Synthesizing least-limiting guidelines for safety of semi-autonomous systems
This addresses safety challenges in semi-autonomous systems, such as human-machine collaboration, but appears incremental as it builds on existing game-theoretic methods.
The paper tackles the problem of synthesizing safe-by-design control strategies for semi-autonomous systems where safety requires collaboration from uncontrollable parts like human operators, proposing a systematic solution to generate least-limiting guidelines that minimize restrictions on the human operator in worst-case long-term executions.
We consider the problem of synthesizing safe-by-design control strategies for semi-autonomous systems. Our aim is to address situations when safety cannot be guaranteed solely by the autonomous, controllable part of the system and a certain level of collaboration is needed from the uncontrollable part, such as the human operator. In this paper, we propose a systematic solution to generating least-limiting guidelines, i.e. the guidelines that restrict the human operator as little as possible in the worst-case long-term system executions. The algorithm leverages ideas from 2-player turn-based games.