Channel-Aware OTA Diagnosis of a Defective Cluster in RIS via Noisy Twenty Questions

Reconfigurable intelligent surfaces (RIS) are envisioned as a key enabler for 6 G networks; however, they are susceptible to clustered hardware impairments. Traditional circuit-level element-wise diagnosis approaches are invasive and infeasible for RIS with a large number of elements. In this paper, we propose a channel-aware over-the-air (OTA) diagnosis framework that localizes the boundaries of a contiguous cluster of defective RIS elements using one-bit feedback. Specifically, we first propose a dual-slot pilot transmission protocol to extract the responses of defective elements from the measurement signals. Subsequently, we formulate the boundary localization problem as a noisy twenty questions problem, and develop a dyadic posterior matching (DyaPM) algorithm to solve the resultant problem with logarithmic complexity. Furthermore, we incorporate an SNR-dependent probability of lie into the DyaPM algorithm to improve belief updates, which is in contrast to prior works that rely on a fixed probability of lie. Simulation results demonstrate that the proposed diagnosis framework enables a correct detection probability exceeding 90 % even at an SNR of -15 d B and asymptotically approaches the efficiency of the noiseless bisection search as the SNR improves.