Efficient DRL-Based Selection Strategy in Hybrid Vehicular Networks

Emerging V2X applications, like Advanced Driver Assistance Systems (ADASs) and Connected Autonomous Driving (CAD) require Ultra-Reliable Low Latency Communications (URLLC). Unfortunately, none of the existing V2X communication technologies, such as ETSI ITS-G5 or C-V2X (Cellular V2X including 5G NR), can satisfy these requirements independently. In this paper, we propose a scalable hybrid vehicular communication architecture that leverages the performance of multiple Radio Access Technologies (RATs). To this purpose, we propose a novel ITS station protocol stack and a decentralized RAT selection strategy that uses Deep Reinforcement Learning (DRL). The proposed approach employs a double deep Q-learning (DDQN) algorithm that allows each vehicle to determine the optimal RAT combination to meet the specific needs of the V2X application while limiting resource consumption and channel load. Furthermore, we assess the ability of our architecture to offer reliable and high throughput communication in two different scenarios with varying traffic flow densities. Numerical results reveal that the hybrid vehicular communication architecture has the potential to enhance the packet reception rate (PRR) by up to 30% compared to both the static RAT selection strategy and the multi-criteria decision-making (MCDM) selection algorithm. Additionally, the selection strategy exhibits about a 20% improvement in throughput and a 10% reduction in the channel busy ratio (CBR).