Development and Validation of a Fully Autonomous Navigation Framework for Autonomous Underwater Vehicles: Improvements in Path Planning and Algorithm Optimization

Autonomous underwater vehicles (AUVs) with fully autonomous navigation are widely used in ocean exploration, resource development, national defense, and environmental monitoring. The ability to plan effective paths is a critical factor in the performance of an AUV. However, current path planning often encounters several issues, including slow convergence, redundancy, and lack of smoothness. Therefore, we construct a fully autonomous navigation framework for AUVs, with a particular focus on improving the path planning component. Specifically, we propose a new path planning algorithm, Underwater-RRT, which is an enhancement of the traditional Rapidly-exploring Random Tree (RRT) algorithm, incorporating goal-biased sampling to accelerate convergence. The proposed algorithm also reduces the length of the path by avoiding unnecessary intermediate nodes and uses a cubic B-spline path smoother to generate smooth trajectories that comply with the dynamic model of the AUVs, preventing abrupt changes in the angular velocity. We validated the effectiveness of Underwater-RRT on the HoloOcean underwater experimental platform in three-dimensional underwater scenarios, where it was shown to align with the actual motion trajectories of AUVs. Experimental results demonstrate that our Underwater-RRT establishes a new state-of-the-art performance in underwater path planning tasks, improving the convergence speed by 2.81 times and reducing the path length by 29.16%.