Towards an Automatic Health Monitor for Autonomous Underwater Vehicles using Parameter Identification

In the last few years, interest has grown in the use of autonomous underwater vehicles for commercial, scientific and military missions. Reliability is critical and autonomous fault detection with programmed recovery procedures have to be built into their control logic. It is important that the mission controller have information concerning the current status of the maneuverability subsystems of the vehicle to perform requested motions. The normal techniques of servo error monitors, limit and trend checks, and Kalman filter state estimators with innovations checks go a long way to providing sensor fault detection. However, the inherent capability of a vehicle to determine the state of health of its steering, diving, and speed subsystems (including fin jams) is not easily discovered by these methods. This paper discusses the use of both batch least squares and Kalman Filters for system parameter identification as a means to detect a change in performance. Applied to the experimental maneuvering responses of the NPS AUV II autonomous underwater vehicle we wish to determine the range of varability of key steering system response parameters that would form the basis of a health monitor. In this application we are not seeking parameter values for the purpose of adaptive control. Instead, we wish to determine if key parameters such as input gain have changed or are out of range.