Dual-Agent AI framework slashes power use in RIS-aided tracking

A team of researchers led by George Stamatelis has proposed a novel Dual-Agent (DA) deep learning framework for energy-efficient tracking of power-limited mobile users assisted by a Reconfigurable Intelligent Surface (RIS). The core challenge is that localization pilot transmissions dominate the energy budget of battery-constrained devices. To reduce this overhead, the framework introduces a low-overhead feedback link from the base station to the user, enabling dynamic uplink power control. The DA framework jointly optimizes the discrete RIS phase profiles and the user equipment's transmit power in real time.

The key innovation is a hybrid training methodology that integrates neuroevolution with supervised learning. This approach overcomes two major obstacles: the non-differentiability of discrete phase responses from RIS unit elements and the strict information bottleneck of single-bit feedback messages for pilot power control. Extensive numerical simulations demonstrate that the proposed scheme achieves highly accurate and robust tracking across diverse target motion models, significantly outperforming extended Kalman filters, particle filters, and machine learning–based trackers. In static localization, it also beats traditional fingerprinting, deep reinforcement learning, and standard backpropagation-based estimators. The framework works with both single- and multi-antenna base stations with minimal modifications.