一种欠驱动AUV的三维实时避碰方法

针对以测距声纳为避碰传感器的一类欠驱动型AUV，提出了一种水平面和垂直面相结合的三维实时避碰方法。根据测距声纳和欠驱动AUV 的特殊性，首先从运动规划和路径规划2 个层次提出了AUV 混合型实时避碰结构，并分别设计了基于事件反馈监控的避碰自动机和基于免疫遗传的局部路径规划算法。多种典型障碍场景的半物理仿真实验表明，论文所提方法能够实现AUV 安全、稳定的三维避碰过程。

Aiming at an under-acturated autonomous underwater vehicle (AUV) with ranging sonar as its obstacle avoidance sensor, this paper presents a 3D real-time obstacle avoidance method combining horizontal plane with vertical plane. Firstly a hybrid real-time obstacle avoidance structure is designed from the view of motion planning and path planning levels based on the sonar’s and AUV’s particularities. Then an obstacle avoidance automata based on event feedback supervision and an immune genetic algorithm for generating local path are proposed respectively. A large number of tests are carried out on a semi-physical simulation platform. The simulations with diverse typical obstacle scenarios demonstrate that the presented method can make AUV avoid unknown obstacles safely and stably in 3D space.

国家自然科学基金资助项目（60805050）．