A Dynamic and Incentive Policy for Selecting D2D Mobile Relays

Abstract: User-to-network relaying enabled via Device-to-Device communications (D2D) is a promising technique for improving the performance of cellular networks. Since in practice relays are in mobility, a dynamic relay selection scheme is unavoidable. In this paper, we propose a dynamic relay selection policy that maximizes the performance of cellular networks (e.g. throughput, reliability, coverage) under cost constraints (e.g. transmission power, power budget). We represent the relays' dynamics as a Markov Decision Process (MDP) and assume that only the locations of the selected relays are observable. Therefore, the dynamic relay selection process is modeled as a Constrained Partially Observed Markov Decision Process (CPOMDP). Since the exact solution of such framework is intractable to find, we develop a point-based value iteration solution and evaluate its performance. In addition, we prove the submodularity property of both the reward and cost value functions and deduce a greedy solution which is scalable with the number of discovered relays. For the muti-user scenario, a distributed approach is introduced in order to reduce the complexity and the overhead of the proposed solution. We illustrate the numerical results of the scenario where throughput is maximized under energy constraint and evaluate the gain that the proposed relay selection policy achieves compared to a traditional cellular network.