Deep Reinforcement Learning Based Robust Communication for Internet of Vehicles

Abstract

The high number of connected nodes in Internet of Vehicles (IoVs) drives to high data exchange between nodes, which increases the network overhead. Moreover, the recurrent change in vehicle mobility in Internet of Vehicles (IoVs) drives to frequent changes in network topology which in turn causes frequent link disconnections. Therefore, the most addressed issues in IoVs are to manage the high quantity of packets sent by the huge number of vehicles connected with IoT devices, to reduce communication delays and guarantee the longest communication stability. Clustering techniques have been utilized to reduce network overhead in IoVs networks. Classical clustering algorithms have been proposed to enhance network performances. However, IoVs environment is characterized by the high dynamicity of nodes, therefore, the optimization methods already proposed cannot perfectly deal with the characteristics of IoVs. Reinforcement learning (RL) is a machine learning algorithm, where the agent learns from its environment and tries to enhance its policies to obtain the best reward. In this paper, we propose to use deep reinforcement learning (DRL) to select the best cluster heads based on node’s degree, node’s buffer size, and signal strength. In the proposed work, the vehicle can perfectly select the cluster heads by choosing the best state-action values taking in consideration the high dynamicity of the network.