Wire-Driven Robotic Gripper of Sensorized Flexure Joints With Kinematic Contact Detection and Underactuated Feedback Control

Abstract

This paper presents an underactuated robotic gripper with Paired Cross Strip Pivot (PCSP) sensorized joints using embedded strain gauges. Conventional underactuated grippers provide passive adaptability but lack sensing and controllability, limiting grasping precision and dexterity. We develop a sensing-integrated gripper based on sensorized flexure joints and feedback control. The PCSP large-deformation behavior is incorporated into displacement/torque sensing, modeling, and control, and sensing performance is validated with 1.90% static error and 2.02% dynamic error. A wire-driven underactuated finger model is derived to consider workspace constraints and compute actuation. For robust control transition, we propose kinematic contact detection that switches between task-space position control and DOB-based force control regulating scalar contact force. Experiments demonstrate accurate position tracking, stable force regulation for different objects, multi-object grasping of 16 objects, and can-lid opening with 60% success. These results show that sensing and feedback control can be integrated into underactuated grippers without compromising structural simplicity.