and the analysis of the emerging challenges in this research field are reported, highlighting the identified gaps and promising future developments in the context of the digital evolution.
|
|
16:30-16:50, Paper TuBT7.4 | |
>Overcoming Uncertainties and Disturbances: An Adaptive Control Approach for Mobile Robots (I) |
|
Naveed, Kanwal | NUST |
Zaman, Uzair Khaleeq uz | NUST |
Kumar, Atal Anil | University of Luxembourg |
Keywords: Robust/Adaptive Control, Motion Control, Motion and Path Planning
Abstract: An adaptive control technique is introduced in this paper in order to efficiently overcome the parametric uncertainties and external disturbances which might prove to be catastrophic in the tracking control task of Mobile Robots (MR). The adaptive control technique is implied is independent of the underlying model of the robot thus making it easier to implement. Moreover, the asymptotic stability is guaranteed by making use of Lyapunov approach. The supremacy of adaptive control is also established through simulations by comparing it with a state-dependent Riccati equation (SDRE) based robust controller.
|
|
16:50-17:10, Paper TuBT7.5 | |
>Visual Scene Understanding for Enabling Situation-Aware Cobots (I) |
|
Eisenbach, Markus | Ilmenau University of Technology |
Aganian, Dustin | University of Technology Ilmenau |
Köhler, Mona | Ilmenau University of Technology |
Stephan, Benedict | Ilmenau University of Technology |
Schroeter, Christof | MetraLabs GmbH |
Gross, Horst-Michael | Ilmenau University of Technology |
Keywords: Collaborative Robots in Manufacturing, Deep Learning in Robotics and Automation, Intelligent and Flexible Manufacturing
Abstract: Although in the course of Industry 4.0, a high degree of automation is the objective, not every process can be fully automated - especially in versatile manufacturing. In these applications, collaborative robots (cobots) as helpers are a promising direction. We analyze the collaborative assembly scenario and conclude that visual scene understanding is a prerequisite to enable autonomous decisions by cobots. We identify the open challenges in these visual recognition tasks and propose promising new ideas on how to overcome them.
|
|
17:10-17:30, Paper TuBT7.6 | |
>Improvements in Robotic Manufacturing Systems by Using a Force Control (I) |
|
Noureddine, Farid | Ecole Nationale d'Ingénieurs De Tarbes - F |
Keywords: Factory Automation, Intelligent and Flexible Manufacturing
Abstract: Abstract--- Nowadays, competitive market comes under a high pressure to produce more and more quickly and cost-effectively. For this purpose, improvements in robotic manufacturing systems play an important part. Two main streams are developped at the Industrial Robotics Platform in our engineering school in Tarbes. The first deals with the use of force control in machining, especially for grinding, deburring and polishing. During parts machining and particularly during finishing operations, surface roughness is highly important and imposed by the design drawing. To improve surface quality, the use of a sensing force with its control is highly suitable. Some results, using a force control, are shown in experimental work. The second aims at reducing the robot programming time as much as possible to have a benefit from the robot investment. It is then appropriate to bring more flexibility to the path generation, on one hand, to automatically generate the tool path, on the other hand to compensate all possible errors occurring on the workcell during the deburring. These errors can be due to uncertainties relative to the raw part, to the fixture or more generally due to the robot itself. To overcome the challenge faced by the automatic trajectory generation, measurement systems must be used. Both options are available, binocular stereo vision or laser scanner. These 2 options are finally discussed according to obtained experimental results.
