Abstract: In the face of escalating nuclear safety concerns, especially during nuclear emergency response and decommissioning, precise and efficient nuclear pollution situation awareness is crucial. Traditional nuclear radiation detection methods, like full-coverage scanning, are flawed. They consume excessive time, taking hours or days for large-area scans, thus delaying emergency responses and heightening radiation exposure risks. Additionally, they lack environmental and geographical information integration, impeding comprehensive assessment of nuclear pollution’s ecological impact. Their low-intelligence nature also renders them unsuitable for complex and dynamic environments. Our study aims to develop an advanced method to surmount these limitations. A novel approach relying on particle filter source seeking, with the goal of swiftly and accurately mapping nuclear radiation situations, was presented. To this end, a sophisticated robot platform integrating autonomous navigation and radiation detection was constructed. Equipped with LiDAR and cameras, the robot can navigate diverse terrains and avoid obstacles. A unique detection mechanism was devised to precisely feedback the angle of arrival of nuclear radiation, which is vital for subsequent data processing. The angle-of-arrival information into the particle filter algorithm was then introduced, effectively constraining the estimation area. This significantly enhanced detection efficiency and source seeking accuracy. By incorporating source seeking estimation results into the Gaussian-model-based interpolation process and leveraging the particle filter’s advantages, measurement data were reduced, enabling rapid generation of high-precision radiation situation maps. In experiments, cross-evaluation indicators from measurement and training sets were used. The mean absolute percentage error was applied to analyze map quality. Extensive simulations in various virtual scenarios and real-world radioactive source search tests in controlled environments verified the method’s effectiveness. The generated maps had a far lower mean absolute percentage error than traditional methods, highlighting its high accuracy. In summary, our particle-filter-based source seeking method offers a more efficient, accurate, and intelligent solution for nuclear pollution situation assessment, holding great promise for nuclear emergency and decommissioning applications.