A comparison between a centralised multilayer LPV/ℋ∞ and a decentralised multilayer sliding mode control architectures for vehicle's global chassis control

Abstract

This paper investigates new achievements in global chassis control, involving Active Front steering (AFS) and Direct Yaw Control (DYC), to improve the overall vehicle performance, i.e. the vehicle manoeuvrability, lateral stability and rollover avoidance. Two multilayer control architectures, each formed by three hierarchical layers, are developed, validated and compared. The lower layer represents the actuators implemented into the vehicle which generate their control inputs based on the orders sent from the middle layer. The middle layer is the control layer which is responsible to generate the control inputs that minimise the errors between the desired and actual vehicle yaw rate, side-slip angle and roll angle, regardless of the driving situation. The control layer is the main difference of the proposed architectures, where a centralised and a decentralised controllers are developed. In the centralised architecture, the novelty is that one single Multi-Input-Multi-Output MIMO optimal controller generates the optimal additive steering angle provided by the AFS and the optimal differential braking provided by the DYC to minimise -at once- all the vehicle state errors . The optimal control technique based on offline Linear Matrix Inequality optimal solutions, in the framework of Linear-Parameter-Varying systems, is applied to synthesise the controller. In the decentralised architecture, a heuristic solution is proposed by decoupling the control problem where the Super-Twisting Sliding Mode (STSM) technique is applied to derive the AFS control input which minimises only the errors on the yaw rate, and the roll angle. Similarly, the DYC control input is privileged to minimise only the error on the side-slip angle. The higher layer of both architectures is the decision making layer which instantly monitors two criteria laying on lateral stability and rollover risks. Then, it generates two weighted parameters which adapt the controller(s) performance(s) according to the driving conditions in order to improve the vehicle's manoeuvrability, lateral stability and rollover avoidance. Both control architectures are tested and validated on the professional simulator ‘SCANeR Studio’. Simulation shows that both architectures are relevant to the global chassis control. The centralised one is optimal, complex and overall closed-loop stability is guaranteed, while the decentralised one does not guarantee the overall closed-loop stability, but it is intuitive, simple and robust.