IEEE Robotics & Automation Magazine - December 2022 - 127

By Ziwei Xia , Bin Fang , Fuchun Sun , Huaping Liu ,
Weifeng Xu, Ling Fu, and Yiyong Yang
and Pose Recognition
Utilizing a Multipole Magnetic Tactile Sensor
With a Metalearning Model
S
oft magnetic tactile sensors have been gradually
applied to robotic systems due to their low-cost
and simple fabrication. The previous soft
magnetic tactile sensor was developed for tactile
features of a single point (i.e., force/location)
estimation and proved the feasibility by experiments.
However, extracting tactile features of a surface
(i.e., contact shape) by magnetic sensors
remains a challenge, which limits the
application. In this article, a soft magnetic
tactile sensor that can extract contact surface
shape and pose features is fabricated, and
a multipole magnetization method is
developed to improve the performance of
the tactile sensor. Furthermore, we propose
a metric-based metalearning method to
extract the tactile feature of the contact
surface shape and pose from magnetic data
under limited sample conditions, and the
method is verified by a series of experiments.
The experimental results show that our method can
achieve more than 80% accuracy in contact shape
recognition and more than 95% accuracy in contact pose
Digital Object Identifier 10.1109/MRA.2022.3198372
Date of current version: 29 August 2022
1070-9932/22©2022IEEE
recognition. The experimental results demonstrate that
our method can extract tactile features under limited
data conditions and has a certain generalization ability
for new contact data.
Introduction
As an essential element, robotic tactile sensors have
received keen attention and have been shown to
greatly enhance the performance of dexterous
manipulation for robots [1]. Perception and
control for dexterous hands are complementary
when grasping and manipulating an
object [2]. To date, several types of tactile sensors
have been proposed that involve diverse
principles, such as piezoresistive sensors [3]
and capacitive sensors [4]. They have been
used in different robotic applications (e.g.,
force estimation, slip detection, and in-hand
manipulation [5]). However, most traditional sensors
are commonly not soft enough. Constrained by
its principle of sensing, traditional tactile sensors are
usually rigid. Therefore, when in contact with an object, only
part of the sensor is in contact with the object, and it is difficult
for these sensors to collect complete tactile information.
A challenging area for robotic tactile sensors is acquiring
richer contact information [6]. Therefore, the sensor
DECEMBER 2022 * IEEE ROBOTICS & AUTOMATION MAGAZINE *
127
Contact Shape
©SHUTTERSTOCK.COM/CORONA BOREALIS STUDIO

https://orcid.org/0000-0003-3798-0877 https://orcid.org/0000-0002-9149-7336 https://orcid.org/0000-0003-3546-6305 https://orcid.org/0000-0002-4042-6044 http://www.SHUTTERSTOCK.COM/CORONA

IEEE Robotics & Automation Magazine - December 2022

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2022