基于改进PF-RRT*算法的果园复杂地形下移动机器人路径规划

Path Planning for Mobile Robots in Complex Terrain of Orchards Based on Improved PF-RRT* Algorithm

摘要:

为提高果园自主移动机器人在山区、丘陵等复杂地形环境下运动路径规划效果,增强机器人野外工作性能,针对果园非结构化布局和起伏地形,本文提出了一种基于改进PF-RRT*（Plane fitting-rapidly exploring random trees star）算法的果园机器人路径规划方法。利用自适应采样步长策略,在果园不同障碍物密度环境下能够灵活地调节随机树扩展步长。采用自适应目标偏置的势场引导采样方法,能够有效引导随机树避开障碍物并向目标点扩展。此外,在随机树扩展过程引入地形评估函数。通过平面拟合技术对地形坡度、稀疏度和粗糙度进行综合评估,确保规划的路径在复杂地形中可通过性和安全性。最后,采用三次样条插值与高斯过程回归对路径进行平滑和优化。在仿真环境中,改进PF-RRT*算法与RRT*、Q-RRT*和PF-RRT*算法相比,轨迹偏差最多减少46.97%,路径长度最多减少7.64%,规划时间最多减少23.58%,且成功率和避障效果均优于对比算法。真实果园行间试验结果表明,与RRT*、Q-RRT*和PF-RRT*算法相比,改进PF-RRT*算法与理想轨迹偏差量减少34.04%、26.19%、27.91%、路径长度减少6.58%、3.16%和4.10%、规划时间减少30.56%、18.25%和33.52%。表明本文算法在果园复杂地形下能够实现自主移动机器人最优路径规划与导航。

Abstract:

Aiming to improve the motion path planning performance of orchard autonomous mobile robots in complex terrain environments such as mountainous and hilly areas,and enhance their field operation performance,an improved PF-RRT* (plane fitting-rapidly exploring random trees star) algorithm for the unstructured layout and undulating terrain of orchards was proposed. An adaptive sampling step size strategy was adopted,which can flexibly adjust the expansion step of the random tree in orchard environments with different obstacle densities. An adaptive goal-biased potential field-guided sampling method was employed to effectively guide the random tree to avoid obstacles and expand toward the target point. In addition,a terrain evaluation function was introduced during the expansion of the random tree. The terrain slope,sparsity,and roughness were comprehensively evaluated through plane fitting technology,ensuring the passability and safety of the planned path in complex terrain. Finally,cubic spline interpolation and Gaussian process regression were used to smooth and optimize the path. In the simulation environment,compared with the RRT*,Q-RRT*,and original PF-RRT* algorithms,the proposed improved PF-RRT* algorithm reduced the trajectory deviation by 46.97%,the path length by 7.64%,and the planning time by 23.58%;meanwhile,its success rate and obstacle avoidance performance were superior to those of the comparison algorithms. Field tests in real orchard rows showed that,compared with the RRT*,Q-RRT*,and original PF-RRT* algorithms,the improved PF-RRT* algorithm reduced the deviation from the ideal trajectory by 34.04%,26.19%,and 27.91%,the path length by 6.58%,3.16%,and 4.10%,and the planning time by 30.56%,18.25%,and 33.52%,respectively. The results demonstrated that the proposed algorithm can achieve optimal path planning and navigation for autonomous mobile robots in complex orchard terrain.