Using Sensor Feedback to Accurately Estimate Part Pose in Gripper

Published

September 28, 2017

Author(s)

Nithyananda Bhat Kumbla, Jeremy Marvel, Satyandra K. Gupta

Abstract

High volume manufacturing use custom made fixtures during assembly operations. Small volume manufacturing cannot use fixtures in order to keep setup time and production cost low. The performance of the task in fixture-less assemblies depends on the accuracy of the pose estimate of the part obtained from the perception system. Due to uncertainty in the estimated pose, the robot is not aware of the exact position of the part within the grippers after grasp. This may lead to undesirable effects during the task execution. We propose a cost effective method to re-estimate the pose of the part within the grippers by determining two contact points on the part and using sensor feedback along simple contact moves. We use an inertial measurement unit to estimate the orientation of the part and a force-sensitive resistor to estimate the position of the part relative to the gripper. We demonstrate the method for a peg-in-hole task. We also use a simulator to predict the performance of pose re-estimation on the peg-in-hole task. Results show that our method performs better than sensor-less systems and the predictions from the simulator match well with physical experiments.

Proceedings Title

2017 IEEE/RSJ International Conference on Intelligent Robots and Systems

Conference Dates

September 24-28, 2017

Conference Location

Vancouver, BC, CA

Pub Type

Conferences

Keywords

manufacturing robotics, robotic assembly, robot performance, peg-in-hole assembly

Citation

Kumbla, N. , Marvel, J. and Gupta, S. (2017), Using Sensor Feedback to Accurately Estimate Part Pose in Gripper, 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems, Vancouver, BC, CA (Accessed December 9, 2024)