|
|
TuBT9 Regular Session, St Clair 3A |
|
Planning, Scheduling and Coordination 1 |
Chair: Dutta, Ayan | University of North Florida |
Co-Chair: Liu, Pengcheng | University of York |
|
15:30-15:50, Paper TuBT9.1 | |
>A Harmony Search Based Algorithm for a Stochastic Two-Sided Assembly Line Balancing Problem |
|
Wu, Jiaxi | Peking University |
Jiang, Wei | Peking University |
Shi, Leyuan | Univ. of Wisconsin-Madison |
Keywords: Planning, Scheduling and Coordination, Assembly, Task Planning
Abstract: We address a two-sided assembly line balancing problem with the consideration of stochastic task time and multi-product production mode. The objective is to minimize the expected cycle time. In view of the difficulty in evaluating the objective in a stochastic environment, we model the target problem as a simulation optimization problem. A Harmony Search (HS) based algorithm is proposed to solve the problem. Besides, an Optimal Computing Budget Allocation (OCBA) method is adopted to enhance the simulation efficiency. Numerical experiments are conducted to evaluate the performance of the proposed HS-based algorithm against other well-known algorithms in the literature. The results indicate that the proposed algorithm outperforms the benchmark algorithms both in quality and computing efficiency.
|
|
15:50-16:10, Paper TuBT9.2 | |
>Time-Efficient Mars Exploration of Simultaneous Coverage and Charging with Multiple Drones |
> Video
|
|
Chang, Yuan | National University of Defense Technology |
Yan, Chao | National University of Defense Technology |
Liu, Xingyu | National University of Defense Technology |
Wang, Xiangke | National University of Defense Technology |
Zhou, Han | National University of Defense Technology |
Xiang, Xiaojia | National University of Defense Technology |
Tang, Dengqing | NUDT |
Keywords: Planning, Scheduling and Coordination, Autonomous Agents, Reinforcement
Abstract: This paper presents a time-efficient scheme for Mars exploration by the cooperation of multiple drones and a rover. To maximize effective coverage of the Mars surface in the long run, a comprehensive framework has been developed with joint consideration for limited energy, sensor model, communication radius and safety radius, which we call TIME-SC 2 (TIme-efficient Mars Exploration of Simultaneous Coverage and Charging). First, we propose a multi-drone coverage control algorithm by leveraging emerging deep reinforcement learning and design a novel information map to represent dynamic system states. Second, we propose a near-optimal charging scheduling algorithm to navigate each drone to an individual charging slot, and we have proven that there always exists feasible solutions. The attractiveness of this framework not only resides on its ability to maximize exploration efficiency, but also on its high autonomy that has greatly reduced the non-exploring time. Extensive simulations have been conducted to demonstrate the remarkable performance of TIME-SC 2 in terms of time-efficiency, adaptivity and flexibility.Video is available at: https://github.com/changmsdn/Coverage-control
|
|
16:10-16:30, Paper TuBT9.3 | |
>Min-Max Task Assignment and Sequencing with Heterogeneous Unmanned Vehicles |
|
Battistini, Jarrett | Texas a & M |
Rathinam, Sivakumar | TAMU |
Tatum, Richard | NSWCPCD |
Keywords: Planning, Scheduling and Coordination, Optimization and Optimal Control, Motion and Path Planning
Abstract: This article addresses a planning problem for a team of heterogeneous, unmanned surface vehicles whose time costs are attributable to either transiting or task execution costs. Given a set of target regions and a team of unmanned vehicles, the Heterogeneous Multi-vehicle Planning Problem (HMPP) aims to find a tour for each vehicle such that each target is visited at least once by some vehicle and the maximum mission cost of any unmanned vehicle is minimized. The mission cost incurred by each unmanned vehicle in this work includes its travel costs as well as the costs involved in performing the tasks in the regions visited by the vehicle. This problem is a generalization of the single vehicle Traveling Salesman Problem and is NP-Hard. We develop a fast approximation algorithm that provides a feasible solution with a bound on the cost of solution found and improve on it further through variable neighborhood search heuristics. We also present numerical results to corroborate the performance of the proposed approaches.
