IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 3, NO. 4, OCTOBER 2018 3293
Dynamic Bilateral Teleoperation of the Cart-Pole: A
Study Toward the Synchronization of Human
Operator and Legged Robot
Joao Ramos and Sangbae Kim
Abstract—This letter presents the experimental evaluation of a
bilateral feedback law for the teleoperation of an underactuated
dynamic system: the Cart-Pole. This physical system illustrates
another simple model, the Linear Inverted Pendulum (LIP); a
popular template for legged robot control and, in this letter, the
mapping channel between the operator and robot. We develop a
scaling strategy based on geometric and kinematic similarity in
order to generate dynamically feasible trajectories for the LIP
with a natural frequency different than the human’s. Moreover,
by modifying the classic equations for the Cart-Pole, we show how
it can competently represent the proposed template quantitatively
and visually. Experiments where a human operator dynamically
controls slave systems with slower or faster natural frequencies
illustrate the efficacy of the proposed method. This study is a step
toward building a human–machine interface that dynamically
synchronize operator and legged robot in order to eventually
achieve complex motor behaviors.
Index Terms—Humanoid and bipedal locomotion, telerobotics
and teleoperation, haptics and haptic interfaces, human factors
and human-in-the-loop, natural machine motion.
I. INTRODUCTION
T
O BUILD a legged machine that can navigate unstructured
terrains and robustly utilize tools originally made for hu-
mans, is the dream of many roboticists. The applications for
such capable robot are endless. They can be used for search and
rescue, firefighting, space exploration or any other condition
that endangers the life of the human responder.
Although current technology has reached a state that we can
build these sophisticated machines, their autonomy, adaptability,
and robustness is yet far from what we evidence from its human
counterparts. A short-term solution to bridge this gap is to utilize
human motor control as a reference for robot motion in real-time.
This strategy has been used widely in computer graphics and
robotics [1], [2]. But we argue that these approaches miss the
utilization of physical real-time feedback to the operator, such
that s/he can adapt the motion to that particular human-machine
configuration. Yet, very few studies have attempted to provide
meaningful feedback to the operator in real-time [3], [4]. In
Manuscript received February 24, 2018; accepted June 19, 2018. Date of
publication July 4, 2018; date of current version July 19, 2018. This letter was
recommended for publication by Associate Editor N. Tsagarakis and Editor
K. Hashimoto upon evaluation of the reviewers’ comments. (Corresponding
author: Joao Ramos.)
The authors are with School of Engineering, Mechanical Engineering De-
partment, Massachusetts Institute of Technology, Cambridge, MA 02139 USA
(e-mail:, jlramos@mit.edu; sangbae@mit.edu).
Digital Object Identifier 10.1109/LRA.2018.2852840
Fig. 1. Left: Linear inverted pendulum model. This template is defined by
the total robot mass m
R
and its nominal height h
R
. The center of mass x
R
is
controlled by regulating the position of the center of pressure p
xR
. Right: 3D
representation of the LIP for Little HERMES [5], a bipedal robot capable of
dynamic behaviors.
addition, properly mapping the information between master and
slave online is still an open problem in the literature, specially
when both system have considerably different scales, dynamics
or morphologies.
To address these issues, we propose: (i) to utilize a simple
template model as the mapping channel between both systems;
and (ii) to impose geometric and kinematic similarity between
them via bilateral feedback. The first item utilizes a simple
model that captures the core dynamic behavior of each system
in order to reduce the complexity and the amount of data that
must be mapped from master to slave. Additionally, item (ii)
provides means to perform dynamic state and force transfor-
mations between human and robot, while still allowing them to
respect their inherent characteristic time responses. The Prin-
ciple of Dynamic Similarity has been extensively utilized for
the study of animal motor behavior across different scales [6],
[7], the modeling of fluid dynamics [8] and also for bilateral
feedback teleoperation of micro-scale manipulators [9]. In this
work we utilize a similar idea in order to dynamically couple
operator and robot.
The simple model utilized in this letter is the Linear Inverted
Pendulum (LIP) (see Fig. 1), a popular template for legged
robot control due to its simple but competent ability to capture
core Center of Mass (CoM) behavior [10], [11]. To evaluate
the proposed similarity mapping, we utilize a physical cart-pole
system, illustrating the promising results from the teleoperation
2377-3766 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.