Scholarlink

<p>Channel knowledge map (CKM) is a promising technology for enabling environment-aware wireless communications and sensing by providing a priori location-specific channel information. One fundamental problem lies in constructing high-quality and complete CKMs for all locations of interest from limited and noisy on-site channel measurements, which constitutes a long-standing ill-posed inverse problem. By utilizing the recent advances of solving inverse problems with learned priors using generative artificial intelligence (AI), this paper proposes CKMDiff, a conditional diffusion-based generative framework for CKM construction. CKMDiff supports multiple CKM reconstruction tasks, including denoising, inpainting, and super-resolution. In addition, a wireless semantic-guided feature augmentation strategy is introduced to enhance the model’s ability to capture implicit relationships between electromagnetic propagation characteristics and spatial-geometric structures. Experimental results on the CKMImageNet and RadioMapSeer datasets demonstrate that the proposed CKMDiff achieves superior performance compared with existing benchmark methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CKMDiff: a generative diffusion model for CKM construction via inverse problems with learned priors

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Shen Fu,
  • Yong Zeng,
  • Zijian Wu,
  • Di Wu,
  • Jie Xu,
  • Shi Jin,
  • Cheng-Xiang Wang,
  • Xiqi Gao

摘要

Channel knowledge map (CKM) is a promising technology for enabling environment-aware wireless communications and sensing by providing a priori location-specific channel information. One fundamental problem lies in constructing high-quality and complete CKMs for all locations of interest from limited and noisy on-site channel measurements, which constitutes a long-standing ill-posed inverse problem. By utilizing the recent advances of solving inverse problems with learned priors using generative artificial intelligence (AI), this paper proposes CKMDiff, a conditional diffusion-based generative framework for CKM construction. CKMDiff supports multiple CKM reconstruction tasks, including denoising, inpainting, and super-resolution. In addition, a wireless semantic-guided feature augmentation strategy is introduced to enhance the model’s ability to capture implicit relationships between electromagnetic propagation characteristics and spatial-geometric structures. Experimental results on the CKMImageNet and RadioMapSeer datasets demonstrate that the proposed CKMDiff achieves superior performance compared with existing benchmark methods.