Sensor fault detection in vehicle lateral control systems via switching Kalman filtering

Vehicle lateral control for automated highway systems (AHSs) is concerned with lane keeping and lane changing. It is of critical importance for safe highway operations; it must have fault tolerant capability in order to maintain stability in the event of malfunction of its components. In this paper, we investigate detection, isolation and accommodation of sensor faults. We propose a stochastic framework to which switching Kalman filtering and the EM algorithm are applied to detect faulty sensors and achieve good state estimation. With prior knowledge about the sensor failure modes, multiple sensors can be integrated into the framework, and the single failure assumption is no longer required. Simulations show that the sensor faults are detected immediately after their occurrences. Stability and performance are observed to be satisfactory in the event of sensor failures.