Reinforcement Learning-based Multi-hop Intelligent Route Selection for Vehicle-to-Vehicle Visible Light Communication

In this paper, our study focuses on the multi-hop routing problem in long-distance visible light communication. We propose a Q-learning intelligent route selection algorithm based in visible light communication, with the goal of maximizing the communication quality of vehicle-to-vehicle. Firstly, the communication range of the current vehicle is determined based on its operational state information, allowing to obtain the set of neighboring vehicles. Secondly, a new routing metric is defined by combining the inter-vehicle signal-to-noise ratio factor and the distance factor. Finally, during training, the Q-value is updated using a dynamic learning rate and exploration rate, enabling faster convergence. To evaluate the performance of the proposed algorithm, a Python-based open-source program is connected to the Simulation of Urban Mobility traffic simulator and controls the simulation. The simulation results demonstrate that the proposed algorithm outperforms existing solutions and can effectively improve the communication quality and delivery efficiency of the communication link.