Improved model predictive control trajectory-tracking algorithm with adaptive horizon

Abstract

Aiming at the problems of poor tracking accuracy and low real-time performance of existing model predictive controllers for autonomous vehicles under high-curvature conditions, an Adaptive Horizon-Based Improved Model Predictive Control (AH-BMPC) method was proposed. In this method, the weight coefficient matrix of Model Predictive Control (MPC) is designed based on the block matrix strategy. Then, based on the lateral stability class errors and the lateral trajectory errors, the adaptive horizon system was constructed to realize the proposed method. Simulation results demonstrate that under speeds of 10m/s, 15m/s, and 17m/s, the proposed method achieves an average reduction of 42.54% in lateral trajectory error and 41.06% in heading angle error compared to the MPC method. In addition, the computational efficiency is improved by an average of 10.53%. In practical applications, lateral trajectory errors can be calculated based on navigation information, and lateral stability evaluation indicators can be obtained through vehicle body sensors. Thus, the AH-BMPC method can be deployed in applications.