|
|
16:30-16:50, Paper TuBT9.4 | |
>Manufacturing Line Design Configuration with Optimized Resource Groups |
|
Nakano, Takahiro | HITACHI |
Kajita, Daiki | Hitachi |
Chen, Heming | Hitachi America Ltd |
Kovalenko, Ilya | University of Michigan |
Balta, Efe | University of Michigan |
Qamsane, Yassine | University of Michigan |
Barton, Kira | University of Michigan at Ann Arbor |
Keywords: Planning, Scheduling and Coordination, Task Planning, Factory Automation
Abstract: This research aims to develop methods to quickly build new manufacturing lines in response to changes in product varieties and manufacturing fluctuations in a factory. We propose a meta-heuristic algorithm for solving large-scale optimizations of the line design process, which includes resource configuration, process design, control design, and line configuration. The proposed framework improves the automation and system-level interactions of the line design process as compared to conventional methods that manually solve each step in the process design problem individually using skilled line engineers with previous experience. This research introduces the concept of a resource group or module that consists of various manufacturing resources such as robots, tools, autonomous guided vehicles, and conveyors. The line design process is then reconfigured for module or group configuration. To demonstrate the proposed framework, a case study is conducted in which the proposed framework is applied to the line design of an assembly manufacturing facility with production costs and manufacturing lead times selected as the key performance indicators of interest. Results indicate improved line costs and manufacturing lead times concurrently.
|
|
16:50-17:10, Paper TuBT9.5 | |
>Robot Motion Planning Benchmarking and Optimization through Motion Planning Pipeline |
|
Liu, Shuai | University of York |
Liu, Pengcheng | University of York |
Keywords: Planning, Scheduling and Coordination, Optimization and Optimal Control, Industrial and Service Robotics
Abstract: Motion planning algorithms have been designed with adaptability to different problems. However, how to choose a suitable planner for a scene has always been a question worth exploring. This paper aims to find a suitable motion planner under two different scenes and three different queries. The work lies in optimization of sampling-based motion planning methods through Motion Planning Pipeline and Planning Request Adapter. The idea is to use the pre-processing of the planning request adapter, to run OMPL as a pre-processer for the optimized CHOMP or STOMP algorithm, and connect through the motion planning pipeline, to realize the optimization of the motion trajectory. The optimized trajectories are compared with original trajectories through benchmarking, which determines the most suitable motion planning algorithm for different scenarios and different queries. Experimental results show that after optimization, although the planning time of the algorithm is longer, the generated path quality is significantly improved. In the low-complexity scenes, STOMP optimizes the sampling algorithm very well, improves the trajectory quality greatly, and has a higher success rate. CHOMP also has a good optimization of the sampling algorithm, but it reduces the success rate of the original algorithm. However, in more complex scenes, the performance of the two optimization methods may not be as good as the original algorithm.
|
|
17:10-17:30, Paper TuBT9.6 | |
>Distributed Hedonic Coalition Formation for Multi-Robot Task Allocation |
|
Dutta, Ayan | University of North Florida |
Ufimtsev, Vladimir | East Central University |
Said, Tuffa | University of North Florida |
Jang, Inmo | Samsung Electronics |
Eggen, Roger | University of North Florida |
Keywords: Planning, Scheduling and Coordination, Task Planning, Robot Networks
Abstract: In this paper, we study the problem of allocating multiple heterogeneous robots to tasks. Due to the limited capabilities of the robots, a task might need more than one robot to complete it. The fundamental problem of optimally partitioning the set of n robots into m disjoint coalitions for allocating to m tasks is proved to be NP-hard. To solve this computationally intractable problem, we propose a distributed hedonic game formulation, where each robot decides to join or not join a team based on the other robots allocated to that particular task. It uses a bipartite matching technique to establish an initial set of coalitions before letting the robots coordinate asynchronously and change teams if they want. Our proposed solution is proved to converge to a Nash-stable solution. We have tested the proposed solution in simulation with up to 50 robots and 15 tasks within a Message Passing Interface (MPI) framework. Results show that our proposed approach is fast and handles asynchronous robot-to-robot communication while earning more utility (up to 23%) than an existing technique in the majority of the test cases.
