Human-vehicle collaborative driving(HVCD) is a transitional technology of non-autonomous/autonomous driving. The semi-physical simulation can significantly reduce the experiment consumption of real vehicles. However, most simulation platforms are lacking four-wheel vehicle models and certain limitations in conventional coordinate system transformation methods. To fill the gap, this paper uses a nonlinear back-stepping control method based on the pose errors between the two-wheel differential mobile vehicle model and the four-wheel vehicle model, which achieves efficient and real-time tracking for the four-wheel vehicle motion trajectory. Physical steering wheel and pedals are applied for HVCD simulation under virtual reality environment, which provides reference for the development of the more realistic system. Vehicle forwarding velocity and guide wheel angle are utilized as the system input. Experiments are processed to compare and verify the method validity under both numerical and semi-physical simulation environments. Through the investigation of the pose errors between the two-wheel differential mobile vehicle and the four-wheel vehicle, the average accumulate error in heading direction is 4.56m in 10km.