Manufacturing smarter and more reliable vehicles is a progressing trend in the automotive industry. Many of today’s vehicles are equipped with driver assistant, automated driving and advanced stability control systems. These systems rely on measured or estimated information to accomplish their tasks. Evidently, reliability of the sensory measurements and the estimate information is essential for desirable operation of advanced vehicle subsystems.

This thesis proposes a novel methodology to detect vehicle sensor faults, reconstruct the faulty sensory signals and deliver fault-tolerant estimation of vehicle states. The proposed method can detect failures of the longitudinal, lateral and vertical acceleration sensors, roll rate, yaw rate and pitch rate sensors, steering angle sensor, suspension height sensors, and motor torque sensors. The proposed structure can deliver fault-tolerant estimations of the vehicle states including the longitudinal, lateral and vertical tire forces, longitudinal and lateral velocities, roll angle, and pitch angle. Road grade and bank angles are also estimated in this method even in presence of sensor faults.

The unified structure in this thesis is realized by fusion of analytical redundancy relations, fault detection observers and adaptive state estimation algorithms. The proposed method can isolate the faults for vehicle stability and control systems and deliver accurate estimation of vehicle states required by such systems despite sensor failures.

The methods developed in this thesis are validated through experiments and can operate reliably in various driving scenarios.