Online Path Planning for AUV in Dynamic Ocean Scenarios: A Lightweight Neural Dynamics Network Approach

In this study, the online path planning problem for autonomous underwater vehicle, which is constrained by the limited hardware computation and energy-carrying capabilities, is studied under the interference of dynamic ocean currents. To address this issue, an online lightweight neural dynamics approach is proposed to plan paths with low time and energy consumption in ocean currents scenarios. Firstly, the lightweight rapid propagation neural dynamics network, which involves the low complexity structure and the rapid propagation mechanism, is constructed. The proposed low connection-computation complexity neural dynamics network structure can reduce the number of adjacent neurons and the computation of neural connection weights by the customized division. The proposed rapid propagation mechanism can enhance the propagation directionality to speed up the convergence of the neural dynamics network. Then, the fused-ocean-currents path-generation mechanism is proposed to fuse the local adjacent ocean currents information into the neural activity values to reconstruct the neural activity value gradient, which can timely reflect the low time and energy consumption paths. In this way, the relatively advantageous ocean currents can be fully utilized and the relatively adverse ocean currents can be actively avoided to efficiently save the navigational time and energy. Furthermore, the event-trigger-based path-screening mechanism is proposed to adaptively avoid detecting unnecessary ocean currents information, thereby reducing the detecting energy consumption. Finally, the superior performance of the proposed approach is verified by the numerical results.