Fruit grasping evaluation based on a humanoid underactuated manipulator and an adaptive grasping algorithm

Abstract

Grasping adaptability and stability are critical performance metrics for humanoid underactuated manipulators. Conventional fruit grasping robots and control algorithms lack generalizability as they are predefined for a specific type of fruit. In this study, a humanoid underactuated fruit grasping robot was designed, accompanied by the proposal of an adaptive grasping (AG-H) algorithm. The robot features a humanoid structure and is controlled by a rope-driven mechanism, enabling adaptability to diverse fruit grasping scenarios. The AG-H algorithm emulates human grasping logic, incorporating comprehensive evaluations of finger quantity, grasping angle, and depth parameters for determining optimal grasping positions based on fruit point cloud targets. Theoretical modeling calculations were rigorously verified and seamlessly integrated with practical AG-H algorithm computations in simulation experiments. The AG-H algorithm demonstrates good grasping performance under uncertain fruit poses. Comparative analyses were conducted between the AG-H algorithm and conventional grasping methods, revealing a 94.75% average grasping success rate and 3.53-second average cycle time achieved by the proposed system. Experimental validation was performed on a dedicated grasping test bed, confirming the manipulator's capability for accurate fruit grasping across arbitrary orientations. The proposed humanoid underactuated robotic grasping method demonstrates promising results, providing new solutions for efficient fruit sorting.