Application of machine learning to object manipulation with bio-inspired microstructures - Journal article

Aug 5, 2026

Upload by Aurora

Bioinspired fibrillar adhesives are proposed for gripping systems that improve scalability and resource efficiency. The work presents an in-situ optical monitoring approach for contact signatures, combining image processing and machine learning. Visual features are extracted from contact images recorded at maximum compressive preload and after lifting glass objects, while handling misalignment and off-center gripping with unbalanced weight distributions. Classifiers achieve about 90% accuracy for attachment prediction depending on object mass, supporting more reliable handling in difficult scenarios.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

Introduction

Smart grippers and microfibrillar adhesion

Challenges: misalignment and interfacial defects

Vision-based tactile sensing and contact signatures

Prior work and motivation

journal of materials research and technology 2023;27:1406 e1416 Available online [at](at www.sciencedirect.com)[ www.sciencedirect.com](at www.sciencedirect.com) journal [homepage:](homepage: www. elsevier. com/locate/jmrt)[ www. elsevier. com/locate/jmrt](homepage: www. elsevier. com/locate/jmrt) | Application of machine learning to object
manipulation with bio-inspired microstructures | | | | --- | --- | --- | | Manar Samri a,b, Jonathan Thiemecke a, Ren Hensel a, Eduard Arzt a,b,c, * a INM e Leibniz Institute for New Materials, Campus D2 2, 66123 Saarbru¨cken, Germany
b Department of Materials Science and Engineering, Saarland University, Campus D2 2, 66123 Saarbru¨cken, Germany
c Department of Mechanical and Aerospace Engineering, Program in Materials Science and Engineering, University of California, San Diego, CA 92093, USA | | | | a r t i c l e i n f o | a b s t r a c t | | | Article history: | Bioinspired ﬁbrillar adhesives have been proposed for novel gripping systems with | | | Received 24 March 2023 | enhanced scalability and resource efﬁciency. Here, we propose an in-situ optical moni- | | | Accepted 30 September 2023 | toring system of the contact signatures, coupled with image processing and machine | | | Available online 6 October 2023 | learning. Visual features were extracted from the contact signature images recorded at maximum compressive preload and after lifting a glass object. The algorithm was trained to cope with several degrees of misalignment and with unbalanced weight distributions by | | | Keywords: | | | | Bioinspired-adhesives | off-center gripping. The system allowed an assessment of the picking process for objects of | | | Microstructures | various mass (200, 300, and 400 g) . Several classiﬁers showed a high accuracy of about 90 % | | | Pick and place | for successful prediction of attachment, depending on the mass of the object. The results | | | Machine learning | promise improved reliability of handling objects, even in difﬁcult situations. | | | Classiﬁcation | © 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ([http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)). | | 1. Introduction We are right on the edge of the fourth industrial revolution. As industries are transitioning towards automation and digitalization of their production lines, the need for smart grippers has grown rapidly. The goal for this transformation is to increase efﬁciency, speed, and quality of objects manipulation [1,2] . For nearly half a century, robotic grippers have relied on various technologies, such as suction and vacuum, electrostatic and magnetic attraction, and, most widespread, mechanical gripping [3] . Since the development of gecko-inspired ﬁbrillar polymer surfaces [4e12], a new gripping principle is now in the process of entering the market [13e19]: manipulation of objects by microﬁbrillar elastomer surfaces, whose adhesion can be switched on and off. Such surfaces achieve, after application of small compressive preloads, strong adhesion by van der Waals interaction and allow residue-free and silent handling, effective in both air and vacuum conditions [20e22] . These properties promise signiﬁcant beneﬁts over conventional gripping technologies, especially in manipulation of delicate and fragile objects of diverse sizes and geometries. Object manipulation has to also work under non-ideal conditions. It has to tolerate loss of the intimate contact with the target object due to interfacial defects or due to inevitable alignment inaccuracies [23e25] . As opposed to previous assumptions, it has been proven by Tinnemann et al. [26] that the different ﬁbrils behave largely independent from each other and can have widely distributed individual * Corresponding author. INM e Leibniz Institute for New Materials, Campus D2 2, 66123 Saarbru¨cken, Germany. E-mail address: [earzt@ucsd.

FAQs

What performance level is reported for predicting attachment success?

Several classifiers reach about 90% accuracy in predicting successful attachment, depending on the object mass.

Aug 5, 2026

Upload by Aurora