Data-Driven Lateral Fault-tolerance Control of Autonomous Vehicle System Using Reinforcement Learning

In this paper, a data driven lateral fault-tolerant control (LFTC) method is proposed for four wheels drive vehicles, which considering both vehicle velocity and trajectory tracking performance. The novel design of the lateral control employs with zero-sum game and adaptive dynamics programming technique to solve the Riccati equation without requiring the knowledge of system, only using online data. The LFTC consists of the data driven off-policy and adaptive control. The adaptive parameters adjusted online to compensate the actuator faults automatically. The tracking system is asymptotically stable with the disturbance attenuation level γ. Finally, simulation is provided to show the effectiveness of the proposed method.

I. INTRODUCTION

In recent decades, the trajectory tracking control of four wheel drive vehicle has attracted the attention of many researchers at home and abroad. Tracking control has always been a basic problem in collaborative control [1], formation control [2], [3] and path planning [4], [5]. However, due to the following reasons: first, the system is subject to various physical constraints at the level of motion; second, the current model is difficult to obtain accurate dynamic models and the strong coupling and high nonlinearity. These difficulties constitute obstacles to the control system. Effective control of four wheel drive vehicle trajectory tracking is still a challenging problem.