|
|
TuBT10 Regular Session, St Clair 3B |
|
Intelligent and Flexible Manufacturing 2 |
Chair: Li, Jingshan | University of Wisconsin - Madison |
Co-Chair: Frigerio, Nicla | Politecnico Di Milano |
|
15:30-15:50, Paper TuBT10.1 | |
>A Novel Architecture for Cyber-Physical Production Systems in Industry 4.0 |
|
Huertos Izquierdo, Francisco Javier | LORTEK |
Chicote, Beatriz | LORTEK Technological Centre, Basque Research and Technology Alli |
Masenlle, Manuel | Lortek S.Coop |
Ayuso, Mikel | Lortek |
Keywords: Cyber-physical Production Systems and Industry 4.0, Intelligent and Flexible Manufacturing, Cognitive Automation
Abstract: The Hyperconnected Architecture for High Cognitive Production Plants (HyperCOG) project aims at the process industry's complete digital transformation through an advanced Industrial Cyber-Physical Infrastructure. It is based on advanced technologies that allow a hyperconnected network of digital nodes to be created improving the classic automation hierarchy of communication layers. The nodes will collect data streams in real-time, offering cognitive sensing and information along with high performance computing capabilities making the process industry businesses solid in different scenarios. The system is validated in three fields of the process industry: steel, cement and chemical where optimization in the use of energy and raw materials is obtained, among other benefits.
|
|
15:50-16:10, Paper TuBT10.2 | |
>GAN Based Data Analysis and Mining for Smart Shop Floor Scheduling (I) |
|
Ma, Yumin | Tongji University |
Li, Shengyi | Tongji University |
Lu, Xiaoyu | Tongji University |
Liu, Juan | Tongji University |
Keywords: Big-Data and Data Mining, Intelligent and Flexible Manufacturing, Semiconductor Manufacturing
Abstract: Mining knowledge from data is an important way to solve scheduling problems. Adequate samples are a prerequisite for mining effective scheduling knowledge. However, it is difficult and time-consuming to obtain high-quality samples from the production process. To solve this problem, we propose a data mining method for smart shop floor scheduling based on GAN (Generative Adversarial Network). First, GAN is used to learn the distribution of initial samples and generate enough simulation samples to meet the data requirements for scheduling knowledge mining. Then, SVR (Support Vector Regression) is applied to perform scheduling knowledge mining, that is to establish the mapping relationship between shop floor production state and optimal scheduling strategy. The proposed method has been validated on the production system model MiniFab. Compared with the traditional sample acquisition method, the proposed method can effectively shorten the sample acquisition time and ensure the validity of the mined scheduling knowledge.
|
|
16:10-16:30, Paper TuBT10.3 | |
>Bayesian Ontology Network for Diagnostic Inference in Additive Manufacturing |
|
Chen, Ruimin | Pennsylvania State University |
Lu, Yan | National Institute of Technology and Standards |
Witherell, Paul | NIST |
Simpson, Tim | Pennsylvania State University |
Kumara, Soundar | The Pennsylvania State University |
Yang, Hui | The Pennsylvania State University |
Keywords: Intelligent and Flexible Manufacturing, Computer Vision for Manufacturing, Product Design, Development and Prototyping
Abstract: Additive manufacturing (AM) enables the creation of complex geometries that are difficult to realize using conventional manufacturing techniques. Advanced sensing is increasingly being used to improve AM processes, and installing different sensors onto AM systems has yielded more data-rich environments. Transforming data into useful information and knowledge (i.e., causality detection and process-structure-property (PSP) relationship identification) is important for achieving the necessary quality assurance and quality control (QA/QC) in AM. However, causality modeling and PSP relationship establishment in AM are still in early stages of development. In this paper, we develop an ontology-based Bayesian network (BN) model to represent causal relationships between AM parameters (i.e., design parameters and process parameters) and QA/QC requirements (e.g., structure properties and mechanical properties). The proposed model enables engineering interpretations and can further advance AM process monitoring and control.
