With the recent advancements in Internet of Things (IoTs) as well as cyber physical systems (CPS), the research and development of autonomous driving have achieved tremendous progress. In this new era, the realization of communication schemes requires low-latency and ultra-reliable communication, which is quite challenging, especially in highly dynamic and complex communication environments. This paper investigates a collaborative communication scheme in vehicular cyber physical systems (VCPS). Firstly, we introduce a vehicle density prediction model according to the spatio-temporal features of trajectories. Then, we divide the map into grids and present a routing path evaluation scheme that jointly considers the vehicle density, link quality, and routing length. Moreover, a grid-based data transmission method is proposed to select the optimal relay nodes according to real-time traffic information. Finally, experiments using real-world vehicle trajectories are conducted to validate the effectiveness of our method.