Joint Active RIS Configuration and User Power Control for Localization: A Neuroevolution-Based Approach

TL;DR

This work tackles localization using reconfigurable intelligent surfaces (RIS) with discrete phase shifts and introduces a neuroevolution-based, multi-agent framework to jointly configure RIS phases and uplink transmit power under a 1-bit feedback constraint. The BS-side policy and UE power controller, together with an LSTM-based estimator, are evolved via CoSyNE, while a final estimator is retrained under learned policies to achieve accurate UE localization. Results show this approach surpasses fingerprinting and constrained deep learning baselines and nearly matches performance of schemes with full power signaling, all under practical RIS hardware limitations and minimal feedback overhead. The method offers a scalable, gradient-free solution for active sensing in RIS-aided localization with potential impact on energy efficiency and privacy in future 6G deployments.

Abstract

This paper studies user localization aided by a Reconfigurable Intelligent Surface (RIS). A feedback link from the Base Station (BS) to the user is adopted to enable dynamic power control of the user pilot transmissions in the uplink. A novel multi-agent algorithm for the joint control of the RIS phase configuration and the user transmit power is presented, which is based on a hybrid approach integrating NeuroEvolution (NE) and supervised learning. The proposed scheme requires only single-bit feedback messages for the uplink power control, supports RIS elements with discrete responses, and is numerically shown to outperform fingerprinting, deep reinforcement learning baselines and backpropagation-based position estimators.

Figures (2)

• Figure 1: Graphical illustration of the proposed MA algorithm.
• Figure 2: Root Mean Squared Error (RMSE) for all localization schemes considering different observation formats.