KBQ‐RRT*: A Smoothness‐Enhanced Kinematic Bidirectional Quick‐RRT* Via Dual‐Tree Optimization for Autonomous Navigation in Complex Orchards

Abstract

ABSTRACT The autonomous navigation of mobile robots in orchards faces significant challenges due to the unstructured nature of these environments, characterized by complex obstacles and narrow feasible areas. To address these challenges, this paper presents a novel kinematic‐constraint bidirectional quick (KBQ)‐RRT* path planning algorithm specifically designed for complex orchard environments. The algorithm innovatively integrates three key improvements: (i) a lightweight smooth‐expansion cost that penalizes waypoint heading changes under kinematic constraints, (ii) an efficient branch pruning strategy during iteration, and (iii) a dual‐tree smooth optimal connection strategy that removes T/V‐shaped junctions at bidirectional merges. The KBQ‐RRT* algorithm was extensively validated through experiments conducted in four orchard environments. The ancestor search depth was set to 3, the expansion step size to 0.2 m, and the optimization range to 3 m. The results of these experiments demonstrate that the KBQ‐RRT* algorithm improves performance, particularly in terms of computational efficiency. Compared with Quick‐RRT*, KBQ‐RRT* reduces the computational time by 1.41 s in four experimental scenarios, improves the path smoothness by 16.14%, and maintains a similar path length; relative to Bi‐RRT*, it increases the computational time by 0.094 s in four scenarios at a negligible computational cost, and improves the path smoothness by 23.46% while reducing the path length by 5.04%. The algorithm's practical effectiveness is further validated through extensive mobile robot path following experiments, confirming the excellent executability of the planned paths in real‐world orchard scenarios.