Opportunistic Planning with Recovery for Robot Safety
Publikation aus Robotics
KI2017: Advances in Artificial Intelligence, Lecture Notes in Computer Science, vol 10505, Springer Int. Publishing, pp. 352-358 , 9/2017
Emerging applications for various types of robots require them to operate autonomously in the vicinity of humans. Ensuring the safety of humans is a permanent requirement that must be upheld at all times. As tasks performed by robots are becoming more complex, adaptive planning components are required. Modern planning approaches do not contribute to a robot system's safety capabilities, which are thus limited to means like reduction of force and speed or halt of the robot. We argue that during a plan's execution there should be a backup plan that the robot system can fall back to as an alternative to halting. We present our approach of generating variations of plans by optional exploitation of opportunities, which extend the initially safe plan for extra value when the current safety situation allows it, while maintaining recovery policies to get back to the safe plan in the case of safety-related events.
Keywords: Automated planning, Collaborative robot, Opportunistic planning, Robot safety