The tyre model based on neural network was built. Inference rules of working condition were established based on the classification of electric power steering (EPS) conditions, which could lead to a correct judgment. Furthermore, according to the different working conditions of EPS, the fuzzy rules of fuzzy PID were changed to achieve active switching of multi-assist characteristics. The whole multi-body dynamic vehicle model was built. The SIMPACK-Matlab co-simulation was carried out on the snakelike road, and then compared with the car test based on dSPACE. The simulation result showed that the control strategy based on the proposed model could effectively reduce the driver’s operating torque and improve the returnability of vehicle. 