Cooperative search and rescue for deep-sea submersibles with multi-source localization, probabilistic prediction, and smooth path planning

Search and rescue (SAR) of deep-sea submersibles is a critical challenge in marine safety, hindered by inaccurate localization, poor prediction, and inefficient search strategies. This study develops an efficient cooperative framework to enhance SAR success rates. An integrated framework is proposed, combining multi-source localization, probabilistic prediction, and intelligent planning. Its core includes three components: (1) A multi-sensor and multi-mode integrated localization model to accurately determine the disabled submersible's initial pose; (2) A Markov Chain Monte Carlo-based probabilistic prediction model simulating the time-varying positional distribution of the submersible under power-loss and complete-failure modes, considering ocean current disturbances; (3) A novel Dynamic-guided Multi-strategy Elite Ant Colony Optimization (DMS-EACO) algorithm considering smoothness to solve the 3D SAR path planning problem. This algorithm improves search efficiency and path smoothness via a Sigmoid decay factor, dynamic guidance mechanism, and turning heuristic function. Moreover, sensitivity analysis evaluates the submersible's dynamic behavior under propulsion failure scenarios. Simulation results show that compared to mainstream optimization algorithms, the proposed algorithm reduces the optimal path length by 19.6% to 77.6% and improves convergence speed by over 54% in two typical failure scenarios, generating significantly smoother trajectories. • Integrates 3 key technologies to build an end-to-end search-rescue framework. • Designs a positioning and navigation model via multi-sensor fusion and multi-mode. • Proposes an MCMC-based position prediction model, covering 2 scenarios. • Proposes DMS-EACO for fast and low-energy coverage of high-probability regions.