Correction Redistribution Mechanism Based on Forward-Reverse Solutions and Real-Time Path Dynamic Adaptive Re-Planning for Multi-AUVs Collaborative Search

Addressing challenges like ocean currents, dynamic obstacles, collision avoidance, and computing limitations, a hybrid algorithm is proposed for multi-AUVs to search for random static or dynamic targets in a 3D marine environment and the environment is characterized by our Task Urgency Biological Neural Network (TUBNN) model, encapsulates factors such as currents, obstacles, and targets using neuronal activity representations. For optimizing task assignments in the face of unpredictable oceanic variables, a Double Layer Biological Self-organizing Map (DLBSOM) algorithm is proposed. Post assignment, the initial path, generated via DLBSOM, is refined by the end-to-end direct connection model to eliminate superfluous routes. Enhanced by the A* algorithm and Fermat’s spiral curve, our approach ensures paths align with AUV navigational constraints. Integrating real-time sonar data, our Distance-Speed Comprehensive Evaluation (DSCE) strategy, combined with the local dynamic neural network, facilitates the avoidance of unforeseen barriers, ensuring safe AUV navigation. The correction redistribution mechanism, leveraging live data and assessing AUV and target statuses, refines task assignment. We holistically evaluate environmental, cooperative, and AUV-specific constraints, with our algorithm’s efficacy validated through rigorous theoretical analysis, simulations, and field tests.