A path tracking method based on robust model predictive control（RMPC） was designed aiming at the problems of low tracking accuracy and poor robustness caused by model parameter uncertainties, external disturbance, system state time-varying and nonlinearities in the process of path tracking of intelligent vehicles. The nonlinear characteristic of tire was considered, and the cornering stiffness of the tire was modified. Then the time-varying longitudinal speed was considered and a polytope with finite vertices was used to describe the longitudinal vehicle speed. A discrete vehicle polytopic uncertainty model was established. Linear matrix inequalities（LMI）with slack variables were derived to solve the optimization problem according to the Lyapunov asymptotic stability and quadratic performance index in infinite horizon. An improved offline robust model predictive control algorithm was proposed to improve the real-time performance of the controller and reduce its conservativeness. The effectiveness of the controller was verified by Simulink-Carsim co-simulation and hardware-in-the-loop tests. The simulation results showed that the designed controller had good tracking accuracy and robustness.