Codebook-Based Configuration of Reconfigurable Intelligent Surfaces

Reconfigurable intelligent surfaces (RISs) are envisioned to become a promising technology for future wireless communication systems as they have the ability to provide cost-efficient solutions for the growing demand for network coverage and high data rates. A typical RIS acts as a smart reflector, capable of inducing a controllable phase shift to impinging electromagnetic (EM) waves at each of its unit cells. The key challenge in RIS-assisted networks is the efficient configuration of these unit cells with optimal phase shifts. This thesis provides a comprehensive overview of RIS technology and investigates the use of RISs for extending the coverage in wireless communication networks through codebook-based configuration of the RIS unit cells. In particular, this thesis is based on three main contributions, which are summarized below. First, a grant-free access scenario is studied, where devices located within a given coverage area try to access a base station (BS) via a RIS-assisted link. The goal is to determine the optimal phase-shift configuration for a RIS codebook that maximizes the probability of device activity detection at the BS for all potential device locations within the coverage area. To this end, a generalized likelihood ratio test (GLRT)-based detection scheme is proposed to derive the detection performance as a function of the RIS phase shifts and statistical channel state information (CSI). Subsequently, the problem statement is formulated as a min-max optimization problem, and locally optimal solutions are found by applying the majorization-minimization principle. Numerical results show that the proposed scheme outperforms existing phase-shift designs. Second, codebook-based RIS configuration by means of beam training is investigated, considering a RIS-assisted link between a BS and a mobile user traveling on an unknown trajectory within a given coverage area. In this scenario, the impact of RIS beam design on the fundamental trade-off between the achievable signal-to-noise ratio (SNR) and the beam training overhead is analyzed, taking into account various system parameters including RIS size, codebook size, coverage area size, number of pilot symbols, RIS response time, and user velocity. A key result is that the achievable SNR can be characterized by two different scaling laws, while the overhead for reliable beam training is either dependent on or independent of the SNR. Based on these findings, the considered trade-off is examined for full search (FS), hierarchical search (HS), and tracking-based search (TS) beam training using both analytical and numerical methods. While TS typically provides the best performance, it is shown that FS is an efficient scheme for large RISs and wide beam designs, which can outperform HS beam training. Third, codebook-based RIS configuration is analyzed for a self-sustainable RIS that integrates energy harvesting capabilities to operate independently of an external power supply. The goal is to determine the optimal splitting ratio for the incident signal at the RIS and the optimal power allocation at the BS. The proposed analytical model adopts a tile-based RIS architecture and takes different transmit strategies for beam training and data transmission into account. Moreover, for the common splitting schemes power splitting, element splitting, and time splitting, it is shown that the optimal splitting ratio at the RIS and the optimal power allocation at the BS can be efficiently determined by a one-dimensional grid search. Subsequent performance evaluations reveal that the considered power consumption model for the RIS and the number of tiles have a significant impact on the minimum transmit power at the BS. Notably, the results demonstrate that each of the considered splitting schemes can be the most efficient.