Cooperative Reinforcement Learning for Adaptive Power Allocation in Device-to-Device Communication

Mobile devices are an intrinsic part of the Internet of Things (IoT) paradigm. Device-to-device (D2D) communication is emerging as one of the viable solutions for the radio resource optimization in an IoT infrastructure. However, it also comes with the challenges associated with power allocation as it causes severe interference by reusing the spectrum with the cellular users in an underlay model. Therefore, efficient techniques are required to reduce the interference with proper power allocation. In this paper, we propose a cooperative reinforcement learning algorithm for adaptive power allocation in D2D communication which helps to provide better system throughput as well as D2D throughput with less interference. We perform cooperation by sharing the value function between devices and incorporating a neighboring factor. We design our states for reinforcement learning with appropriate application-defined variables which provide a longer observation space. We compare our work with the existing distributed reinforcement learning method and random allocation of resources. Simulation results show that the proposed algorithm outperforms the distributed reinforcement learning and the random allocation both in terms of overall system throughput as well as D2D throughput by adaptive power allocation.