Sliding Mode Approach for Partitioned Cost Function-based Fault-Tolerant Control of Automated Driving

Abstract

This paper presents sliding mode and partitioned cost function-based fault-tolerant controller of automated driving. A proper strategy for ensuring functional safety of autonomous vehicles is needed when there exist sensor faults in acceleration information used for longitudinal autonomous driving. The data-driven fault-tolerant control algorithm proposed in this study is based on the upper-level controller decoupled with the lower-level controller. The adaptive sliding mode observer (ASMO) using recursive least squares (RLS) for reconstruction of acceleration sensor fault signal has been designed with gradient descent method. The reconstructed fault signal has been used to compute the desired acceleration for fault-tolerant longitudinal control with the Lyapunov stability condition. In order to compute the lower-level control inputs such as acceleration and brake pedal inputs, the desired and current acceleration values have been used based on the PID control law. It is assumed that the longitudinal acceleration of the preceding vehicle can be obtained using V2V communication. The performance evaluation environment has been constructed using Matlab/Simulink and CarMaker software. The evaluation results shows that the desired acceleration can be tracked reasonably by the proposed fault-tolerant control algorithm despite of existence of fault signal in longitudinal acceleration value.