Diagnosis of action execution failures for cognitive robots

Execution failures are likely in robotic applications due to dynamic and partially observable structure of the physical world. These failures should be detected by the robot, and a reasoning procedure should take place to diagnose the causes of the failures. In this paper, we propose a Hierarchical Hidden Markov Model (HHMM) based failure diagnosis method to identify the cause of a failure. Parallel HHMMs are used in the proposed method in order to track different type of failures. The performance of the proposed method is evaluated on our Pioneer 3-AT robot in several failure scenarios. The results reveal that using a probabilistic method ensures diagnosing multiple failures when there are more than one cause of a failure. Furthermore, using relations between the failure types and actions decreases memory requirements of the method by reducing the state space.