8th Vienna International Conference on Mathematical Modelling

8th
8th Vienna
Vienna18International
International Conference
Conference on
on Mathematical
Mathematical Modelling
Modelling
February
8th Vienna - 20, 2015. Vienna
International University
Conference on of Technology,
Mathematical Vienna,
Modelling
February
February 18
18 -- 20,
20, 2015.
2015. Vienna
Vienna University
University of
of Technology,
Available online at Vienna,
Technology, www.sciencedirect.com
Vienna,
Austria
February 18 - 20, 2015. Vienna University of Technology, Vienna,
Austria
Austria
Austria
ScienceDirect
IFAC-PapersOnLine 48-1 (2015) 328–333
Reduced
Reduced equations
equations of
of motion
motion for
for a
a wheeled
wheeled
Reduced
Reduced equations
equations of
of motion
motion for
for a
a wheeled
wheeled
inverted
inverted pendulum
pendulum
inverted pendulum
inverted pendulum
∗ ∗∗ ∗∗
Sergio
Sergio Delgado
Delgado ∗∗∗∗ ,,, Sneha
Sneha Gajbhiye
Gajbhiye ∗∗ , Ravi N. Banavar
Banavar ∗∗
Sergio ∗∗ , Ravi N.
∗∗ ∗∗
Sergio Delgado
Delgado , Sneha Sneha Gajbhiye
Gajbhiye ∗∗ ,, Ravi
Ravi N.
N. Banavar
∗∗
Banavar ∗∗
∗
∗ Technische Universität München, Boltzmannstr. 15, D-85748,
∗ Technische
∗
∗ Technische Universität
Universität München,
München, Boltzmannstr.
Boltzmannstr. 15,
15, D-85748,
D-85748,
Garching
Garching (Tel:
Technische
(Tel: +49-89-289
Universität
+49-89-289 15679;
München,
15679; e-mail: s.delgado@tum.de).
Boltzmannstr.
e-mail: 15, D-85748,
s.delgado@tum.de).
Garching
∗∗
Garching
∗∗
(Tel:
(Tel:
Indian +49-89-289
+49-89-289
Institute of 15679;
15679;
Technologye-mail:
e-mail:
Bombay,s.delgado@tum.de).
s.delgado@tum.de).
India (e-mail:
∗∗ Indian Institute of Technology Bombay, India (e-mail:
∗∗ Indian
∗∗
Indian Institute
Institute of
of Technology
Technology Bombay,
Bombay, India
{banavar,sneha}@sc.iitb.ac.in) India (e-mail:
(e-mail:
{banavar,sneha}@sc.iitb.ac.in)
{banavar,sneha}@sc.iitb.ac.in)
{banavar,sneha}@sc.iitb.ac.in)
Abstract:
Abstract: This
This paper
paper develops
develops the
the equations
equations of
of motion
motion in
in the
the reduced
reduced space
space for
for the
the wheeled
wheeled
Abstract:
inverted
Abstract: This
pendulum,
This paper
paper whichdevelops
is
develops an the equations
underactuated
the equations of
of motion
mechanical
motion in
in the reduced
system
the reduced subject space
space tofor
for the wheeled
nonholonomic
the wheeled
inverted
inverted pendulum,
pendulum, which
which is
is an
an underactuated
underactuated mechanical
mechanical system
system subject
subject to
to nonholonomic
nonholonomic
constraints.
inverted
constraints. The
pendulum,
The equations which
equations are are
is derived
an from
underactuated
derived from the Lagrange-d’Alembert
mechanical
the Lagrange-d’Alembert system principle
subject
principle tousing variations
nonholonomic
using variations
constraints.
constraints.with
consistent The
The equations
equations are are derived from
from the
derivedequations the Lagrange-d’Alembert
Lagrange-d’Alembert principle
principlespace, using
using variations
variations
consistent with
consistent with the the constraints.
the constraints. The
constraints. The equations are
The equations are first
are first derived
first derived in
derived in the
in the shape
the shape space,
shape space, and and then,
and then,
then,
aconsistent
a coordinate
coordinate with transformation
the
transformationconstraints. is
is performed
The equations
performed to
to get
are
get the
first
the equations
derived
equations in of
the
of motion
shape
motion in more
space,
in more andsuitable
then,
suitable
a
a coordinate
coordinates
coordinate fortransformation
the purpose
transformation of is
is performed
control.
performed to
to get
get the
the equations
equations of
of motion
motion in
in more
more suitable
suitable
coordinates
coordinates for the purpose of control.
coordinates for for the
the purpose
purpose of of control.
control.
