Deep Reinforcement-based Q-Learning Approach for Improving Energy Efficiency and Resource Management in V2X Communication Networks

Recent advancements in technologies have paved the way for developing intelligent transportation systems that offer safety and autonomous services on the road through the utilization of Internet and cellular technologies. Vehicles equipped with sensors are capable of transmitting data to other vehicles or infrastructure, serving various types of services. Efficient radio resource management schemes have become more important due to the increase in the number of vehicles on the road, along with other wireless devices. This paper introduces an innovative method utilizing Deep Reinforcement Q-Learning to enhance Energy Efficiency and Resource Management within V2X communication networks, in conjunction with established mobile networks. Initially, the network congestion of the V2X framework such as vehicle positions, traffic conditions and available energy is defined for the representation of state information along with the existing mobile network. To optimize energy efficiency along with resource management we define the action space for varying the transmission power, selecting the number of vehicles along with their movement status and mode of communication. The reward function is initiated to provide incentivized actions that can reduce resource usage and provide optimization in energy consumption. The Q-Learning approach using a deep neural network (DNN) is implemented by selecting state as an input value and for every action provided in state information, the Q value in the output is updated. The results obtained from the numerical simulations demonstrate that the DRL-based Q-learning approach makes energy-efficient decisions and allocates resources efficiently by improving system throughput and avoiding collisions.