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

TL;DR

This work tackles single-view 3D reconstruction of occluded objects using wireless sensing enhanced by a Reconfigurable Intelligent Surface (RIS). The system employs a large RIS to improve spatial resolution and create reflection paths that bypass occluders, while a generative diffusion-based model completes the full shape from the RIS view. A learning-based error predictor maps RIS configurations to amodal reconstruction error, enabling gradient-based optimization of phase shifts under a discrete phase-set constraint. The approach combines compressed sensing for visible-shape recovery with occlusion-aware GAMP and conditional diffusion for shape completion, achieving up to $65.54\%$ and $56.73\%$ improvements over baselines in experiments on ShapeNet, demonstrating robust amodal sensing with RIS under occlusion.

Abstract

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.

Figures (4)

• Figure 1: System model of a RIS-aided wireless amodal sensing system.
• Figure 2: Illustration of the target object and the occlusion among voxels.
• Figure 3: The training and inference of the proposed model for RIS configuration optimization.
• Figure 4: (a) Reconstruction error of the proposed scheme and other sensing schemes. (b) Reconstruction results of a lamp under different schemes (N = 10×10×10). (c) Reconstruction error of our RIS configuration optimization scheme and the correlation minimization-based scheme.