Intelligent Safety Message Dissemination with Vehicle Trajectory Density Predictions in VANETs

Integration of wireless communication systems and machine learning techniques are generating new applications and services in vehicle ad-hoc networks (VANETs). By analyzing data transmission in vehicle-to-vehicle (V2V) communications and vehicle-to-infrastructure (V2I) communications, an intelligent transportation system (ITS) can provide better safety applications. This work explores machine learning approaches to estimate vehicle density on predicted trajectories, which is further utilized to provide intelligent safety message dissemination. With our approach, the traffic safety message, such as accident notifications, will only be disseminated to relevant vehicles that are predicted to pass by the accident areas. Depending on the network connectivity, our system adaptively chooses vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I) or hybrid communications to disseminate a message to relevant vehicles. We evaluate the system by using real-world VANET mobility datasets, and experiment results show that our system outperforms other mechanisms without considering predicted vehicle trajectory density information