八叉树
计算机科学
运动规划
路径(计算)
人工智能
计算机视觉
计算机网络
机器人
作者
Wenna Wang,Huaming Qian,Shuai Feng
标识
DOI:10.1109/tiv.2024.3367771
摘要
Unmanned aerial vehicle (UAV) path planning has a wide range of applications, and it constitutes a key element of autonomous aerial navigation. The intricate arrangement and irregular nature of unstructured spaces can cause a significant computational burden on UAV path planning. This paper proposes a three-dimensional (3D) A* algorithm of UAV path planning incorporating octree overlap detection and real-time avoidance strategies based on simplified unstructured mesh. A mesh simplification method of feature-extraction based quadratic error metrics (QEM) is proposed to reduce the constructing time of the navigation space octree, combined with a novel constructed process of unstructured terrain. Furthermore, the cost function of the 3D A* algorithm is optimized to solve the problem of unstructured environments with many path inflection points and high path cost. The resulting paths are combined with Catmull-Rom spline curves to smooth the path through the initial positioning points. An obstacle avoidance strategy is proposed in conjunction with unstructured terrain, where the speed is planned by a spherical linear interpolation method to achieve dynamic obstacle avoidance. Simulation results demonstrate that the proposed method reduces the flight path cost and the number of inflection points, and achieves dynamic obstacle avoidance in real time.
科研通智能强力驱动
Strongly Powered by AbleSci AI