Drift Control of Autonomous Vehicle Under Extreme Condition Based on Neural Network Tire Model

Under extreme conditions, professional rally drivers will intentionally slip the wheels and use the special driving technique of "drifting" to control the vehicle and improve its safety of the vehicle. However, tire force has strong nonlinear characteristics under extreme drift conditions, and traditional tire models cannot accurately describe its characteristics. For this reason, this paper proposes a drift control algorithm based on the Neural Network (NN) tire model to improve the accuracy of vehicle drift. First, the 3-state-single-track vehicle model with the Magic formula is adopted to design the basic drift controller of the linear quadratic regulator (LQR) based on the solution of the drift equilibrium state. The controller is used to realize the drift of the vehicle and collect data to train the NN tire model. By using the NN tire model, more accurate lateral tire forces and corresponding drift equilibria states could be obtained. At the same time, the NN tire model updates the tire parameters of the LQR controller in real-time according to the vehicle state. Finally, the drift controller is verified by co-simulation with MATLAB/Simulink and CarSim, and the results show that the proposed method can significantly improve the accuracy of drift motion control.