Optimization of Reconfigurable Intelligent Surfaces Through Trace Maximization

Abstract

Reconfigurable Intelligent Surfaces (RIS) have received significant attention recently as an innovation for enhanced connectivity, capacity, and energy efficiency in future wireless networks. Recent works indicate that such RIS-augmented communications can significantly enhance performance by intelligently shaping the characteristics of the multipath propagation environment to focus the energy in a desired direction and to circumvent impediments such as blockage, especially for communication at millimeter-wave (mmW), Terahertz (THz) and higher frequencies. In this paper, we investigate optimized (amplitude and phase) RIS design in a point-to-point multipath MIMO link and study the impact on link capacity under the assumption of perfect channel state information at the transmitter (TX), receiver (RX) and RIS. Specifically, we propose RIS design based on the maximization of the trace of the composite TX-RIS-RX link matrix which is a measure of the average power at the RX. We propose two RIS designs: a diagonal RIS matrix, and a general RIS matrix representing a more advanced architecture. The optimum design, in both cases, corresponds to calculating the dominant eigenvector of certain Hermitian matrices induced by the component channel matrices. We illustrate the capacity performance of the optimized RIS designs and compare them to a baseline design (random amplitudes and phases) and a recently proposed low-complexity phase-only design. We present results for sparse and rich multipath, and also consider the impact of line-of-sight paths. Our results show that while all designs offer comparable capacity at high signal-to-noise ratios (SNRs), the proposed optimum designs offer substantial gains at lower SNRs.