Journal
IEEE INTERNET OF THINGS JOURNAL
Volume 10, Issue 19, Pages 17323-17337Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/JIOT.2023.3275818
Keywords
Deep reinforcement learning (DRL); Internet of Things (IoT); multiuser multiple-input-single-output (MU-MISO); nonconvex optimization; reconfigurable intelligent sur-face (RIS)
Ask authors/readers for more resources
This article introduces a method of creating virtual Line-of-Sight channels in an IoT network using reconfigurable intelligent surfaces (RIS). It aims to maximize the sum-rate of communication systems by jointly optimizing the phase shift matrix of the RIS and transmit beamforming. A piecewise-deep reinforcement learning algorithm is proposed to solve the nonconvex problem, which avoids local optima and achieves faster convergence.
With the widespread connectivity of everyday devices realized by the advent of the Internet of Things (IoT), communication between users of different devices has become increasingly close. In practical scenarios, obstacles present between the transceiver may cause a deterioration in the quality of the received signals. Therefore, the reconfigurable intelligent surface (RIS) is employed to create virtual Line-of-Sight (LoS) channels in an IoT network. Specifically, this article aims at maximizing the sum-rate of the RIS-assisted multiuser multiple-input-single-output (MU-MISO) communication systems by jointly optimizing the phase shift matrix of the RIS and transmit beamforming. To solve the formulated nonconvex problem, a piecewise-deep reinforcement learning (DRL) algorithm is proposed in this article. Unlike the existing alternative optimization (AO) algorithms, the proposed algorithm avoids falling into the local optimal by using an exploration mechanism. Moreover, piecewise-DRL can reduce the action dimension, allowing the algorithm to obtain faster convergence. Simultaneously, this algorithm also ensures that the parameters of the two-part networks are updated to generate a larger system sum-rate by unsupervised joint optimization. Simulations in various circumstances reveal that the proposed approach is more robust and presents better stability and faster convergence than previous state-of-the-art algorithms while obtaining competitive performance.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available