Ensuring Task-Independent Safety for Multi-Agent Systems by Feedback

This paper presents a new safety feedback design for multi-agent systems. In contrast to existing methods, the proposed approach follows the idea to decouple the design of low-level safety features from abstract high-level tasks of the agents. This leads to a modular design that preserves the flexibility of multi-agent systems while increasing their usability in safety critical applications by giving systems theoretic safety guarantees. The approach is illustrated using collision avoidance of two vehicles as an example.