Nuclear Pollution Control Evolutionary Algorithm v2.0: Dynamic Tracking Based on Multi-source Joint Particle Filter

To address the escalating nuclear waste pollution amid extreme weather, this study develops the Nuclear Pollution Control Evolutionary Algorithm v2.0 (purify_enterprise_ultimate_v2), achieving a qualitative leap from static monitoring to dynamic probabilistic tracking. The core breakthrough lies in the integration of multi-source joint particle filter technology, which initializes 2000 particles to simultaneously track up to 2 radioactive sources, encoding position, intensity, and decay rate into the particle swarm for comprehensive coverage of potential pollution scenarios. Enabled by Numba JIT hardware acceleration, the algorithm realizes millisecond-level real-time response and centimeter-level precise positioning even under high background noise. Key functionalities include multi-modal data fusion (AFM and spectrum data), DarkCandidate-based anomaly detection, adaptive robot scheduling, and stable particle state update (Effective Sample Size, ESS = 2000.0). Compared with Version 1.0, it solves industry pain points such as dynamic source term evolution and Gaussian noise interference, supporting multi-target joint tracking and predictive governance of pollution diffusion. The algorithm completes full-process self-inspection and HIL testing within 22.5 seconds, maintains an optimal aggregated risk score of 10.0 in complex mixed nuclide scenarios, and provides precise decision support for nuclear pollution control, marking a paradigm shift from passive response to active prediction.