Authors

Abstract

Irregular branches in agricultural picking environments require rapid collision-free path planning and smooth manipulator motion. This study developed an improved RRT algorithm guided by grid numbering, termed Opti-mized-RRT, based on a unified indexed-set representation of numbered primary and backup free-grid regions. Within this framework, dynamic region adjustment, target biasing, node occupancy, and local burst expansion jointly guide the search, while greedy pruning and quintic polynomial interpolation are applied to post-process the resulting path. Unlike geometric-region-based methods that generate coordinate samples within an updated sampling window, Optimized-RRT selects a valid grid number from the active set and maps it directly to the corresponding state. Compared with RRT, GB-RRT*, and Informed-RRT*, the complete algorithm reduced planning time, the number of iterations, and the final number of path nodes by 40.61%–94.32%, 57.51%–88.66%, and 89.29%–94.59%, respectively, in two-dimensional scenarios, and by 88.95%–98.92%, 79.26%–94.67%, and 67.31%–90.83%, respectively, in three-dimensional scenarios. The reduction in the number of nodes resulted from the combined effects of guided search and path pruning. Manipulator simulations and a demonstration on a single experimental platform confirmed the executability of the generated trajectories under the tested conditions.

Abstract in Chinese

农业采摘环境中复杂枝条分布要求机械臂快速生成无碰路径并平稳运动。本文提出栅格编号引导的改进RRT 算法，以编号栅格集合表示主、备用采样区域，并融合动态区域调整、目标偏置、节点占用和局部突发扩展，随后采用贪心剪枝与五次多项式插值优化路径。与 RRT、GB-RRT* 和 Informed-RRT* 相比，二维规划时间、迭代次数和节点数分别降低 40.61%–94.32%、57.51%–88.66% 和 89.29%–94.59%，三维分别降低 88.95%–98.92%、79.26%–94.67% 和 67.31%–90.83%。机械臂仿真与平台试验验证了该方法的可行性。