Learning-based routing approach for direct interactions between wireless sensor network and moving vehicles

Wireless sensor network (WSN) has been applied to traffic information collection in many researches. Moving vehicles in the road need to acquire real time traffic information directly from WSN to make proper decisions and avoid potential accidents. Due to the large scale deployment of WSN, routing is necessary to deliver real time information to vehicles in a multi-hop way. Taking the moving vehicle as a mobile sink, a reinforcement learning-based routing approach is proposed to support sink mobility and enable direct interactions between WSN and vehicles. Multiple metrics including time delay, network lifetime and reliability are ensured by designing a comprehensive reward function for learning. The convergence speed of learning is improved with sink announcement. Simulation results show the feasibility of the proposed approach for direct interactions between WSN and moving vehicles. Comparisons are also carried out to show the superiority of the proposed approach over other approaches.