In context
In 2023, robotic grasping research was shifting toward soft, compliant end-effectors to handle complex tasks that rigid grippers could not. While underactuated hands reduced control complexity, achieving reliable multi-finger grasping in unstructured environments remained a challenge, motivating new approaches that combine soft hardware with intelligent grasp planning.
What was reported
Researchers from Zhejiang University of Technology and Northwestern Polytechnical University presented a five-fingered grasping system based on regional pose solving. The system uses a pneumatic soft hand made from materials of varying stiffness, driven by one main and five branch air tubes, enabling independent finger control with low complexity. The hand's thumb is actuated by a DC motor to rotate inward during grasping, mimicking human hand motion.
The proposed grasping strategy predicts contact regions on objects based on human grasping patterns. A neural network with feature extraction, approach, and contact region prediction modules processes scene point clouds to output approach vectors and contact areas. A regional pose solving algorithm then calculates finger bending angles by matching the contact regions with pre-recorded bending trajectories of the soft fingers, generating robust grasp poses.
To train the network, the authors created a contact dataset with objects annotated for human fingertip contact areas, designed to be scene-independent for generality. Experiments in complex scenarios demonstrated the system's effectiveness and reliability, with improved grasping success compared to simpler methods.
Why it mattered
This work addressed two key barriers in dexterous manipulation: the high complexity of fully actuated hands and the difficulty of learning finger joint angles directly. By combining a soft underactuated hand with a vision-based regional pose solving approach, it offered a practical path toward reliable five-fingered grasping in industrial settings, potentially enabling more versatile automation for tasks like bin picking and assembly.
“The experimental results demonstrate the effectiveness and reliability of the designed five-fingered soft hand grasping system in complex environments.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2023-11-10 · “基于区域姿态解算的五指机械手抓取系统设计与实现”