logo SBA


Digital archive of theses discussed at the University of Pisa


Thesis etd-05302021-233152

Thesis type
Tesi di dottorato di ricerca
Thesis title
Angular Momentum Based Balancing Control and Shock-Proof Design of Legged Robots
Academic discipline
Course of study
tutor Prof. Caldwell, Darwin
supervisore Prof. Featherstone, Roy
commissario Prof. Hutter, Marco
commissario Dott. Ajoudani, Arash
commissario Prof. Saccon, Alessandro
  • Legged robots
  • Balancing
  • Hopping
  • Shock-proof design
  • Legged Locomotion
  • Angular momentum based control
Graduation session start date
Release date
Legged robots aim to be ubiquitous, while they are still trying to find their feet
within the field of robotics itself, almost quite literally. The excitement that
legged robots bring to robotics research is in their promise to dynamically balance,
walk, step over obstacles, jump and/or a combination of these. As can be easily
observed, wheeled robots won’t be able to do any of these agile and dexterous motions.
Unfortunately, apart from the robots of Boston Dynamics, MIT Cheetah and a handful
of others, legged locomotion remains just that, a "promise" and it seems like quite an
uphill task, something that the robotics community at large still grapples with, owing to
the community’s research focus and lack of hardware resources and diverse technical
expertise that is required to build complicated legged robots. all in one research group.
Without the express ability to perform high performance balancing, walking or running
motions, legged robotic machines don’t really have much potential to be applied
to practical and useful scenarios. A quadrupedal or a bipedal robot have four or two
legs respectively, which is still a very complicated robotic systems with many actuators.
Other than the dynamics associated with robot legs, both quadrupeds and bipeds
have significant torso mass as well. which makes designing and controlling such robots
quite challenging. A way to reduce this complexity is to design legged robots from the
ground-up by designing one-legged robots. With that objective as the centerpiece of
our pursuits, two self-balancing robots have been conceptualized at IIT, named Tippy
and Skippy.

As their names imply, Tippy is a balancing machine that tips, while its next version,
Skippy, both balances on a single leg and hops (skips). Tippy already exists and has
been used to implement some new, high-performance 2D balance algorithms and that
give it the capability of being able to follow fast motion commands while staying balanced
on a point contact or a flat edge. However Skippy, which is in the process of
being built, is capable of a vast range of swift motions, all while maintaining its balance.
This thesis aims to experimentally demonstrate the implementation of these high-performance
balancing controllers, and to demonstrate the utility of dynamic measures
of balancing on Tippy in its various balance configurations and a hopping robot from
UC Berkeley named Salto1P. The leaning controller presented here enables Salto1P to
accurately launch to targets and balance on a point, even when the robot stops hopping.
The high-performance position and angular momentum reference tracking balance
controllers developed in this thesis will also be implemented on Skippy in the
future. Several numerical experiments are conducted to investigate the performance
of various balance controllers on robots with realistic physical parameters. Moreover,
experimental implementation of the balance controller for Tippy in a special mode of
operation known as the Reaction Wheel Configuration (RWP) is also presented, along
with the leaning controller for accurate targeted hops for Salto1P.

Legged robot design is another challenging area of research, more so for highly
athletic robots that are subjected to high impulses when they make contact with the
ground. A springy foot is a good solution to reduce the propagation of shock from
the foot of the robot to its torso. The presence of a spring helps in absorption of peak
impacts. However, some shocks can still propagate to the torso of a legged robot when
the robot makes contact with the ground, be it during walking, running or hopping.
So an instrumented shock-testing rig was et up for testing fragile components within
a robot such as absolute and quadrature position sensors, Inertial Measurement Unit
(IMU) and LiPo batteries. These shock-testing experiments provide crucial information
on the suitability of sensors or other similar components that will be able to handle the
high impacts that a highly athletic hopping robot will face. A leg design criterion that
can be used in conjunction with the passive shock-absorption methods based on the
concept of the Centre of Percusssion (CoP) is also presented in this thesis.