Automatic Failure Recovery for End-User Programs on Service Mobile\n Robots

摘要

For service mobile robots to be most effective, it must be possible for\nnon-experts and even end-users to program them to do new tasks. Regardless of\nthe programming method (e.g., by demonstration or traditional programming),\nrobot task programs are challenging to write, because they rely on multiple\nactions to succeed, including human-robot interactions. Unfortunately,\ninteractions are prone to fail, because a human may perform the wrong action\n(e.g., if the robot's request is not clear). Moreover, when the robot cannot\ndirectly observe the human action, it may not detect the failure until several\nsteps after it occurs. Therefore, writing fault-tolerant robot tasks is beyond\nthe ability of non-experts.\n This paper presents a principled approach to detect and recover from a broad\nclass of failures that occur in end-user programs on service mobile robots. We\npresent a two-tiered Robot Task Programming Language (RTPL): 1) an expert\nroboticist uses a specification language to write a probabilistic model of the\nrobot's actions and interactions, and 2) a non-expert then writes an ordinary\nsequential program for a particular task. The RTPL runtime system executes the\ntask program sequentially, while using the probabilistic model to build a\nBayesian network that tracks possible, unobserved failures. If an error is\nobserved, RTPL uses Bayesian inference to find the likely root cause of the\nerror, and then attempts to re-execute a portion of the program for recovery.\n Our empirical results show that RTPL 1) allows complex tasks to be written\nconcisely, 2) correctly identifies the root cause of failure, and 3) allows\nmultiple tasks to recover from a variety of errors, without task-specific\nerror-recovery code.\n