King's College London

Research portal

In-Hand Object Pose Estimation Using Covariance-Based Tactile To Geometry Matching

Research output: Contribution to journalArticle

Original languageEnglish
Article number7378871
Pages (from-to)570-577
Number of pages8
JournalIEEE Robotics and Automation Letters
Volume1
Issue number1
DOIs
Accepted/In press21 Dec 2015
Published12 Jan 2016

King's Authors

Abstract

This letter presents a strategy to represent data from a tactile array sensor and match it to an object's geometric features. Using that representation, a method is presented to localise a grasped object within a robot hand. The method consists of computing the covariance matrix in the tactile sensors' pressure data and computing the eigenbasis from its principal axes. A global search is carried out to find a pose in which the object's local geometry in the vicinity of the contact is coherent with that basis, i.e., is aligned with the principal axes and has similar variances. This approach, which can be used as a measurement model for tactile sensors, is compared and outperforms methods using the distance between the tactile sensor elements and object surface.

View graph of relations

© 2018 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454