In the obstacle avoidance condition of the human-machine cooperative driving system， a human-machine cooperative obstacle avoidance controller is designed considering the driver’s characteristics， aiming at the problem that the driver cannot control the car in time due to the driver’s inattention. Based on the dual-driver and dual-control man-machine co-driving structure， the automatic driving system uses a Linear Quadratic Regulator （LQR） to control the vehicle to track the planned trajectory， and the driver model is established based on the optimal preview lateral acceleration model.To allocate driving rights more reasonably in the process of obstacle avoidance， a driving simulator is used to collect fatigue driving data to train a BP neural network to identify the driver’s fatigue operation behavior. The human-machine cooperative obstacle avoidance strategy is designed by modeling the driver fatigue factor and space risk of collision. Carsim，PreScan and Simulink co-simulation platform is built to conduct obstacle avoidance simulation experiments under high-speed conditions. The simulation results show that， compared with the traditional method， the proposed strategy reduces the lateral acceleration， sideslip angle and yaw rate by 8.9%， 18.2% and 11.1%， respectively， during the static obstacle avoidance process， and in the process of dynamic obstacle avoidance， the corresponding indicators are reduced by 51.5%， 53.4% and 50.6%， respectively. Therefore， the stability of the vehicle during obstacle avoidance is improved.