Object Recognition-Based Grasping with a Soft Modular Gripper

Soft modular grippers play a significant role in multiple fields due to their excellent adaptability and flexibility. This paper proposes a modular soft modular gripper driven by pneumatically actuated multi-chambers. The designed soft modular gripper features three operational modes, with its modular fingers employing independently controlled dual chambers. The distal and proximal dual-chamber structure enhances the fingertip force of the modular fingers. Based on classical laminated plate theory and incorporating the large deformation characteristics of soft materials, a relationship between the bending centerline of the fingers and the driving pressure is established, providing a theoretical foundation for grasping tasks executed by the soft modular gripper. The Denavit-Hartenberg (D-H) parameter method is utilized to develop the coordinate system of the soft modular gripper, thereby defining its operational workspace. Visual sensing technology is introduced, incorporating improvements to the YOLOv8-based object recognition and localization framework, which enhances recognition accuracy for target objects and ensures grasping stability.