Development of optimization-based algorithms for radiation source localization in nuclear emergencies

Abstract

Accurate radiation source localization is a fundamental requirement for nuclear emergency response, environmental radiation monitoring, radioactive waste management, and homeland security. However, reliable localization remains challenging because of measurement uncertainty, background radiation, sensor-network limitations, and the nonlinear nature of radiation propagation. This work presents an integrated optimization-based framework for radiation source localization that combines experimentally characterized gamma-ray measurements with physically representative radiation modeling and intelligent optimization. High-purity germanium (HPGe) measurements of Bi-207 and Ti-44 radioactive sources were first employed to characterize detector response, energy calibration, and radiation statistics. These experimentally derived parameters were subsequently incorporated into a simulation framework to establish realistic localization scenarios under different signal-to-noise ratios (5–20 dB), radiation intensities, and sensor-network configurations. Radiation source localization was formulated as a nonlinear optimization problem and solved using Particle Swarm Optimization (PSO) and Biogeography-Based Optimization (BBO). Performance was quantitatively evaluated using localization error, root mean square error (RMSE), convergence behavior, localization accuracy, and computational efficiency. PSO achieved an RMSE of 0.85 ± 0.04 m over 30 independent runs, whereas BBO obtained 0.90 ± 0.06 m. A paired statistical analysis confirmed that the observed difference was statistically significant (p < 0.05). The results demonstrate a progressive reduction in RMSE from approximately 0.58 at 5 dB to nearly 0.10 at 20 dB, confirming the strong influence of measurement quality on localization performance. The computational efficiency analysis showed that PSO completed the localization process in 12.5 s, whereas BBO required 14.2 s under the same simulation conditions, demonstrating the lower computational overhead of PSO while both algorithms maintained reliable localization performance. Statistical analyses further indicate that the two algorithms provide comparable localization accuracy, with each exhibiting advantages under different operating conditions. The principal contribution of this work is the development of a unified experimental–simulation optimization framework that bridges experimentally characterized radiation measurements with optimization-based source localization, providing a reproducible methodology for evaluating localization accuracy, computational efficiency, and robustness under realistic nuclear emergency conditions. The proposed framework offers a practical foundation for intelligent radiation monitoring systems and autonomous robotic platforms employed in future nuclear safety and emergency-response applications.