Jonker, Marissa (University of Twente), de Jong, Jan (University ot Twente), Roozing, Wesley (University of Twente), Hakvoort, Wouter (University of Twente)

Vision-Based State Estimation for Underactuated Grippers

Scheduled for presentation during the Regular Session "Robot Hand Control" (WeAT5), Wednesday, July 16, 2025, 11:40−12:00, Room 109

Joint 10th IFAC Symposium on Mechatronic Systems and 14th Symposium on Robotics, July 15-18, 2025, Paris, France

This information is tentative and subject to change. Compiled on July 16, 2025

Abstract

Grasping and handling of delicate objects is a challenge, especially in agri-food. The required force sensing to carefully handle these objects is ideally not done by tactile sensors, to upkeep cleanability and easy replacement of the grippers. To avoid the use of such sensors, we propose to estimate the contact forces from the actuator force and the configuration of the gripper using computer vision. We track integrated fiducials to obtain a pose estimate using an Extended Kalman Filter. By using a redundant set of fiducial markers, the method is robust to partial occlusions. We quantify its performance on an underactuated compliant gripper, which has a range of motion of 272 mrad for the joint angles, and 17.3 mm for the actuation phalanx movement. The results show we are able to estimate the configuration with an RMSE of approximately 29 and 17 mrad for the gripper angles, and 0.67 mm for the actuation phalanx displacement.