Secure Relay Selection with Outdated CSI in Cooperative Wireless Vehicular Networks: A DQN Approach

Cooperative communications is a core research area in wireless vehicular networks (WVNs), thanks to its capability provide certain degree of fading mitigation and improve spectral efficiency. In cooperative scenario, the intercept probability system can be reduced by optimizing relay selection scheme order select optimal from set available relays for data transmission. However, due mobility WVNs, best often selected practice based on outdated channel state information (CSI), which turn affects overall performance. Therefore, there need robust (RSS) that guarantees satisfactory achievable performance presence an CSI. Motivated this considering advantageous features autoregressive moving average (ARMA), present contribution models communication scenario with as Markov decision process (MDP) proposes two deep Q-networks (DQNs), namely DQN-RSS DQN-RSS-ARMA. proposed framework, reinforcement learning (RL)-based RSS are trained probability, aiming multiple relays. We then compare conventional methods evaluate network security point view. Simulation results show DQN-RSS-ARMA perform better than approaches, they reduce approximately 15% 30%, respectively, compared standard ARMA approach.