Resource management semi-parallel method in RIS-assisted user centralization cellular removal system based on DRL

Abstract

The invention provides a deep reinforcement learning (DRL)-based semi-parallel method used in a reconfigurable intelligent surface (RIS)-assisted user centralized cellular removal (UCCF) system. According to the method, an optimization problem is decomposed into two iterative sub-tasks, namely an access point (AP) and user equipment (UE) association (AUA) sub-task and a transmit power and RIS reflection coefficient management (PRCM) sub-task. The method specifically comprises the following steps: adopting a binary particle swarm optimization (BPSO) algorithm for integer nonlinear programming AUA; for a PRCM of multi-continuous variable joint optimization, a parallel algorithm based on a DRL is provided, a double-delay depth deterministic strategy gradient (TD3) algorithm is adopted to improve convergence, and a new state preprocessing mechanism is provided. According to the method, a plurality of variables are optimized by adopting a DRL-based semi-parallel method, the limitation of solving an NP-hard problem by a traditional algorithm is overcome, the accuracy is improved, and meanwhile, the intelligent method can be quickly adjusted when facing environment change without a large amount of priori knowledge.

Description

Technical field [0001] The present invention designs a DRL-based semi-parallel joint optimization method for the RIS assisted UCCF system. Specifically, the program considers the complexity of multivariate (including discrete variables and continuous variables), and designs a DRL-based semi-parallel framework and tips for the converging of neural networks. To maximize system availability, joint optimization of AUA, UES transmit power and RISS reflection coefficient, belong to artificial intelligence wireless communication technology. Background technique [0002] In recent years, with the development of wireless communications and artificial intelligence, the concept of "Wan I Zhilin" has been proposed, and the cell-centric cellular system is limited by the edge effect, and the increasing system capacity demand is no longer. The CF system has been deployed by a large number of distributed APS services in UES, eliminating the concept of the cell boundary, increasing the spatial ma...

AI Technical Summary

Problems solved by technology

However, RIS also brings the challenge of optimizing the RIS reflection coefficient matrix to the UCCF system, which increases the design complexity and convergence time of the algorithm

[0004] Due to the complex characteristics between variables in RIS-assisted UCCF system, it is difficult to capture the potential i

Images