Adaptive Channel Switching for Connected Vehicles under Extreme Weather Conditions: A Reinforcement Learning Based Approach

Due to the enormous volume of data generated by connected vehicles (CVs), future vehicle-to-vehicle (V2V) commu-nications will require a high-throughput communication channel. While 5G millimeter-wave (mmWave) technology generally meets these requirements, it suffers from the risk of performance degradation under harsh weather conditions such as heavy rain, snow or sand storms. This decline is primarily because of the high frequency of mmWave, which results in significant propagation attenuation loss. In contrast, 4G LTE is less susceptible to these environmental challenges. In an effort to improve communication performance reliability under severe weather conditions, we propose a Deep Reinforcement Learning (DRL)-driven approach. Our proposed framework utilizes the Received Signal Strength Indicator (RSSI), short-term throughput, and weather parameters as input variables. It employs cumulative throughput as the reward metric in the reinforcement learning process. As the framework interacts with the environment, it learns to dynamically switch between 5G mmWave and 4G LTE channels to maintain a robust and reliable communication link between CVs. Our approach has been validated through simulations using the ns-3 network simulator, enhanced with a customized weather model for 5G mmWave and 4G LTE channels. The simulation results confirm that our DRL framework substantially improves the reliability and performance of V2V communication within minutes, even in harsh environmental conditions.