Towards Robust Task Execution for Domestic Service Robots

In the field of domestic service robots, recovery from faults is crucial to promote user acceptance. In this context we will focus, in particular, on some specific faults which arise from interaction of the robot with its real world environment. In these situations even a well modelled robot may fail to perform its tasks successfully due to external faults which occur while interacting. We reason along the most frequent failures in typical scenarios which we have observed in real-world demonstrations and competitions using the autonomous service robot Jenny. We propose four different fault classes caused by disturbances, imperfect perception, inadequate planning or chaining of action sequences. The faults are first classified and then mapped to a small number of fault handling techniques partly known, partly extended by us. In addition to existing techniques we present two approaches to handle external faults from inadequate descriptions of the planner operator class. The first approach uses naive physics concepts to find information about detected external faults. The second approach is simulation based, utilising a single simulation that shows a manipulated object’s behaviour for successfully completing an action. The approach uses the N-Bins learning algorithm to suggest a releasing state of the object that avoids the occurrence of external faults. We apply the proposed approaches to the scenarios where a robot performs the pick-and-place manipulation tasks. The results of these applications show that both approaches hold great promises for handling external faults in domestic service robotics.