Radiation source estimation using a connectivity-aware decentralized multi-mobile sensor system

This paper proposes a cooperative search method utilizing multiple mobile sensors to estimate the characteristics of unknown radiation source. Low-level radiation signals are sparse, intermittent, and highly time-varying due to environmental refraction and sensor noise, making source estimation a significant challenge. To address these stochastic noise conditions, a particle filter, a nonlinear estimation technique, is employed for source term estimation. Furthermore, to enable successful exploration in environments lacking wireless communication infrastructure, each distributed agent estimates the overall communication connectivity and conducts the search within its communication range. To achieve this, the Laplacian matrix from graph theory is utilized to estimate communication connectivity. Additionally, to generate efficient search trajectories, each agent applies Shannon entropy, an information-theoretic measure, to guide the search process effectively. The performance of the proposed approach has been validated through extensive ablation studies comparing scenarios with and without communication range constraints, as well as the presence and absence of communication estimation. The results demonstrate that estimating communication connectivity significantly enhances the search success rate to 94% and reduces the search time from 180.8 to 156.0 under communication constraints in the best case. Moreover, it has been observed that search performance improves as the number of mobile agents increases.