A self-adaptive RBF neuron network compensation control strategy based on the Lyapunov function is proposed in order to solve the path tracking problem of intelligent vehicles which is much complicated with nonlinear and time-varying characteristics. Firstly, the nominal dynamic model of vehicle’s path tracking is built. Then, RBF neuron network is used to compensate this nominal model’s inaccuracy parts. Finally, the learning rule is obtained based on the Lyapunov function, and the stability of this system is proved at the same time. The simulation results show that this strategy is much more accurate and with higher feasibility and practicability.