Date of Award

Spring 1-1-2017

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical, Computer & Energy Engineering

First Advisor

Nikolaus Correll

Second Advisor

Christopher Heckman

Third Advisor

Daniel Szafir

Abstract

This paper presents our results towards fusing RGB-D images with data from contact and proximity sensors embedded in a robotic hand for improved object perception, recognition and manipulation. Optical depth information from multiple sensors is often inaccurate and inconsistent. These problems arise from problems with sensor calibration, but also occlusion of objects by other objects or the robot arm itself. In this paper, we propose to combine global pose information from RGB-D sensing with local proximity sensing during approach. Here, we use contact information based on a novel contact sensor and additional pose information provided by the arm's pose.

Included in

Robotics Commons

Share

COinS