Heterogeneous Sensor-Based Multi-Robot Exploration for Radiation Source Localization in Unknown Environments
Abstract
Autonomous radiation source localization in unknown environments is an important problem in nuclear decommissioning and post-accident inspection.However, exploration based solely on directional sensors or dose-rate sensors has inherent limitations. Therefore, this paper proposes a heterogeneous sensor-based multi-robot exploration method that combines these complementary sensing modalities. A directional-sensor robot first generates coarse candidate source points using a von Mises mixture model, and multiple small robots equipped with dose-rate sensors refine them through local measurements based on Gaussian Process Regression. Candidate points are assigned by market-based task allocation, and D* Lite is used for navigation in unknown environments. In four simulation cases with two to four radiation sources, the proposed method detected all true sources. Compared with directional sensing alone, the cooperation with dose-rate-based local exploration reduced the average localization error from 0.796–2.127 m to 0.083–0.141 m. Compared with a coverage-based baseline, the use of directional sensing reduced the coverage required to detect all sources by 47.8%–92.0% and the total travel distance by 63.1%–96.3%.
