Abstract: The on-board edge computing technology brings low delay and high reliability service to on-board users. However, the rapid movement of users in the vehicle-mounted environment will interrupt the information transmission between vehicles and RSU. Therefore, a vehicle network architecture based on mobile edge computing is proposed, and a vehicle prediction mechanism is designed for low delay and small task applications. In this mechanism, particle filter algorithm was used to predict vehicle movement, and the optimal prediction position of vehicle nonlinear motion was obtained through finite iteration and modification of prediction and update. During task processing, the most suitable RSU was selected to ensure the smooth transmission of information. Simulation results show that this mechanism greatly improves the probability of successful connection between vehicle and RSU, and improves the service quality of vehicle network system.