|
|
16:30-16:50, Paper TuBT10.4 | |
>A Novel Implementation Architecture of Intelligent Predictive Maintenance with Continuous Integration and Delivery Based on Container Technology |
|
Hung, Min-Hsiung | Chinese Culture University |
Lin, Yu-Chuan | National Cheng Kung University |
Lin, You-Cheng | Institute of Manufacturing Information and Systems, National Che |
Shieh, Zih-Yan | National Cheng Kung University |
Suryajaya, Benny | National Cheng Kung University |
Chen, Chao-Chun | National Cheng Kung University |
Cheng, Fan-Tien | National Cheng Kung University |
Keywords: Intelligent and Flexible Manufacturing, Cyber-physical Production Systems and Industry 4.0, Sustainable Production and Service Automation
Abstract: The new-generation virtualization technology-Docker Container has many advantages over the traditional virtual machine: The booting time of a container is only a few seconds. The performance of a containerized application is close to the native one. A single host could run thousands of isolated containers simultaneously. Docker containers are less resource-intensive than virtual machines and are thus more suitable to package and ship software. Docker containers allow the software to be built, shipped, and run anywhere in a standard way. Also, the container orchestrator Kubernetes can run and manage containerized applications in an automatic manner and provide them with robustly operational functionalities, such as load balance, health check, failover, and resource limit. Thus, leveraging container technology to tackle the issues of the software supply chain is becoming a trend for software development and operations (DevOps). In this paper, we propose a novel implementation architecture of intelligent predictive maintenance (IPM) [3] with continuous integration and delivery (CI/CD) based on container technology for intelligent manufacturing. Specifically, Docker is utilized to package and run the containerized IPM application (i.e., IPMC) that is composed of several microservices, and Kubernetes is employed to manage and orchestrate the IPMC. Besides, Drone and Helm are adopted to design a CI/CD mechanism for efficient deployment and updates of the IPMC. Finally, testing results of an industrial case study of applying the proposed IPMC with CI/CD for the TFT-LCD manufacturing industry are demonstrated to validate the efficiency and efficacy of the proposed methods.
|
|
16:50-17:10, Paper TuBT10.5 | |
>Machine Learning in Manufacturing Ergonomics: Recent Advances, Challenges, and Opportunities |
|
Lee, Sujee | University of Wisconsin-Madison |
Liu, Li | Northeast University |
Radwin, Robert | University of Wisconsin-Madison |
Li, Jingshan | University of Wisconsin - Madison |
Keywords: Logistics, Intelligent and Flexible Manufacturing, Human Factors and Human-in-the-Loop
Abstract: The rapid development of machine learning (ML) technology has introduced substantial impact on ergonomics research in manufacturing. Numerous studies and practices have been carried out to apply ML techniques to address manufacturing ergonomics issues, which has brought extensive opportunities as well as significant challenges. To incentivize future research in this area, this paper reviews the recent advances of ML applications in manufacturing ergonomics, and discusses future research opportunities and challenges from ML, ergonomics, and manufacturing systems perspectives.
|
|
17:10-17:30, Paper TuBT10.6 | |
>An Online Policy for Energy-Efficient State Control of Manufacturing Equipment |
|
Frigerio, Nicla | Politecnico Di Milano |
Marzano, Lorenzo | Politecnico Di Milano |
Matta, Andrea | Politecnico Di Milano |
Keywords: Intelligent and Flexible Manufacturing, Factory Automation, Model Learning for Control
Abstract: Machine state control is one of the most promising energy-efficient measures for machining processes. A proper control reduces the energy consumed during idle periods by switching off/on the machines. A critical barrier for practical implementation is related to the knowledge of part arrival process that is affected by uncertainty. The stochastic processes involved in the system are usually assumed to be known. However, real production environments are subject to several sources of randomness that are difficult to model a priori. This work provides an online time-based algorithm that is able to control the machine state. Through a method for the estimation of the stochastic process, the algorithm provides the optimal control parameters based on a collected set of observations. A new policy is formulated to manage the control over time such that changes in the control parameters are applied only under certain conditions. Potential benefits are discussed using realistic numerical cases.
