Flexible Composition of Robot Logic with Computer Vision Services
Doctoral thesis, 2018
Vision-based robotics is an ever-growing field within industrial automation. Demands for greater flexibility and higher quality motivate manufacturing companies to adopt these technologies for such tasks as material handling, assembly, and inspection. In addition to the direct use in the manufacturing setting, robots combined with vision systems serve as highly flexible means for realization of prototyping test-beds in the R&D context.
Traditionally, the problem areas of robotics and computer vision are attacked separately. An exception is the study of vision-based servo control, the focus of which constitutes control-theoretic aspects of vision-based robot guidance under assumption that robot joints can be controlled directly. The missing part is a systemic approach to implementing robotic application with vision sensing given industrial robots constrained by their programming interface.
This thesis targets the development process of vision-based robotic systems in an event-driven environment. It focuses on design and composition of three functional components: (1) robot control function, (2) image acquisition function, and (3) image processing function. The thesis approaches its goal by a combination of laboratory results, a case study of an industrial company (Kongsberg Automotive AS), and formalization of computational abstractions and architectural solutions.
The image processing function is tackled with the application of reactive pipelines. The proposed system development method allows for smooth transition from early-stage vision algorithm prototyping to the integration phase. The image acquisition function in this thesis is exposed in a service-oriented manner with the help of a flexible set of concurrent computational primitives. To realize control of industrial robots, a distributed architecture is devised, which supports composability of communication-heavy robot logic, as well as flexible coupling of the robot control node with vision services.
Traditionally, the problem areas of robotics and computer vision are attacked separately. An exception is the study of vision-based servo control, the focus of which constitutes control-theoretic aspects of vision-based robot guidance under assumption that robot joints can be controlled directly. The missing part is a systemic approach to implementing robotic application with vision sensing given industrial robots constrained by their programming interface.
This thesis targets the development process of vision-based robotic systems in an event-driven environment. It focuses on design and composition of three functional components: (1) robot control function, (2) image acquisition function, and (3) image processing function. The thesis approaches its goal by a combination of laboratory results, a case study of an industrial company (Kongsberg Automotive AS), and formalization of computational abstractions and architectural solutions.
The image processing function is tackled with the application of reactive pipelines. The proposed system development method allows for smooth transition from early-stage vision algorithm prototyping to the integration phase. The image acquisition function in this thesis is exposed in a service-oriented manner with the help of a flexible set of concurrent computational primitives. To realize control of industrial robots, a distributed architecture is devised, which supports composability of communication-heavy robot logic, as well as flexible coupling of the robot control node with vision services.
robotics
data flow
service-oriented architecture
distributed systems
machine learning
discrete-event systems
Machine vision
publish-subscribe
Room ED, Hörsalsvägen 11, Campus Johanneberg
Opponent: Prof. Robert Harrison, The University of Warwick
Author
Oleksandr Semeniuta
Chalmers, Electrical Engineering, Systems and control
Oleksandr Semeniuta, Sebastian Dransfeld, Kristian Martinsen, and Petter Falkman. Towards increased intelligence and automatic improvement in industrial vision systems. Procedia CIRP, 67:256–261, 2018. ISSN 22128271. doi: 10.1016/j.procir.2017.12.209
Oleksandr Semeniuta, Sebastian Dransfeld, and Petter Falkman. Visionbased robotic system for picking and inspection of small automotive components. In 2016 IEEE International Conference on Automation Science and Engineering (CASE), pages 549–554. IEEE, aug 2016. ISBN 978-1-5090-2409- 4. doi: 10.1109/COASE.2016.7743452
Oleksandr Semeniuta and Petter Falkman. EPypes: a framework for building event-driven data processing pipelines. Submitted to: PeerJ Computer Science
Oleksandr Semeniuta and Petter Falkman. Flexible image acquisition service for distributed robotic systems. In 2018 Second IEEE International Conference on Robotic Computing (IRC), pages 106–112. IEEE, jan 2018. ISBN 978-1-5386-4652-6. doi: 10.1109/IRC.2018.00024
Oleksandr Semeniuta and Petter Falkman. Event-driven industrial robot control architecture for the Adept V+ platform. Submitted to: Frontiers in Robotics and AI
Areas of Advance
Information and Communication Technology
Production
Subject Categories
Communication Systems
Robotics
Computer Science
Computer Systems
ISBN
978-91-7597-777-5
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4458
Publisher
Chalmers
Room ED, Hörsalsvägen 11, Campus Johanneberg
Opponent: Prof. Robert Harrison, The University of Warwick