IEEE Robotics & Automation Magazine - December 2022 - 92

Simulation to Real
Learning Energy-Efficient Slithering Gaits for a Snake-Like Robot
By Zhenshan Bing , Long Cheng , Kai Huang , and Alois Knoll
T
o resemble the body flexibility of biological snakes,
snake-like robots are designed as a chain of body
modules, which gives them many degrees of
freedom (DoF) on the one hand and leads to
a challenging task to control them on the other.
Compared with conventional model-based control methods,
reinforcement learning (RL)-based ones provide pro -
mising solutions to design agile and energy-efficient gaits
for snake-like robots as RL-based methods can fully
exploit the hyperredundant bodies of the robots. However,
RL-based methods for snake-like robots have rarely been
investigated even in simulations, let alone been deployed
on real-world snake-like robots. In this work, we introduce
a novel approach for designing energy-efficient gaits for a
snake-like robot, which first learns a policy using an RL
algorithm in simulation and then transfers it to the realworld
testing, thereby leveraging a fast and economical
gait-generation process. We evaluate our RL-based
approach in both simulations and real-world experiments
to demonstrate that it can generate substantially more
energy-efficient gaits than those generated by conventional
model-based controllers.
Introduction
Inspired by the versatile locomotion skills of biological snakes,
researchers have designed many kinds of snake-like robots to
mimic the structural and functional properties of their counterparts
in nature. With distinctive skills, snake-like robots are
expected to be deployed in special task scenarios, where other
kinds of robots are incapable to be used, such as disaster rescue,
military surveillance, and pipeline maintenance in factories. As
snake-like robots can only carry limited energy resources for
field operations, it is essential to design energy-efficient gaits to
reduce the impact of the power constraints. An energy-efficient
movement, on the one hand, can maximize the service
time of a robot while maintaining its performance. On the
other hand, it can, in return, allow us to optimize the design of
the robot by lowing its weight or adding other hardware [1].
However, it is difficult to design energy-efficient gaits for
snake-like robots due to their redundant long mechanisms
and complex dynamic interactions with their surroundings.
The most commonly adopted method for controlling a
snake-like robot is to model its discrete body as a continuous
curve with sideslip constraints in the planar environment,
which is also widely known as the serpenoid curve [2]. This
©SHUTTERSTOCK.COM/ANDREY SUSLOV
Digital Object Identifier 10.1109/MRA.2022.3204237
Date of current version: 20 September 2022
92 * IEEE ROBOTICS & AUTOMATION MAGAZINE * DECEMBER 2022
1070-9932/22©2022IEEE

https://orcid.org/0000-0002-0896-2517 https://orcid.org/0000-0002-9993-0077 https://orcid.org/0000-0003-3347-204X https://orcid.org/0000-0003-4840-076X http://www.SHUTTERSTOCK.COM/ANDREY

IEEE Robotics & Automation Magazine - December 2022

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