Touch Identification on Sensitive Robot Skin Using Time Domain Reflectometry and Machine Learning Methods

Research output: Contribution to book/conference proceedings/anthology/reportChapter in book/anthology/reportContributedpeer-review

Abstract

The article presents the proof of concept of a novel sensor system for robotic HMI applications, mimicking the human sense of touch. An artificial sensitive skin, consisting of a robust and simple part of the sensing hardware based on electrical TDR, is mounted on the robot. In combination with adaptive AI algorithms, it enables for localisation of touch events on the sensor surface as well as determination of the touch-force magnitudes. Sensor data, obtained from a robotised test stand, are utilised to train and validate regressive DNNs for touch position recognition and classification DNNs for discrete force level classification. The results demonstrate that a high level of accuracy can be obtained, but some additional work is needed to reduce the gap between training and validation accuracy.

Details

Original languageEnglish
Title of host publicationArtificial Intelligence for Digitising Industry - Applications
PublisherTaylor and Francis Group
Chapter3.4
Pages239-248
Number of pages10
ISBN (electronic)9788770226639
ISBN (print)978-87-7022-664-6
Publication statusPublished - 2021
Peer-reviewedYes

External IDs

ORCID /0000-0002-8854-7726/work/142242059
Scopus 85179126878

Keywords

Keywords

  • AI at the edge, Industrial internet of things (IIoT), Machine vision, Silicon-born AI Industrial sectors, food and beverage, industrial machinery, machine learning, Artificial intelligence, Deep learning, Neural Networks, semiconductor