|
|
TuBT11 Regular Session, St Clair 4 |
|
AI-Based Methods 2 |
Chair: Hsu, Chia-Yu | National Taipei University of Technology |
Co-Chair: Julius, Agung | Rensselaer Polytechnic Institute |
|
15:30-15:50, Paper TuBT11.1 | |
>Learning to Discover Task-Relevant Features for Interpretable Reinforcement Learning |
|
Zhang, Qiyuan | Harbin Institute of Technology |
Ma, Xiaoteng | Tsinghua University |
Yang, Yiqin | Tsinghua University |
Li, Chenghao | Tsinghua University |
Yang, Jun | Tsinghua University |
Liu, Yu | Harbin Institute of Technology |
Liang, Bin | Tsinghua University |
Keywords: Reinforcement Learning, AI-Based Methods, Deep Learning Methods
Abstract: Reinforcement Learning (RL) agents are often fed with large-dimensional observations to achieve the ideal performance in complex environments. Unfortunately, the massive observation space usually contains useless or even adverse features, which leads to low sample efficiency. Existing methods rely on domain knowledge and cross-validation to discover efficient features which are informative for decision-making. To minimize the impact of prior knowledge, we propose a temporal-adaptive feature attention algorithm (TAFA). We adopt a non-linear attention module, automatically choosing task-relevant components of hand-crafted state features without any domain knowledge. Our experiments on MuJoCo and TORCS tasks show that the agent achieves competitive performance with state-of-the-art methods while successfully identifying the most task-relevant features for free. We believe our work takes a step towards the interpretability of RL.
|
|
15:50-16:10, Paper TuBT11.2 | |
>Utilization of Semantic Planes: Improved Localization and Dense Semantic Map for Monocular SLAM in Urban Environment |
|
Bao, Yaoqi | Zhejiang University |
Pan, Yun | Zhejiang University |
Yang, Zhe | Zhejiang University |
Huan, Ruohong | Zhejiang University of Technology |
Keywords: AI-Based Methods, Computer Vision for Transportation, Intelligent Transportation Systems
Abstract: In this paper, we propose a novel semantic direct monocular simultaneous localization and mapping (SLAM) system that fuses the semantic information obtained by an advanced deep neural network (DNN) into direct sparse odometry with loop closure(LDSO), with the purpose of improving the localization accuracy and building a dense semantic map of the urban environment. For localization, we apply a point reselection strategy based on coarse semantic plane (CSP) constraints to discard static points inconsistent with the nearby co-plane points of the same semantic class and dynamic points beyond the visible range. Moreover, a point group movement consistency (PGMC) check is utilized to decrease the impact of moving dynamic objects. For the dense semantic map, we model numerous small semantic planes from well-estimated points to measure the depth of each static pixel, rather than conduct stereo matching. Experimental results show that our method is more accurate than LDSO and comparable with ORB-SLAM in terms of localization. Moreover, it is capable of building a dense semantic map of the urban environment for better scene understanding.
|
|
16:10-16:30, Paper TuBT11.3 | |
>A Neuro-Symbolic Architecture for Signal Temporal Logic Inference |
|
Julius, Agung | Rensselaer Polytechnic Institute |
Yan, Ruixuan | Rensselaer Polytechnic Institute |
Keywords: AI-Based Methods, Big-Data and Data Mining
Abstract: In this talk, we present a framework that combines the characteristics of neural networks and temporal logics. The framework is called Weighted Signal Temporal Logic Neural Network (wSTL-NN). Our framework generalizes the existing framework of Signal Temporal Logic (STL) by introducing the use of weights in logical Boolean operations (AND and OR) and extending it to the temporal operators ALWAYS and EVENTUALLY. Further, the process of computing the satisfaction of a wSTL formula by a signal is cast in a neural network-like architecture, where each neuron corresponds to a wSTL subformula, and its output corresponds to the quantitative satisfaction of the formula. The main idea of this talk is to use wSTL formulas as classifier for time series. We infer (or, in the terminology of neural networks, we train) the wSTL formula, which is represented by a wSTL-NN in a supervised learning fashion. The wSTL-NN is end-to-end differentiable, which allows inference of wSTL formulas to be done using backpropagation.
