Autoencoder and Matching-Based Resource Allocation Scheme for D2D Communication

A fifth-generation (5G) mobile network facilitates key enable technology for future communication systems known as device-to-device (D2D) communication, where two nearby devices communicate with each other without any involvement of the base station (BS). Here, the device's physical proximity establishes a communication link that enhances the system's spectral efficiency. However, it faces interference issues from the other device. In this paper, we proposed an efficient resource allocation scheme using a matching theory-based Hungarian algorithm. We resolve the random weight selection (when they are the same consecutive weight) issue in the Hungarian algorithm using a vanilla autoencoder. For the performance evaluation of the adopted scheme, we considered different parameters such as loss function, accuracy, and sum rate. The simulation results reveal the proposed autoencoder scheme recreates original input data with minimum reconstruction loss with 88% accuracy. Further, the Hungarian algorithm efficiently allocates D2D resources by maximizing the system's overall sum rate.