© 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords:
Keywords: Underactuated
Underactuated mechanical
mechanical systems,
systems, nonholonomic
nonholonomic constraints.
constraints.
Keywords: Underactuated
Keywords: Underactuated mechanical mechanical systems, systems, nonholonomic
nonholonomic constraints.constraints.
1. 1
1. INTRODUCTION
1. INTRODUCTION straints
straints 1 . While traditional
1 ... While
1 traditional approaches
approaches like
like the
the Lagrange-
Lagrange-
1. INTRODUCTION
INTRODUCTION straints
straints
d’Alembert 1 While
While traditional
traditional
equations lead approaches
approaches
to the like
like the
equations the Lagrange-
Lagrange-
d’Alembert equations
d’Alembert equations lead lead to to the equations of
the equations of motion
of motion
motion
of
of nonholonomic
d’Alembert
nonholonomic equationsmechanical
lead
mechanical tosystems
the
systems (see,
equations
(see, e.
e. g.,
of
g., Pathak
motion
Pathak
The Wheeled Inverted Pendulum (WIP) - and its com- of nonholonomic mechanical systems (see, e. g., Pathak
The Wheeled Inverted
The Inverted Pendulum Pendulum (WIP) (WIP) -- and and its com- com- et al. [2005]), geometric approaches help to understand the
et
of al. [2005]),
nonholonomic geometric approaches
mechanical systems help to
(see, understand
e. g., Pathak
the
The Wheeled
mercialWheeled
version, Inverted
the Segway Pendulum -- has (WIP)
gained and its
-interest itsincom-
the et al.
al. [2005]),
structure and geometric
the intrinsic approaches
properties help
of to
to understand
the system. the
There
mercial
mercial version,
version, the
the Segway
Segway - has
has gained
gained interest
interest in
in the et
the structure[2005]), and geometric
the intrinsic approaches
properties help
of the understand
system. the
There
mercial
past version,
several years thedue Segwayto its -maneuverability
has gained interest and simple structure
in the structure
is a lot of and
and
workthe
the intrinsic
intrinsic
regarding properties
properties
the modeling of
of the
the
of system.
system. There
There
nonholonomic
past several years due to its maneuverability and simple
past
past several
several years
construction yearse.g. due
due to to its
its maneuverability
maneuverability and simple is
and simple is aa lot
a lot
lot of
of workwork regarding
regarding the
the modeling of
the modeling of nonholonomic
of nonholonomic
construction (see (see e.g. Grasser
Grasser et et al.al. [2002],
[2002], Segway
Segway [2015, [2015, is systems, ofsee for
work example
regarding Bloch [2003],
modeling Ostrowski [1999],
nonholonomic
construction
construction
Jan]). Other (see
(see e.g.
robotic Grasser
e.g.systems
Grasserbased et
et al.al.on[2002],
[2002],
the WIP Segway
Segway
are [2015, systems,
[2015,
becom- systems,
Bloch
systems, et
see
see
al.
see
for
for
[1996],
for
example
example
Bloch
example
Bloch
Bloch
et
Bloch al.
[2003],
[2003],
[2009]
[2003],
Ostrowski
Ostrowski
and the
Ostrowski
[1999],
[1999],
references
[1999],
Jan]). Other
Jan]). Other robotic
robotic systems
systems based based on on the
the WIP
WIP are are becom-
becom- Bloch et al. [1996], Bloch et al. [2009] and the references
Jan]).
ing Other as
popular robotic
well systems
in the based on
robotic the WIP are
community for becom-
human Bloch
Bloch
therein. et
et al.
al.
These [1996],
[1996], Bloch
Bloch
geometric et
et
tools al.
al. [2009]
[2009]
help and
and
understand the
the references
references
the mech-
ing popular as well in the robotic community for human
ing
ing popular
popular
assistance
assistance or
or
as
as well
well in in the
transportation
transportation
the robotic
robotic
as can
as can be community
community
be seen in
seen in the
thefor human
forworks
human
works of therein.
of
therein.
anism
These geometric
therein.ofThese
These geometric
geometric
locomotion, i. e.,
tools
tools
tools
the
help
help understand
understand
helpmotion
way understand is
the
the
generated
mech-
the mech-
mech-by
assistance
assistance
Li et al. or
or
[2012], transportation
transportation
Nasrallah et as
as al. can
can be
be
[2007], seen
seen
Baloh in
in the
the
and works
works
Parent of
of anism
anism of
of locomotion,
locomotion, i.
i. e.,
e., the
the way
way motion
motion is
is generated
generated by
by
Li et al.
al. [2012], Nasrallah
Nasrallah et et al.al. [2007], Baloh Baloh and and Parent
Parent changing
anism of the shape
locomotion, of i.the
e., mechanical
the way motionsystem. is generated by
Li
Li et
et al.A[2012],
[2003]. [2012],
WIP Nasrallah
consists of a al. [2007],
etvertical [2007],
body Baloh
with Parent changing
and coaxial
two changing the
the shape
shape of
of the
the
changing the shape of the mechanical system.
mechanical
mechanical system.
system.
