Deep Reinforcement Learning Based User Association and Resource Allocation for D2D-enabled Wireless Networks

With the ultra-dense deployment of small-cell base stations (SBSs), it is common today to find a user locates within the coverage area of several SBSs. In this paper, we investigate the joint user association and resource allocation problem of D2D pairs in ultra-dense cellular networks. Specifically, we formulate an optimization problem for D2D pairs that are within the overlapping coverage areas of several SBSs. By jointing optimizing the user association and resource allocation of such D2D pairs, we maximize the overall data rate of both cellular users and D2D pairs. After that, the double-dueling-deep Q-network (D3QN) algorithm is adopted to address the formulated problem, in which we consider the central controller in the network as an agent, and let it interact with the environment to find the optimal user association and resource allocation strategy. Numerical results validate that our proposed D3QN algorithm could achieve near-optimal performance, and is superior to other schemes.