Leveraging Deep Unsupervised Learning for Joint Passive Beamforming and Power Control of RIS-Aided D2D Networks with Energy Constrained Nodes

In this work, we investigate sum-rate for the reconfigurable intelligent surface (RIS)-aided device-to-device (D2D) networks with energy constrained wireless nodes. Specifically, we aim to maximize the sum-rate for the RIS-aided D2D networks while meeting minimum energy requirements of the energy constrained wireless nodes. To this end, we formulate a problem that involves joint optimization of phase-shifts of reflecting elements of the RIS (passive beamforming) and transmit power control of each transmitter. Nonetheless, this optimization problem is a high-dimensional non-convex, thus finding its optimal solution is significantly challenging. In order to solve this optimization problem efficiently, we propose a novel data-driven scheme that is based on the deep unsupervised learning (DUL) mechanism. It is shown through simulations that our proposed data-driven scheme achieves a sum-rate comparable to that of the existing benchmark scheme, yet our scheme notably excels in efficiency, i.e., demonstrating a time-complexity which is at-least ten times lower compared to the existing benchmark.