Performance enhanced adaptive cruise predictive control considering large curvature roads with variable adhesion coefficients

Abstract

This study investigates the challenges faced by adaptive cruise control (ACC) systems on roads with large curves and varying adhesion coefficients. A performance-enhanced predictive control strategy is proposed for the ACC system, considering variable adhesion coefficients and large curved roads. Firstly, a Levenberg–Marquardt backpropagation (LMBP) network trained using the similarity search method is established for online estimation of the road adhesion coefficient, which cannot be directly measured. Subsequently, a bidirectional long-short-term memory intelligent driver model (Bi-LSTM-IDM) is employed to estimate the velocity and position of the preceding vehicle (PV), thereby addressing the issue of not directly measuring these parameters under large curvature road conditions. Then, a coordinated longitudinal and lateral control strategy based on a nonlinear model predictive control (NMPC) is proposed to enhance the vehicle-following performance in terms of safety, comfort, and fuel economy on roads with large curvatures and varying adhesion coefficients. The simulation results, based on virtual real road scenes, as well as comparisons with existing ACC strategies, have verified the effectiveness and advantages of the proposed ACC strategy. In addition, the test platform embedded with Speedgoat verifies the computational efficiency of the proposed strategy for its feasibility in real-time applications.