The vehicular position information plays an important role in the vehicle communication. In open environments without signal blockage, global navigation satellite system (GNSS) has achieved on-road level accuracy and good reliability, However, the vehicle dynamic tracking is difficult in GNSS degraded environments. Usually, vehicle dynamic tracking consists of three portions: static positioning system, manoeuvring model and fusion algorithm. A radiobased positioning system aided by roadside units (RSUs) is employed as static positioning system, while the adaptive Kalman filter algorithm is used as fusion algorithm. In this paper, a new manoeuvring model is studied. Firstly, the problem of 'current' model is analysed. Secondly, a 'current-ellipse' manoeuvring model is proposed to adapt the strong and weak manoeuvring simultaneously. The experiments show that it can improve the tracking accuracy. Finally, a vehicle tracking case is discussed, its results show that root mean square error (RMSE) is smaller than 2 m, and the positioning accuracy can meet the position-based vehicular communication.