The path-planning problem for unmanned underwater vehicles (UUVs) in long-endurance deep-sea missions is investigated, with a focus on the trade-off between navigation safety and energy efficiency. A novel AGB-RRT* algorithm is proposed to address these challenges within complex 3D environments. By leveraging the passive hydrostatic restoring characteristics of UUVs, a simplified 4-DOF planning model is established, and an energy-evaluation function is subsequently derived from fluid resistance theory. A terrain-guided “logical grid” heuristic is designed to enhance search efficiency. A probabilistic hybrid sampling strategy is further introduced to balance gradient-guided expansion with random exploration, which accelerates convergence while effectively avoiding local minima. The cost function integrates a flowadaptive mechanism to prioritize energy-saving trajectories, while 3rd-order B-spline interpolation is applied to ensure curvature-continuous paths that satisfy kinematic requirements. The effectiveness of the AGB-RRT* algorithm is validated through simulations using real seabed bathymetry and vortex models. Results demonstrate that the proposed method significantly reduces both energy consumption and path length compared to standard RRT* and Informed-RRT*, providing a robust and practical solution for the autonomous navigation of UUVs in challenging underwater environments.

unmanned underwater vehicles, path planning, anisotropic sampling, flow-adaptive mechanism, 3rd-order B-spline smoothing