Fault diagnostic system design using LMIs, with applications to health monitoring for GPS-based lateral vehicle control

This paper develops an explicit design methodology for development of observer-based sensor health monitoring systems using Linear Matrix Inequalities. The developed methodology applies to linear parameter-varying systems with three or more sensors in which the state is observable through any one of the sensors. A fault in any of the system sensors can be uniquely identified by the diagnostic system. The developed methodology is applied to health monitoring of the lateral sensors on a GPS-controlled truck. A set of three sensors on the truck consisting of a GPS system, a lateral accelerometer and a yaw-rate gyroscope is considered. The performance of the fault diagnostic system is documented from extensive experimental results. Experimental results show that the fault diagnostic system can correctly detect a failure in any of the 3 sensors and accurately identify the source of the fault.