To improve the dynamic response ability of unmanned vehicle tracking process and ensure that the vehicle can track the reference route quickly and stably, this paper firstly introduces the time-varying forward-looking distance into the traditional line-of-sight (LOS)guidance strategy based on the idea of fuzzy control, and proposes the improved adaptive LOS guidance strategy, which simplifies the tracking of target trajectory to the heading of the target point track. Secondly, a three degree of freedom dynamic model of vehicle is established, and the linear system mathematical model of path tracking is designed by combining the tracking error variable of adaptive LOS. Finally, multi-step prediction, rolling real-time optimization and feedback correction are used to solve the optimal feedback steering wheel control command, based on the principle of model prediction. To verify the effectiveness of the above-mentioned tracking strategy, the straight-line path and the curve path are used as the reference path in the Simulink simulation environment. The results show that the proposed adaptive LOS guidance strategy can make the tracking vehicle's lateral and heading errors converge to zero rapidly, which verifies that the adaptive LOS guidance algorithm can improve the response speed and stability of unmanned vehicle path tracking.