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).

Cyber-Physical Systems (CPS) increasingly leverage Reinforcement Learning (RL) to adapt dynamically to changing environments and optimize performance over time. While RL enhances efficiency and safety by enabling autonomous adjustments to unexpected conditions and hazard avoidance, it also introduces significant risks, as learned behaviors may lead to unpredictable or unsafe actions in real-world deployment. Therefore, integrating risk management into RL system design is essential. In this paper, we propose the QuestRL Framework, a question-driven approach that translates high-level safety guidelines into RL-specific considerations. This framework helps RL practitioners address key risks early in development, informing new or existing system requirements while ensuring traceability to risk management objectives. To evaluate its effectiveness, we conducted a study across two use cases, engaging six RL experts in developing system requirements with and without the framework. Our findings suggest that the framework promotes critical thinking and helps practitioners identify additional risk factors, ultimately supporting safer RL deployment.

Small Unmanned Aerial Systems (sUAS) must meet rigorous safety standards when deployed in high-stress emergency response scenarios; however many reported accidents have involved humans in the loop. In this paper, we, therefore, present the HiFuzz testing framework, which uses fuzz testing to identify system vulnerabilities associated with human interactions. HiFuzz includes three distinct levels that progress from a low-cost, limited-fidelity, large-scale, no-hazard environment, using fully simulated Proxy Human Agents, via an intermediate level, where proxy humans are replaced with real humans, to a high-stakes, high-cost, real-world environment. Through applying HiFuzz to an autonomous multi-sUAS system-under-test, we show that each test level serves a unique purpose in revealing vulnerabilities and making the system more robust with respect to human mistakes. While HiFuzz is designed for testing sUAS systems, we further discuss its potential for use in other Cyber-Physical Systems.

The Human Machine Teaming (HMT) paradigm focuses on supporting partnerships between humans and autonomous machines. HMT describes requirements for transparency, augmented cognition, and coordination that enable far richer partnerships than those found in typical human-on-the-loop and human-in-the-loop systems. Autonomous, self-adaptive systems in domains such as autonomous driving, robotics, and Cyber-Physical Systems, are often implemented using the MAPE-K feedback loop as the primary reference model. However, while MAPE-K enables fully autonomous behavior, it does not explicitly address the interactions that occur between humans and autonomous machines as intended by HMT. In this article, we, therefore, present the MAPE-K HMT framework, which utilizes runtime models to augment the monitoring, analysis, planning, and execution phases of the MAPE-K loop to support HMT despite the different operational cadences of humans and machines. We draw on examples from our own emergency response system of interactive, autonomous, small unmanned aerial systems to illustrate the application of MAPE-K HMT in both a simulated and physical environment, and we discuss how the various HMT models are connected and can be integrated into a MAPE-K solution.