Path planning for mobile robots in unstructured orchard environments: An improved kinematically constrained bi-directional RRT approach

Abstract

In complex orchard environments, efficient path-planning algorithms for mobile robots are crucial for achieving accurate autonomous navigation and mobile operations. The existence of kinematic constraints for mobile robots in orchards and the unstructured layouts of orchard environments, which primarily feature irregular obstacles, pose considerable challenges for path-planning algorithms. Ideal path-planning algorithms must overcome the above problems and plan navigation paths that are safe, efficient, and consistent with the kinematic constraints of mobile robots in an obstacle-filled environment. In this study, to satisfy the above requirements, the Improved Kinematically Constrained Bi-directional Rapidly Exploring Random Tree (IKB-RRT) algorithm is proposed, which includes three innovations. First, an improved artificial potential field method is designed for guidance-node generation in complex orchard environments; hence, the nodes provide the path guidance for the random expansion trees that are subject to the kinematic constraints, avoiding obstacle regions and expanding towards the target point. Second, a motion-control parameter sampling method for the robot configuration is introduced to control the random expansion tree to generate path nodes and trajectories that satisfy the mobile robot constraints. Third, a path optimization strategy for dual-tree connected regions is proposed to avoid sudden changes in the heading angle at dual-tree connected waypoints and to improve the overall path smoothness. Path planning is performed in four orchard environments with different characteristics. The algorithm test results show that, in different scenarios, the proposed IKB-RRT algorithm reduces the path length by a maximum of 13.94% and 12.20% and the planning time by a maximum of 61.41% and 68.84%, significantly improving the overall smoothness of the path relative to the KB-RRT* and KB-RRT algorithms, respectively. The results fully verify that the IKB-RRT algorithm has obvious advantages in handling the path-planning problem that arises for mobile robots in complex orchard environments. The proposed algorithm can be implemented in future orchard mobile robots to enable effective path planning.