[2003].
[2003]. A
A WIP consists of a vertical body with two coaxial
A WIP
WIP consists
[2003]. wheels.
driven consists of of aa vertical
vertical body body with
with two two coaxial
coaxial Symmetries
Symmetries can can bebe exploited
exploited to to develop
develop dynamical
dynamical models models
driven wheels.
driven wheels. Symmetries
in a reduced
Symmetries can be
space.
can be exploited
Roughly
exploited to develop
speaking,
to develop dynamical
the Lagrangian
dynamical models
models L
driven wheels. in aa reduced space. Roughly speaking, the Lagrangian L
The stabilization and tracking control for the WIP is chal- in
exhibits
in a reduced
a
reduced space.
symmetry
space. Roughly
if it
Roughly does speaking,
not depend
speaking, the
the onLagrangian
one configu-
Lagrangian L
L
The
The stabilization
stabilization and
and tracking
tracking control
control for
for the
the WIP
WIP is
is chal-
chal- exhibits a symmetry if it does not depend on one configu-
The stabilization
lenging: the system andbelongs
tracking to control
the forofthe
class WIP is chal- exhibits
underactuated exhibits
ration a
a symmetry
symmetry
variable, lets if
if
say, it
it q does
does
. The not
not depend
depend
variable q on
on
is one
one
calledconfigu-
configu-
cyclic.
lenging:
lenging: the system belongs to the class of underactuated ration variable, lets say, j
qqjj .. The variable qqjj isj called cyclic.
lenging: the
mechanical the system
system belongsbelongs to
to the the class
class ofof underactuated
underactuated ration variable,
variable, lets say, qjj . The
The variable qjj is
is called
called cyclic.
mechanical systems,
mechanical systems, since
systems, since the
since the control
the control inputs
control inputs are
inputs are less
are less than
less than ration
than
The
The Lagrangian
Lagrangian
lets
is
is
say, invariant
thus
thus invariant
variable
under
under transformations
transformations
cyclic.
in
in
mechanical
the number systems,
of since
configuration the control
variables: inputs
There are less
a than
total of The
cyclic
The Lagrangian
coordinates.
Lagrangian is
is thus
Lie
thus invariant
group
invariant actionunder
underand transformations
symmetry
transformations in
reduc-
in
the number of
the of configuration
configuration variables: variables: There There are are aa total
total ofof cyclic coordinates. Lie group action and symmetry reduc-
the number
two number
control of configuration
variables τ and variables:
τ which There
are
1 and τ2 which are the torques ap- the are a
torques total of
ap- cyclic
tion
cyclic coordinates.
has been
coordinates. Lie
successfully
Lie group
group action
applied
action and
to
and symmetry
model
symmetry other reduc-
types
reduc-
two control variables τ
two
two control
plied
plied
control
to
to rotate
rotate
variables
variables
the
the
ττ111 and
wheels,
wheels, 1 and and
and
ττ222 six
which
2 six
are
are the
configuration
which
configurationthe torques variables,
torques
variables,ap- tion
ap- tion
tion
of
has
has been
has
nonholonomicbeen successfully
been successfully
successfully
mechanical
applied
applied
applied
systems
to
to model
to model
model
in the
other
other
other
types
types
types
differential
plied
namely,
plied to
to rotate
the x-
rotate the
and
the wheels,
y- position
wheels, and
and of six
the
six configuration
WIP on
configuration the variables,
horizontal
variables, of nonholonomic
of nonholonomic
nonholonomic mechanical
mechanical systems
systems the in the
in works differential
the differential
differential
namely, the x- and y- position of the WIP on the geometric
of framework. See
mechanical
horizontal geometric framework. See for example the works by Bloch for example
systems in the by Bloch
namely,
namely, the the x-
x- andand y- y- position
position of of the WIP
WIP on on the the horizontal
horizontal geometric framework. See for
plane,
plane, the
the relative
relative rotation
rotation angle angle of
theeach
each of
of each of the
the wheels
wheels with et
with geometric
et al.
al. [1996],
[1996], Ostrowski
framework.
