„DREAMing –Optimization and Learning of Robotic Tasks in Simulation“ supported by KUKA
In general, the application of ICT (information and communication technologies) methods to problems with high uncertainties or not fully known limits comes with two major challenges: First, the applied methods have to be able to cope with the problem’s openness. Second, the applied solution methods should show a suitable performance when operating in a standard situation. Self-optimizing or self-reflecting systems can handle both issues at the same time: during setup – and sometimes even during operation – data is collected, which can be used to interpolate between known solutions or extrapolate to yet unknown situations, the system might encounter. From an economic perspective, the Quality of Service and the Overall Equipment Efficiency will increase, while at the same time the Total Cost of Ownership for setup, programming, and parameter tuning will decrease.
The context of this workshop is the idea that the performance of robotic tasks can be improved in simulation based on the experience while executing the real task at hand. In particular, the RobDREAM project pursues the idea as follows: Sleep! For hominids and most other mammals, sleep means more than regeneration. Sleep positively affects working memory, which in turn improves higher-level cognitive functions such as decision making and reasoning. This is the inspiration of RobDREAM! What if robots could also improve their capabilities in their inactive phases – by processing experiences made during the working day and by exploring – or “dreaming” of – possible future situations and how to solve them best?
The special track lays its focus on learning and self-optimization with respect to flexible (mobile) robotics. It will introduce the concepts to the audience by invited talks from industry, academia, and the RobDREAM project. Subsequently, a public discussion about the key challenges will be held and the results will be made public on the RobDREAM website as a starting point for further discussion and common development.
Special Track Program
9:30 – Introduction: „Why should robots DREAM?“
9:45 – Invited Talks (I)
- The RobDREAM approach and use case (Dr. Daniel Braun, KUKA)
- Continuously optimize robotic tasks using a virtual model of the robot cell (Dr. Alwin Hoffmann, University of Augsburg)
10:30 Coffee Break (including presentation of RobDREAM testbed)
11:00 – Invited Talks (II)
- Novelty detection and active learning as basic mechanisms for robots’ dreams (Prof. Dr. Bernhard Sick, University of Kassel)
- Mobile and flexible robots – Challenges and Opportunities
11:30 Discussion on „Key Challenges and Issues in robotic DREAMing“
The track will take the invited talks to discuss some important topics in bringing together self-adapting and self-optimizing methods on the one hand and robot applications on the other hand. Three key questions will be addressed:
- Where and how can self-adapting/self-optimizing methods be beneficial to robotics applications?
Taking the RobDREAM use case of mobile robotics in an industrial setting as an introduction the possibilities for application of SASO methods and frameworks in for the individual tasks in this setting will be explored and discussed. The goal is to identify possible optimization points in robotic applications where self-adaption and self-optimization could be first applied.
- How can stability and predictability of newly created solutions be rated?
One challenge when applying self-adaptation and self-optimization to robotic systems in a real-world production scenario – or even a human-robot-interaction scenario – is to ensure stability and predictability of newly created solutions. With static programs or standard planning methods the stability and predictability of the robot behavior can at least be rated. This topic aims at bringing together the requirements in real world robotic applications and the prediction about the features of the outcome after applying self-adaption/self-optimization to the execution of robotic tasks.
- Which Interfaces and Frameworks are commonly used in Robotics and SASO?
After exploring use cases and application concerns in the prior two questions, the discussion session will be rounded up by a summary of existing (and commonly used) frameworks for setting up robotic tasks and for carrying out self-adaption/self-optimization operations. Furthermore the possibilities and challenges for bringing together both worlds will be discussed. The goal should be an identification of a common base and necessary extensions for the design, implementation and benchmarking of robotic applications with self-adaption and self-optimization features.
Individual expert statements and the outcome of the discussion will be published on the RobDREAM website for reference and further expanding the discussion on the topic.
- KUKA - Dr. Rainer Bischoff
- KUKA - Dr. Daniel Braun
- University of Augsburg - Dr. Alwin Hoffmann
Primary Contact Email:email@example.com
SASO homepage: https://saso2016.informatik.uni-augsburg.de/