To improve the safety of lateral obstacle avoidance for an intelligent vehicle with a shared driving control between human driver and autonomous driving system, a robust control strategy based on dynamic obstacle avoidance warning is proposed. Considering the dynamic changes of acceleration and speed of the vehicle ahead, the longitudinal safety distance model is established, and a multi-level warning algorithm based on the fusion and complementation of the critical longitudinal safety distance and the reciprocal of collision time is proposed. According to the different warning status and dynamic intervention of the driver, relative controls are put up for the vehicle so as to achieve the safety of the intelligent vehicle. Meanwhile, to realize the precise control of trajectory tracking, the robust control strategy of lateral obstacle avoidance based on parameter perturbation is proposed. Finally, the simulation platform of Carsim and Simulink is constructed, and the proposed control strategy is verified. The results indicate that the proposed control strategy can obtain different warning levels and can dynamically adjust warning states with dynamic intervention of the driver. If the driver does not respond to the warning until the fourth level warning, the control strategy automatically takes over the intelligent vehicle and carries out the lateral obstacle avoidance control, improving the safety and yaw stability of the intelligent vehicle.