Resource Allocation Scheme for Guarantee of QoS in D2D Communications Using Deep Neural Network
Woongsup Lee, Kisong Lee
Abstract
In this letter, we propose a hybrid resource allocation scheme for multi-channel underlay device-to-device (D2D) communications. In our proposed scheme, the transmit power of D2D user equipment (DUE) allocated to each channel is controlled in order to maximize the sum rate of the DUEs for a given Quality of Service (QoS) constraints. We consider two QoS constraints such that the interference caused on cellular user equipment (CUE) is kept to be less than a predefined level and the rate of individual DUE is managed to be larger than a predefined threshold. In order to solve the drawbacks associated with previous deep neural network (DNN)-based approaches in which QoS constraints could be violated with high probability, a heuristic equally reduced power (ERP) scheme, is utilized together with a DNN-based scheme. By means of simulations under various environments, we verify that the proposed scheme provides a near-optimal sum rate while guaranteeing the QoS constraints with a low computation time.