Adaptive Model Predictive Control With Variable Prediction Time for Path Tracking of Autonomous Vehicles

Path tracking is a core technology for autonomous vehicles. Model predictive control is well recognized as one of the promised approaches in this field. This paper presents an adaptive model predictive control with variable prediction time (VPT-AMPC) for path tracking of autonomous vehicles. To acquire more accurate vehicle planar dynamics in a linear form, the linear-equivalent brush tire model is developed. Subsequently, the relationship between prediction time and path curvature, vehicle longitudinal velocity, and tire-road friction coefficient are established through data collection and mapping. An prediction time adaptive adjustment strategy is developed based on the coordinated tuning of the prediction horizon and discrete time interval to improve tracking accuracy without extra computational burden. Meanwhile, real-time adaptive constraints are applied to improve lateral stability. Finally, the simulation and hardware-in-the-loop results demonstrate that the proposed controller improves tracking performance and reduces computation time in the linear region, while enhancing lateral stability and tracking accuracy in the nonlinear region without increasing computational burden.