SS05 - Evaluation Methods for Autonomous Cyber-Physical Systems' Behavior

Special Session Organized by

Alexander Fay, Ruhr University Bochum, Germany, and Felix Gehlhoff, Helmut Schmidt University Hamburg, Germany, and Artan Markaj, Helmut Schmidt University Hamburg, Germany, and Mehmet Mercangöz, Imperial College London, United Kingdom, and André Scholz, Siemens AG, Germany,

Download Call for Papers

Click here to download the session cfp.


Global competition, shorter product life cycles, and highly volatile markets are putting a strain on operations in manufacturing, process, logistics, and energy systems. Many industries are facing a shortage of workers today, and given these trends, production of essential goods as well as the operation of such systems are going to be at risk due to missing workers and operators. To tackle these challenges, a higher degree of autonomy in cyber-physical systems (CPS) operations can be considered. Such autonomous CPS represent technical constructs which make decisions without the intervention of humans. In addition, autonomous CPS can be characterized by the ability to execute processes systematically, the ability to adapt to environmental changes, and the ability to self-govern resources. However, due to these characteristics, the behavior of such systems is not always easily explainable and predictable. Thus, methods for the evaluation of the behavior of such systems are of increasing importance. Evaluating these systems encompasses testing, verifying, diagnosing, and explaining (inter-)actions as well as analyzing performance. Suitable methods can be data-driven (incl. Generative AI) and knowledge-driven as well as simulation-based. Potential use cases can be, for example, a simulation-based selection of the most appropriate goals to follow by autonomous CPS or explaining actions of autonomous CPS to operators.

Topics under this session include (but not limited to)

  • Simulation-based evaluation of autonomous CPS’ behavior
  • Verification of system goals and requirements
  • Testing intelligent control algorithms (multi-agent systems, reinforcement learning, or model-predictive control) for autonomous CPS
  • Data-driven methods for autonomous CPS’ identification and adaption
  • Real-time diagnosis and decision-making in autonomous CPS’ operation
  • Integration of Explainable AI (XAI) for enhanced autonomous CPS’ transparency
  • Real-time decision support approaches for autonomous CPS’ supervision
  • Evaluation of human-autonomy interaction in remote and autonomous operation
  • (Real-time) Safety and regulatory compliance evaluation of autonomous CPS
  • Performance and test-beds for the evaluation of autonomous CPS
  • Evaluation of degrees of autonomy in autonomous CPS