Abstract
In this Letter, a deep reinforcement learning‐based approach is proposed for the delay‐constrained buffer‐aided relay selection in a cooperative cognitive network. The proposed learning algorithm can efficiently solve the complicated relay selection problem, and achieves the optimal throughput when the buffer size and number of relays are large. In particular, the authors use the deep‐Q‐learning to design an agent to estimate a specific action for each state of the system, which is then utilised to provide an optimum trade‐off between throughput and a given delay constraint. Simulation results are provided to demonstrate the advantages of the proposed scheme over conventional selection methods. More specifically, compared to the max‐ratio selection criteria, where the relay with the highest signal‐to‐interference ratio is selected, the proposed scheme achieves a significant throughput gain with higher throughput‐delay balance.