LLMs-based On-off Control and Beamforming Optimization for RRIS-assisted Outdoor-to-Indoor mmWave SWIPT for Mobile Robots

Refracting reconfigurable intelligent surface (RRIS) has emerged as a promising solution for enhancing outdoor-toindoor millimeter-wave (mmWave) simultaneous wireless information and power transfer (SWIPT). However, current studies have not thoroughly examined the on-off control of RRIS elements to support multiple mobile indoor robots, nor have they adequately addressed algorithmic efficiency. In this paper, we explore RRIS-assisted outdoor-to-indoor mmWave SWIPT where an RRIS is mounted on an exterior wall of a building to refract mmWave signals from an outdoor base station (BS) to indoor mobile robots navigating around blockages. We propose an optimization problem aimed at maximizing the minimum harvested power among all robots while ensuring that the robots' downlink data rate requirements are met. This is achieved by jointly optimizing the on-off states and phase shifts of RRIS elements, the beamforming vectors of the BS and the SWIPT time switching factor. To address the problem's nonconvexity and variable dimensionality, we decompose it into two subproblems: one optimizes the phase shifts of RRIS elements and beamforming vectors of the BS, and the other optimizes the RRIS elements' on-off states. We then develop an iterative algorithm, where in each iteration the two subproblems are solved by a successive convex approximation (SCA) and penalty method based algorithm and a large language models (LLMSs) driven algorithm, respectively. Simulation results show that the proposed algorithms significantly enhance the convergence speed compared to existing methods, providing an optimal RRIS on-off control strategy with improved performance.