In order to improve the robustness and stability of autonomous vehicle at high speed, a path tracking approach which combines front steering and differential braking is investigated in this paper. A bicycle model with 3-DOFs is established and a linear time-varying predictive model using front steering as its control input can be derived. Based on model predictive theory, the path tracking issue using linear time-varying model predictive control can be transformed into an online quadratic programming problem with constraints. The expected front steering angle can be obtained from online moving optimization. Then the direct yawing control is adopted to treat two types of differential braking control. The first one investigates steady-state gain of yaw rate in linear 2-DOFs vehicle model, and designs a stable differential braking controller which is based on reference yaw rate. The other one is based on the research of stable area of side-slip angle phase portrait, and corresponding differential braking control strategy tracking target path is designed. The structures of both differential braking systems are designed to be hierarchical, which consists of an upper level controller and a lower level controller. The upper one determines the desired additional yaw moment that tracking target path and the lower one determines the wheels to brake and the corresponding brake pressure required. The proposed controllers can compensate for tracking deviation and instability caused by the front wheel steering under complicated conditions. Simulation results show that the combined control method can significantly improve path tracking robustness and yaw stability of autonomous vehicle at high speed compared with unique front steering.