Detection of unanticipated faults for autonomous underwater vehicles using online topic models

For robots to succeed in complex missions, they must be reliable in the face of subsystem failures and environmental challenges. In this paper, we focus on autonomous underwater vehicle (AUV) autonomy as it pertains to self-perception and health monitoring, and we argue that automatic classification of state-sensor data represents an important enabling capability. We apply an online Bayesian nonparametric topic modeling technique to AUV sensor data in order to automatically characterize its performance patterns, then demonstrate how in combination with operator-supplied semantic labels these patterns can be used for fault detection and diagnosis by means of a nearest-neighbor classifier. The method is evaluated using data collected by the Monterey Bay Aquarium Research Institute's Tethys long-range AUV in three separate field deployments. Our results show that the proposed method is able to accurately identify and characterize patterns that correspond to various states of the AUV, and classify faults at a high rate of correct detection with a very low false detection rate.