|
|
16:30-16:50, Paper TuBT11.4 | |
>Towards Safe Control of Continuum Manipulator Using Shielded Multiagent Reinforcement Learning |
> Video
|
|
Yan, Junyan | The Chinese University of Hong Kong |
Ji, Guanglin | The Chinese University of HongKong |
Du, Jingxin | The Chinese Univerisity of Hong Kong |
Yan, Wanquan | The Chinese University of HongKong |
Chen, Jibiao | Shenyang Institute of Automation, Chinese Academy of Sciences |
Lu, Yongkang | The Chinese University of Hong Kong |
Rojas, Juan | Chinese University of Hong Kong |
Cheng, Shing Shin | The Chinese University of Hong Kong |
Keywords: Reinforcement Learning
Abstract: Continuum robotic manipulators are increasingly adopted in minimal invasive surgery. However, their nonlinear behavior is challenging to model accurately, especially when subject to external interaction, potentially leading to poor control performance. In this letter, we investigate the feasibility of adopting a model-free multiagent reinforcement learning (RL), namely multiagent deep Q network (MADQN), to control a 2-degree of freedom (DoF) cable-driven continuum surgical manipulator. The control of the robot is formulated as a one-DoF, one agent problem in the MADQN framework to improve the learning effi- ciency. Combined with a shielding scheme that enables dynamic variation of the action set boundary, MADQN leads to efficient and importantly safer control of the robot. Shielded MADQN enabled the robot to perform point and trajectory tracking with submillimeter root mean square errors under external loads, soft obstacles, and rigid collision, which are common interaction scenarios encountered by surgical manipulators. The controller was further proven to be effective in a miniature continuum robot with high structural nonlinearitiy, achieving trajectory tracking with submillimeter accuracy under external payload.
|
|
16:50-17:10, Paper TuBT11.5 | |
>Semi-Supervised Learning Framework for Defect Classification and Empirical Study in Printed Circuit Board Manufacturing |
|
Hsu, Chia-Yu | National Taipei University of Technology |
Lu, Yi-Wei | National Taipei University of Technology |
Keywords: AI-Based Methods, Computer Vision for Manufacturing, Deep Learning Methods
Abstract: Supervised deep convolution neural network has achieved great results on image classification task. Supervised model generates a function which can mapping input to output through algorithm. This function learns the relationship between labeled data and the output classes, mapping the vector of predict target to one of the classes to make the inference. It requires a large amount of labeled data to perform multiple iterative to update neural weights and optimize model. The model cannot exert its maximum performance with a lack of labeled data. Acquiring label information often consumes a lot of time and labor cost. Therefore, the labeling problem is one of the key points to be solved in deep learning pragmatic. In the research, we combine neural network and semi-supervised learning framework to accelerate the acquisition of labeled data by pseudo labeling method. In the framework, for unlabeled data, we use a model which has been trained with labeled data, to inference the unlabeled data. Then, take the class with the highest prediction probability as the pseudo label of unlabeled data. The pseudo label will be regarded as the true label of these unlabeled data, and add the data to the next round of model training, and keep looping. The pseudo labeling mechanism can not only help reduce the time and labor cost required of label data, but also achieve the purpose of using unlabeled data to improve the generalization performance of the model. Finally, this research uses real world case of printed circuit board (PCB) defect inspection to verify this architecture, and the results prove that the training results of the model can be improved while reducing labor and time costs.
|