Current detection based on GPR for grasping control of underac-tuated hands

Wenwu Cao, Siyu Zhao, Weidong Wang, Zhijiang Du

Research output: Contribution to conferencePaper

Abstract

In the actual process of grasping, the unexpected movement of objects is the main contributing factor for the failure. In order to eliminate such phenomenon and improve the success rate of grasping, a novel controlling method of current detection based on Gaussian Process Regression (GPR) for underactuated hands grasping is proposed in this paper. Ac-cording to the practical requirements, the criterion for current detection is established firstly. Then based on multiple groups of learning samples, the reference model for controlling is acquired through GPR machine learning, and imple-mented in experiments of a 3-knuckles-2-fingeres underactuated hand compared with current detection method based on the theoretical model. The experimental results prove that the reference model can be used for the collision detection of a variety of objects, and then the unexpected movement of objects can be eliminated to achieve grasping successfully.

Original languageEnglish (US)
StatePublished - 2017
Event49th International Symposium on Robotics, ISR 2017 - Shanghai, China
Duration: Jul 6 2017Jul 7 2017

Conference

Conference49th International Symposium on Robotics, ISR 2017
CountryChina
CityShanghai
Period7/6/177/7/17

All Science Journal Classification (ASJC) codes

  • Human-Computer Interaction
  • Artificial Intelligence
  • Building and Construction

Fingerprint Dive into the research topics of 'Current detection based on GPR for grasping control of underac-tuated hands'. Together they form a unique fingerprint.

  • Cite this

    Cao, W., Zhao, S., Wang, W., & Du, Z. (2017). Current detection based on GPR for grasping control of underac-tuated hands. Paper presented at 49th International Symposium on Robotics, ISR 2017, Shanghai, China.