3DCKMDiff: 3D CKM Construction Based on Generative Diffusion for Environment-Aware Low-Altitude Communication and Sensing

With the rapid development of unmanned aerial vehicle (UAV) assisted low-altitude applications, there has emerged an increasing demand for environment-aware communication and sensing services in complex flight environment to guarantee the robustness and safety of UAV scheduling operations. Channel knowledge map (CKM), which provides location-specific channel knowledge prior, is becoming a promising technique to meet the above requirement. In the context of low-altitude environment, CKM construction needs to be extended from the conventional two-dimensional (2D) plane to three-dimensional (3D) space, which poses a new challenge of a complete 3D CKM based on sparsely sampled data from different slices in the flight region. In this paper, we formulate this problem as a 3D inverse problem and propose 3DCKMDiff via 2D generative diffusion model by learning the joint distribution of multiple slices to gain the 3D prior knowledge. Simulation results demonstrate that compared with benchmark methods, the proposed 3DCKMDiff achieves significant improvements in both CKM construction accuracy and robustness.