To address the low efficiency of existing Unmanned Underwater Vehicles (UUVs) in autonomous obstacle avoidance within complex marine environments,this paper proposes a three-dimensional (3D) obstacle avoidance algorithm that integrates the Velocity Obstacle (VO) method and the Dynamic Window Approach (DWA).First,a collision cone and a polar coordinate system are introduced,and a corresponding collision detection method is developed.This extends both the VO and DWA methods,originally designed for two-dimensional space,into three dimensions.Second,by defining the range of a detection sphere around the UUV,obstacles entering this region are classified as dynamic or static.Collision detection is then performed using the collision cone and polar coordinate motion models,respectively,to generate a set of feasible trajectories.Finally,an evaluation function selects the optimal path from these candidates.This process iterates in real time to enable efficient obstacle avoidance.Three-dimensional simulation experiments demonstrate that the UUV using this fusion algorithm successfully avoids all collisions.In a fully dynamic obstacle scenario,the proposed method reduces the trajectory length by 5.78% compared to the pure VO method and shortens the task completion time by 16.91% compared to the pure DWA method,while also requiring less computational runtime.The proposed algorithm offers valuable insights for enhancing the autonomous obstacle avoidance capability of UUVs in complex underwater environments.