Towards Safety Assessment of Robot Behaviors in SMACH
Eun-Young Kang, Miguel Campusano · 2023
Due to the critical consequences of possible failures, robot systems must be formally verified to guarantee that their behaviors are correct and safe. There is, however, a gap in terms of building safe behaviors between the formal methods and robotic communities as the latter focuses on informal design and its implementation in a manner which is accessible to robotics engineers. In this paper, we present an approach to bridge that gap which enables a tight coupling of informal robot behaviors defined in SMACH, Python state machine API, with formal models through a process of translation. A set of mapping rules, which facilitates transformation is provided and the result is utilized for formal verification of safety properties. We also discuss the current limitations of such work along with recommendations on how these might be addressed.