RHONN-modeling-based predictive safety assessment and torque vectoring for holistic stabilization of electrified vehicles

Abstract

Improving vehicle safety and reducing traffic accidents have always been of cardinal importance in vehicle dynamics control fields. A reasonable and comprehensive safety index that characterizes the vehicle's safe region is the most challenging aspect of research. With the linear dynamics model as the benchmark, this article uses the deviation of yaw rate and vehicle sideslip angle from the corresponding linear response as the lateral stability indices. Meanwhile, the maximum slip ratio of the driven wheels is selected as the longitudinal stability index. The safety indicator , a quantitative index featuring the safety degree of vehicle planar motions, is then inferred by a fuzzy inference system using the lateral and longitudinal stability indices. As such, the recurrent high-order neural network model predicts the vehicle states. Based on the predicted states, a safety indicator is then derived by using fuzzy inference system, which can assess the safety of a driver's control commands. In the case of an improper driving torque demand given by the driver, the torque correction process is immediately conducted to maintain the vehicle in a safe region. Finally, two typical scenarios—slippery curves and double lane changes in low friction roads—are simulated on the MATLAB/Simulink-CarSim cosimulation platform. The hardware-in-the-loop experiments are also conducted on a driving simulator test rig, validating the performance of the developed algorithms. The holistic stability performance of the in-wheel motor driven vehicle is thoroughly analyzed and compared using three existing methods. The simulation and experimental results validate the effectiveness and feasibility of the proposed method.