Abstract. The vehicle collision warning system (CWS), enabled by recent advances in positioning systems and wireless communication technologies, plays a significant role in reducing traffic accidents. Since vehicle position is the most frequently exchanged information in vehicular networks and the key information in collision prediction, it becomes crucial to establish a strong level of trust in the positions, especially in active safety applications. To this end, in this paper we proposed a cooperative collision warning system (CCWS) based on digital map and dynamic vehicle model fusion. In CCWS, vehicles broadcast self-relevant information to neighbor vehicles periodically. In contrast to conventional methods that use only vehicle's dynamic models, the proposed system takes the information provided by a digital map into consideration to improve the accuracy of vehicle position estimation. Each vehicle calculates whether a potential collision may occur by fusing the dynamic vehicle model, digital map and the information received. If such a dangerous situation is predicted, an alarm will be triggered to inform the driver. Our experiment results show that the CCWS has a better performance in vehicle position estimation and collision prediction compared with the CWS without considering cooperation.