Relay Selection for Energy-Harvesting Collaborative Communication Systems: A Deep Reinforcement Learning Approach

In wireless networks, route transmission can be accomplished, and the system's long-term channel capacity can be increased by leveraging the collaboration among numerous relay nodes. While earlier relay selection methods computed all signal-to-noise ratios before selection, they proved to be ineffective. This study addresses the relay selection issue in wireless collaborative communication by employing a deep reinforcement learning (DRL)-based relay selection technique. Initially, this study examines the collaborative communication process of wireless communication networks as a markov decision process (MDP). Subsequently, it proposes a relay selection technique utilizing entropy proximal policy optimization (EPPO), which incorporates policy entropy. The optimal policy is established by interacting with the environment to select the best relay from among numerous relays. Ultimately, EPPO is compared to the original proximal policy optimization (PPO) algorithm, the Deep Q-Network (DQN) algorithm, and a random relay selection scheme under the premise of unknown instantaneous channel state information (CSI). The network performance is evaluated in terms of system capacity and outage probability. The results of the simulation experiment demonstrate that EPPO outperforms current solutions in these areas.