Uncertainty aware multi-agent active radio map estimation

Radio maps are a key element in understanding the electromagnetic environment. For scenarios involving multi-agent cooperative monitoring and constructing radio map, an uncertainty-aware sampling trajectory planning cooperative control method is proposed, encompassing three major modules: map estimation, region exploration, and neighborhood exploration. First, Gaussian process regression is employed to periodically reconstruct the radio map, with the posterior variance quantifying uncertainty to provide a basis for trajectory planning. Second, Voronoi diagrams are utilized to dynamically allocate exploration windows, and a centroid detection and evaluation method for uncertain regions is designed to guide agents in prioritizing the exploration of high-uncertainty subregions. Finally, upon reaching the centroid location, the agent employs a greedy measurement-based search method, incorporating the spatial distribution characteristics of historical sampling points, to prioritize sampling in low-density neighborhood spaces. Simulation results demonstrate that the proposed algorithm can collect high-quality data within a short time and achieves improvements in both map exploration time and estimation accuracy.