Robust NLoS Sensing for RIS-Assisted FMCW Radar With Discrete Phase Constraints and Imperfect CSI

Abstract

This article investigates robust sensing for reconfigurable intelligent surface (RIS)-assisted frequency-modulated continuous-wave (FMCW) radar operating under nonline-of-sight (NLoS) conditions. Practical performance is fundamentally constrained by finite-resolution RIS phase quantization and imperfect cascaded channel state information (CSI), which jointly degrade coherent combining over the transceiver (TR)–RIS–target–RIS–TR link. To address these nonidealities, we develop a statistical sensing framework with explicit channel-uncertainty modeling and derive a closed-form equivalent sensing-gain metric based on second-order error statistics, providing an analytically tractable objective for robust design. Based on this formulation, we propose an alternating-optimization algorithm with discrete dimensionwise sinusoidal maximization (AO–DDSM) to jointly optimize transmit beamforming and RIS phases over a finite codebook, ensuring monotonic objective improvement and finite-step convergence. Numerical results demonstrate that AO–DDSM significantly enhances range–Doppler focusing and detection performance compared with existing designs that assume continuous RIS phases, perfect CSI, or both, while maintaining strong robustness across different RIS sizes, phase resolutions, and CSI uncertainty levels.