RIS-Aided Wireless Amodal Sensing for Single-View 3D Reconstruction

Amodal sensing is critical for various real-world sensing applications because it can recover the complete shapes of partially occluded objects in complex environments. Among various amodal sensing paradigms, wireless amodal sensing is a potential solution due to its advantages of environmental robustness, privacy preservation, and low cost. However, the sensing data obtained by wireless system is sparse for shape reconstruction because of the low spatial resolution, and this issue is further intensified in complex environments with occlusion. To address this issue, we propose a Reconfigurable Intelligent Surface (RIS)-aided wireless amodal sensing scheme that leverages a large-scale RIS to enhance the spatial resolution and create reflection paths that can bypass the obstacles. A generative learning model is also employed to reconstruct the complete shape based on the sensing data captured from the viewpoint of the RIS. In such a system, it is challenging to optimize the RIS phase shifts because the relationship between RIS phase shifts and amodal sensing accuracy is complex and the closed-form expression is unknown. To tackle this challenge, we develop an error prediction model that learns the mapping from RIS phase shifts to amodal sensing accuracy, and optimizes RIS phase shifts based on this mapping. Experimental results on the benchmark dataset show that our method achieves at least a 56.73% reduction in reconstruction error compared to conventional schemes under the same number of RIS configurations.

💡 Research Summary

The paper tackles the problem of reconstructing the complete 3‑D shape of partially occluded objects using wireless signals, a task traditionally hampered by the low spatial resolution of radio frequency (RF) measurements and severe data sparsity when obstacles block propagation paths. To overcome these limitations, the authors propose a novel framework that integrates a large‑scale Reconfigurable Intelligent Surface (RIS) with a generative learning model, thereby enhancing both the sensing resolution and the ability to bypass occlusions.

System architecture. A single‑antenna transmitter (Tx) illuminates a vertically deployed RIS composed of M sub‑wavelength elements, each capable of applying a discrete phase shift from a set of 2^b values. The RIS reflects the incident wave toward a Region of Interest (ROI) that contains the target object, which is discretized into N = N_x × N_y × N_z voxels. Each voxel n is characterized by a scattering coefficient ω_n and a binary occupancy indicator χ_n. The presence of other objects creates blockage; a binary matrix V (M × N) records whether the path from RIS element m to voxel n (and subsequently to the receiver) is obstructed. Only unblocked voxels contribute to the measured channel.

Measurement model. The composite channel from Tx → RIS → ROI → Rx is expressed as
h_ROI(ω, q, V) = Σ_{m=1}^M Σ_{n=1}^N … e^{−jϕ_m} V_{m,n} ω_n,
where q =