Efficient Path Planning for Fiber Sorting Manipulators Based on Improved Bi-RRT* in Narrow Environments

Automated foreign-fiber sorting in textile processing requires efficient robotic path planning, yet conventional sampling-based planners often struggle in narrow passages because of limited entrance accessibility and constrained tree expansion. To address this problem, this study proposes HA-Bi-RRT*, an improved Bi-RRT* planner incorporating a cooperative breakthrough mechanism, hybrid adaptive sampling, and hierarchical path refinement. When tree expansion is blocked, the cooperative mechanism guides the search along obstacle boundaries to facilitate passage-entrance localization. During passage traversal, the adaptive sampler enlarges the local sampling radius after consecutive failures to improve exploration around blocked regions. The method is evaluated in a two-dimensional benchmark, a three-dimensional narrow-passage environment, a ROS-based 5-DOF manipulator simulation, and a physical manipulator experiment. The observed results show favorable finite-budget trade-offs in planning success, runtime, and path cost relative to the selected baselines. The ROS simulation and physical experiments demonstrate the feasibility of converting the generated workspace paths into executable manipulator motions under the tested laboratory conditions.