RL-based Resource Allocation for Improving Throughput in Cellular D2D Communications

With increasing demand of bandwidth-intensive application in cellular networks, coexistence of Device-to-Device (D2D) communications with cellular subscribers is a promising solution for high spectrum efficiency and network throughput. In cellular D2D communications, intelligent resource sharing among the network subscribers and paired devices is of significant importance. The most state-of-the-art works are relied on the exact values of Channel State Information (CSI) and subscribers’ transmission rate feedback which are not available in the real cases. In this paper, we propose a novel reinforcement-learning-based approach for mode selection and spectrum allocation called RL-D2D which shares efficiently resources amongst the D2D users and cellular subscribers with the need for CSI, achieving high network throughput. The results of evaluations show that RL-D2D achieves near-optimal performance and low outage rate in despite of lack of CSI and users’ transmission rate feedback.