Over-the-Air Data Aggregation Aided by Multi-Functional RIS in Internet of Robotic Things

Xinran Zhang, Hui Tian, Wanli Ni, Zhaohui Yang

Abstract: To facilitate fast data aggregation in Internet of Robotic Things, over-the-air computation (AirComp) is a communication-efficient enabler by virtue of its high spectrum efficiency and low transmission latency. In this paper, we introduce the multi-functional reconfigurable intelligent surface (MF-RIS) as a means to mitigate aggregation mean square error (MSE) by enhancing signals across the entire space for robots. However, imperfect channel state information (CSI), stemming from inaccurate channel estimation and robot mobility-induced channel aging, can degrade transmission accuracy. This challenge is further exacerbated when future dynamic CSI is unavailable. We aim to minimize long-term MSE through transceiver beamforming and MF-RIS coefficient design. To confront system dynamics without prior knowledge, we propose a deep reinforcement learning (DRL)-based double-agent algorithm. Our DRL-based algorithm comprises an MF-RIS mode-guided strategy and an asynchronous agent collaboration framework. Specifically, one agent designs MF-RIS modes, guiding the second agent in configuring transceiver and MF-RIS beamforming, and then an asynchronous network update framework is proposed to enhance agent interaction. Numerical results corroborate the performance gain brought by the MF-RIS, as well as the effectiveness and robustness of the proposed algorithm in suppressing long-term MSE under imperfect CSI.

External IDs:dblp:journals/tvt/ZhangTNY25