Beam Training for Self-Sustainable RIS

This paper studies beam training for self-sustainable reconfigurable intelligent surfaces (RISs), where the RIS is en-dowed with energy harvesting capabilities in order to operate independently of external power sources. In particular, we consider codebook-based reconfiguration of the RIS for periodic alignment of the reflection beam with the current location of a mobile user. Moreover, a fraction of the RIS unit cells harvests energy from the incident signal while the remaining unit cells are used for signal reflection. Based on this framework, we analyze the code-book design, beam training overhead, and power consumption. Furthermore, we show that the optimal unit cell split ratio can be obtained based on a convex optimization problem. In addition, for low-overhead beam training, we derive a closed-form expression for the rate loss caused by the energy constraint needed for self-sustainable operation. Finally, we provide simulation results to evaluate the proposed analytical framework and reveal insights on the tradeoff between beam training overhead, RIS response time, and RIS power consumption.