【摘要】 针对传统渐近最优快速扩展随机树算法(RRT~*)随机性大、收敛精度低以及运行时间长等问题，提出一种改进的Informed-RRT~*FN算法。改进算法在找到初始路径前采用基于贪心思想改进的目标偏置采样策略和随机删除叶子节点策略以降低找到初始路径的时间；找到初始路径后进一步在椭圆子空间中采样，使用基于节点权重的节点删除策略优先删除“无效”节点和动态重连半径的策略以提高收敛精度并保持较低的运行时间。改进算法在3种地图开展了仿真实验，结果表明相较于RRT~*FN、Informed-RRT~*和Informed-RRT~*FN算法，该算法收敛精度最高，且运行时间最短。该算法进一步在ROS平台开展全局路径规划实验，验证了其可靠性和实用性。更多还原

【Abstract】 Aiming at the problems of the traditional asymptotically optimal rapidly extended random tree algorithm(RRT~*) with large randomness, low convergence accuracy and long running time, an improved Informed-RRT~*FN algorithm is proposed.It uses an improved goal biased sampling strategy based on greedy ideas and random node deletion strategy before finding the initial path to reduce the time of finding the initial path.After finding the initial path, it further samples in the elliptic subspace, using the node deletion strategy with node weight which in order to remove the "invalid" nodes firstly and the strategy of dynamic reconnection radius to improve convergence accuracy and keep running time low.The improved algorithm is simulated on three maps, and the results show that compared with the RRT~*FN,Informed-RRT~* and Informed-RRT~*FN algorithms, it has the highest convergence accuracy and the shortest running time.The algorithm is further carried out on the global path planning experiment on the ROS platform to verify its reliability and practicability.更多还原