基于改进RRT*算法的菠萝采收机导航路径规划

为了提高菠萝收获的机械化和自动化水平,该研究以菠萝采收机为研究对象,采用改进RRT*（Rapidly-exploring Random Trees Star）算法进行全局路径规划。首先在产生随机采样点时引入自启发式思想约束采样点的生成,在拓展新节点时借鉴人工势场引入方向权重对新节点拓展方向进行约束,同时计算权重wg合适的取值范围,采双向拓展加快迭代速度,后利用贪心算法修剪路径的冗余节点,并利用Cantmull-Rom插值函数对路径进行平滑处理。根据农田道路存在的复杂情况创建多障碍物、迷宫和狭窄通道3种仿真环境,分别对比RRT*算法、双向RRT*算法和改进后RRT*算法的性能进行测试。试验结果表明：3种环境下,本文算法的平均收敛时间是RRT*算法的17.97%,是双向RRT*算法的46.12%,平均规划速度是RRT*算法的4.7倍,是双向RRT*算法的2.2倍左右,平均拓展的节点数量比 RRT* 算法减少87.22%,比双向 RRT* 算法减少 52.52%,平均路径长度比 RRT* 算法减少 3.81%,比双向 RRT* 算法减少 6.08%。田间试验结果表明：本文算法的规划时间仅为RRT*算法的14.12%,为双向RRT*的20.34%,迭代次数比RRT*算法减少80.89%,比双向RRT*减少69.70%。另外,RRT*和双向RRT*算法规划出的路径上大于60°的转角分别是本文算法的1.56和2.06倍,大于100°的转角分别是本文算法的1.55和2.18倍,本文算法规划的路径更平滑。改进RRT*算法在农田里规划的路径符合菠萝采收机的路径导航需求,研究结果可为菠萝采收机的导航研究提供参考。

Navigation path planning of the pineapple harvester based on improved RRT* algorithm

In China, all pineapples are harvested manually. But the agricultural labour force is aging more and more seriously. Automatic navigation is one important direction for the development of pineapple harvest machines. In order to improve the mechanization and automation level of pineapple harvesting, this study proposed a path planning algorithm as navigation scheme for a pineapple harvester. An improved RRT* algorithm was used as the planning algorithm for global path planning. Firstly, the self-heuristic idea was used to constrain the generation range of sampling points. Then, the bias probability pbias was introduced in generating random sampling points. Sampling points were randomly generated with probability p in the space, when p>pbias. Otherwise, the target point was used as the sampling point so as to decrease the blindness of sampling point generation. Thirdly, the gravitational field idea of the artificial potential field algorithm, and the concept of direction weight were introduced in new node expansion. The weights wg and wk were assigned to the directions of the sampling point and the target point respectively, and the direction of expansion of the new node was constrained by those two weights. Fourthly, bidirectional expansion was used to speed up the iteration speed by referring the idea of double-tree expansion. Finally, the greedy algorithm was applied to prune the redundant nodes of the path, and the Cantmull-Rom interpolation function was used to smooth the path corners. Three environments, including multiple obstacles, mazes and narrow passages, were created to simulate the path planning process and to compare the performance among the improved navigation path planning algorithm, RRT* algorithm and bidirectional RRT* algorithm. Planning time, node number and path length were selected as the comparison indicators. Each algorithm was experimented 30 times in every single environment. The average, maximum, minimum and standard deviation of the simulation data of the three indicators were calculated respectively. The simulation results showed that the average planning time of the proposed algorithm in the three environments was 17.97% that of the RRT* algorithm, 46.12% that of the bidirectional RRT* algorithm, which mean that its average programming speed was 4.7 times as quick as that of the RRT* algorithm, about 2.2 times as quick as that of the bidirectional RRT* algorithm. The average of node number of the proposed algorithm was 87.22% less than that of the RRT* algorithm and was 52.52% less than the bidirectional RRT* algorithm. The average path length of the proposed algorithm was 3.81% less than the RRT* algorithm and was 6.08% less than the bidirectional RRT* algorithm. The field test results showed that the planning time of the proposed algorithm was only 14.12% that of the RRT* algorithm and was 20.34% that of the bidirectional RRT*. The iteration number of the proposed algorithm was 80.89% less than that of the RRT* algorithm, and was 69.70% less than that of the bidirectional RRT*. In addition, the rotation angles of bigger than 60° on the path planned by RRT* and bidirectional RRT* algorithms were 1.56 times and 2.06 times as much as that of the proposed algorithm, respectively, and the rotation angles of bigger than 100°on the path were 1.55 times and 2.18 times as much as that of the proposed algorithm, respectively. The improved RRT* algorithm planed the path in the field which can meet the path navigation requirements of agricultural machinery. As moving with speed of 0.2, 0.4, and 0.6m/s, the pineapple harvester can run along the planned path to the target point, but the position deviation and heading deviation increased with moving speed. This study can provide a reference for the navigation study of pineapple harvesters and other agricultural machines.