ROSSMARie: A Domain-Specific Language To Express Dynamic Safety Rules and Recovery Strategies for Autonomous Robots

Ensuring functional safety is a critical challenge for autonomous robots, as they must operate reliably and predictably despite uncertainty. However, existing safety measures can over-constrain the system, limiting the robot’s availability to perform its assigned task. To address this problem, we propose a more flexible strategy that equips robots with the
ability to adapt to system failures and recover from those situations without human intervention. We extend a domain-specific language, Declarative Robot Safety (DeROS), whose runtime stops a robot whenever it violates a safety rule (e.g., proximity to a human). Our extended language, ROSSMARie, adds the capability to monitor whether a rule is no longer violated and to recover and resume robot operation. We validate ROSSMARie on the ROS-based industrial platform SkiROS2 and verify its effectiveness in achieving safety and availability. Our experiments demonstrate that our DSL extension ensures
functional safety while enabling robots to complete their tasks.