This paper investigates the motion control of automated vehicles for both lane-changing and lane-keeping maneuvers. This research is critical because lane keeping and lane changing, which need to be integrated into a unified control system, are still two fundamental control problems on the way to developing the highly automated vehicle. In addition, environment perception, which is highly coupled with motion control, should be introduced into the control loop. A further challenge is to solve the complex optimization problem with constraints of vehicle dynamics and full-dimensional collision avoidance. To solve these issues, this paper proposes a unified model predictive control method that can seamlessly handle lane-keeping and lane-changing maneuvers. The control problem adopts three reference generation approaches to get the perception of the traffic environment involved. Further, a rough-plan-and-fine-check strategy is utilized to reduce the complexity of solving the proposed unified model predictive control problem with constraints of collision avoidance. The proposed method has been implemented on the PreScan-MATLAB/Simulink joint simulation platform, where its performance of lane keeping and lane changing has been evaluated in different driving scenarios. Simulation results verify the capabilities of the proposed method.