IEEE Robotics & Automation Magazine - December 2022 - 110

Experiments
To verify effectiveness of the proposed humanoid control
method based on visual servoing, in this article, a set of
comparative experiments is performed. A classical visual
servoing controller and a proportional-integral-derivative
torque controller are used for control to highlight that our
proposed controller, with constraints, can restrict image features
to stay within the camera's FoV. The human swivelangle
motions are demonstrated during the visual constraint
servoing task, which aims to validate feasibility of the proposed
humanoid control method. The SMC is utilized for
the redundant Sawyer robot and aims to improve stability
and accuracy of the humanoid visual servoing control.
Experimental Setup
To fully verify and study the performance of the FoV constraints
approach of humanoid robot control in practice,
real-world experiments were conducted on a physical visual
servoing platform (ViSP) that we built. We present the
overall system in Figure 5. A RealSense D435i camera is
equipped on the end effector of the Sawyer robot as we
adopted eye-in-hand construction.
The vision section contains the RealSense D435i
camera, AprilTag marker, and the ViSP. The ViSP is a
C++ modular, cross-platform library developed by the
Institut de Recherche en Informatique et Systèmes Aléatoires
(IRISA) Inria Rainbow team, enabling computing
control laws and providing a set of visual features for
tracking real-time image processing or computer vision
algorithms [28]. The AprilTag image is captured online by
the RealSense D435i camera and sent to the ViSP. The
ViSP decodes the tag, extracts the corner coordinates of
the AprilTag as feature points and calculates the visual
part of the control law.
The torque-level controller of the Sawyer robot is implemented
through Intera SDK, a Python application programming
interface developed by Rethink Robotics. The Intera
SDK and the ViSP run on a computer equipped with Ubuntu
18.04 and ROS melodic. Because programming languages
are not compatible, the ROS topic is utilized to transfer data
that interact between two programming platforms.
Experimental Results
Here a comparative experiment is presented to verify the
proposed algorithm's effectiveness and ensure the stability
of the visual servoing system. The controller expression is
vk Jz
Ke tK zt Ke tdt
T
x12
=qs
11
1
=- -+ # () ,
Pq DI q
(( ))
where =- q, and qd
eq ,qqd
(( ))
are the current and desired
joint positions, respectively.
For ease of calculation, we map the feature points from the
pixel plane to the image plane. Parameter vq1 is k1
0.5 0.5 0.5 0.5 0.5 0.5) and parameter 1x is kP
diag(100 80 50 50 36 26 16), kI
5.0 4.0 2.0), kD
parameters of the proposed method are ka1
= diag(0.5
=
= diag(15.0 8.0 10.0 10.0
= diag(10.0 6.0 5.0 3.0 2.6 1.9 1.1). The
= [0.05 0.15 0.1
Intera SDK
Control and Vision Workstation
Data Interaction
Ethernet Switch
Sawyer Robot
Data Interaction
ViSP
RealSense D435i
Figure 5. A physical visual servoing framework. SDK: software development kit.
110 * IEEE ROBOTICS & AUTOMATION MAGAZINE * DECEMBER 2022
AprilTag
IEEE Robotics & Automation Magazine - December 2022

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