Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Diffusion Model-Enhanced Environment Reconstruction in ISAC

Nguyen Duc Minh Quang, Chang Liu, Shuangyang Li, Hoai-Nam Vu, Derrick Wing Kwan Ng, Wei Xiang

TL;DR

This work tackles the challenge of sparse and noisy environment reconstruction in ISAC systems. It introduces NSADM, a diffusion-model framework with a CRB-guided forward process and a conditioned reverse process that incorporate noise variance and detection probabilities to denoise and densify the distance matrix. Experimental results on synthetic ISAC data show that NSADM outperforms model-based and learning-based baselines in RMSE and Chamfer Distance, yielding denser and more structurally coherent point clouds. The approach offers a practical pathway to robust ER in low-SNR ISAC deployments and highlights the value of integrating physical noise models into diffusion-based refinement.

Abstract

Recently, environment reconstruction (ER) in integrated sensing and communication (ISAC) systems has emerged as a promising approach for achieving high-resolution environmental perception. However, the initial results obtained from ISAC systems are coarse and often unsatisfactory due to the high sparsity of the point clouds and significant noise variance. To address this problem, we propose a noise-sparsity-aware diffusion model (NSADM) post-processing framework. Leveraging the powerful data recovery capabilities of diffusion models, the proposed scheme exploits spatial features and the additive nature of noise to enhance point cloud density and denoise the initial input. Simulation results demonstrate that the proposed method significantly outperforms existing model-based and deep learning-based approaches in terms of Chamfer distance and root mean square error.

Diffusion Model-Enhanced Environment Reconstruction in ISAC

TL;DR

This work tackles the challenge of sparse and noisy environment reconstruction in ISAC systems. It introduces NSADM, a diffusion-model framework with a CRB-guided forward process and a conditioned reverse process that incorporate noise variance and detection probabilities to denoise and densify the distance matrix. Experimental results on synthetic ISAC data show that NSADM outperforms model-based and learning-based baselines in RMSE and Chamfer Distance, yielding denser and more structurally coherent point clouds. The approach offers a practical pathway to robust ER in low-SNR ISAC deployments and highlights the value of integrating physical noise models into diffusion-based refinement.

Abstract

Recently, environment reconstruction (ER) in integrated sensing and communication (ISAC) systems has emerged as a promising approach for achieving high-resolution environmental perception. However, the initial results obtained from ISAC systems are coarse and often unsatisfactory due to the high sparsity of the point clouds and significant noise variance. To address this problem, we propose a noise-sparsity-aware diffusion model (NSADM) post-processing framework. Leveraging the powerful data recovery capabilities of diffusion models, the proposed scheme exploits spatial features and the additive nature of noise to enhance point cloud density and denoise the initial input. Simulation results demonstrate that the proposed method significantly outperforms existing model-based and deep learning-based approaches in terms of Chamfer distance and root mean square error.

Paper Structure

This paper contains 14 sections, 17 equations, 5 figures.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Signal Model
  4. Environment Reconstruction Model
  5. Problem Formulation
  6. NSADM: Noise-Sparsity-Aware Diffusion Model-based Point Cloud Enhancement
  7. Noise-Sparsity-Aware Forward Process
  8. Noise-Sparsity-Conditioned Reverse Process
  9. Objective Function
  10. Training and Inference
  11. Experimental Results
  12. Dataset Generation and Benchmark Selection
  13. Performance Evaluation
  14. Conclusion

Figures (5)

  • Figure 1: An illustration of the ISAC ER scenario.
  • Figure 2: Overview of the proposed NSADM framework.
  • Figure 3:
  • Figure 4: Performance comparison of NSADM, DPIR, DnCNN, and MT in terms of (a) average RMSE and (b) average CD across varying transmit powers.
  • Figure 5: Trade-off between fidelity and perceptual quality. Varying perceptual loss weight $\lambda$ in the hybrid loss function affects average RMSE and CD.