Evaluating Reinforcement Learning Safety and Trustworthiness in Cyber-Physical Systems
This addresses safety and trustworthiness issues for RL-based CPS, but it appears incremental as it builds on existing design science approaches without claiming major breakthroughs.
The paper tackles the challenge of constructing safety cases for reinforcement learning (RL) components in cyber-physical systems (CPS) by proposing the SAFE-RL framework, which is demonstrated in three RL applications for small uncrewed aerial systems (sUAS).
Cyber-Physical Systems (CPS) often leverage Reinforcement Learning (RL) techniques to adapt dynamically to changing environments and optimize performance. However, it is challenging to construct safety cases for RL components. We therefore propose the SAFE-RL (Safety and Accountability Framework for Evaluating Reinforcement Learning) for supporting the development, validation, and safe deployment of RL-based CPS. We adopt a design science approach to construct the framework and demonstrate its use in three RL applications in small Uncrewed Aerial systems (sUAS)