Ostrowski See [1999],for example
example
[1999], Gajbhiye
Gajbhiye
the
the works and
works
and
by Bloch
Banavar
by Bloch
Banavar
plane,
respect
plane, the
to
the relative
the body
relative rotation
φ
rotation and angle
φ
angle, the of
of of
orientation
each of the
the wheels
angle
wheels θ, with
and
with et al. [1996], Ostrowski [1999], Gajbhiye and Banavar
respect to the body φ 1 and φ2 , the orientation angle θ, and [2012].
et al.
[2012]. As
[1996],
As shown,
shown, e.
Ostrowski
e. g., by Ostrowski
[1999], [1999],
Gajbhiye the
and resulting
Banavar
respect
the
respect to
to the
tilting angle
the body
body α. φ
In
1 and φ2
and φ222 ,, the
φ111 addition, the
the orientation
system
orientation is angle
angle θ,
restrictedθ, andby
and [2012].
[2012].
equations As can
As shown,
shown,
be e. g.,
e.
put g.,
g.,
in
by
by
by
Ostrowski
Ostrowskiform
Ostrowski
aa simplified
[1999],
[1999],
[1999],
the
the resulting
the
containing resulting
resulting
apart
the
the tilting
tilting angle
angle α.
α. In
In addition,
addition, the
the system
system is
is restricted
restricted by
by equations can be put in simplified form containing apart
nonholonomic
the tilting angle (nonintegrable)
α. In addition, constraints
the system and
is is thus
restricted not
by equations
from the
equations can be
reduced
can be put
put in aa simplified
equations
in simplified
of motion, form
also
form containing
the momentum
containing apart
apart
nonholonomic
nonholonomic (nonintegrable)
(nonintegrable) constraints
constraints and
and is
is thus
thus not
not from the reduced equations of motion, also the momentum
nonholonomic
smoothly (nonintegrable)
stabilizable at a point constraints
as proven and by is thus
Brockettnot from
and
from the reduced
reconstruction
the reduced equations
equation,
equations of
of motion,
which
motion, also
describe
also the
thethemomentum
dynamics
momentum
smoothly stabilizable
smoothly stabilizable at at a a point
point as as proven
proven by by Brockett
Brockett and reconstruction equation, which describe the dynamics
[1983].
smoothly These constraints
stabilizable at do not
a point restrict
as proventhe state space
by Brockett on and
of
and thereconstruction
system
reconstruction along equation,
the group
equation, which describe
directions.
which describe That the
theis,dynamics
how
dynamicsthe
[1983].
[1983]. These
These constraints
constraints do
do not
not restrict
restrict the
the state
state space
space on
on of the system along the group directions. That is, how the
which
[1983]. the
Thesedynamics
constraints evolve, do but
not the motion
restrict the direction
state space atona of
of the
the
system system
system
translates along
along the
the
and group
group
rotates directions.
directions.
in space dueThat
That to is,
is,
the how
how the
the
change
which
which the
the dynamics
dynamics evolve,
evolve, but
but the
the motion
motion direction
direction at
at a
a system translates and rotates in space due to the change
given
which point:
the The
dynamics rolling
evolve,constraint
but theimpedes
motion a sideways
direction mo-
at a system
in the
system translates
shape
translates and
variables.
and rotates
Bloch
rotates in
et
in space
al.
space due
[2009]
due to
to the
further
the change
show
change
given point: The rolling constraint impedes a sideways mo-
given
givenand
tion, point:
point:the The
The
forwardrolling
rolling constraint
constraint
velocity of impedes
impedes
the WIP and a
a sideways
sideways
its yaw mo- in
mo-
rate in
the
in
the
the shape
shape
advantage
the
variables. Bloch
shape variables.
of using
variables. the Bloch
Bloch Hamel
et
et
et al.
al. [2009]
[2009]
equations
al. [2009] to
further
further
obtain
further
show
show
the
show
tion,
tion, and
and the
the forward
forward velocity
velocity of
of the
the WIP
WIP and
and its
its yaw
yaw rate
rate the advantage of using the Hamel equations to obtain the
are
tion,directly
and the given
forward by the
velocity angular of thevelocity
WIP and of the
its wheels.
yaw rate the advantage
reduced
the of
nonholonomic
advantage of using
using the
equations
the Hamel
Hamel equations
of motion:
equations to
The
to obtain
momen-
obtain the
the
are
are directly
directly given
given by
by the
the angular
angular velocity
velocity of
of the
the wheels.
wheels. reduced nonholonomic equations of motion: The momen-
Wheeled
are directly robotsgiven have
by largely
the angular been considered
velocity of the as purely
wheels. reduced
reduced
tum nonholonomic
nonholonomic
equation is in this equations
equations
case given of
of
in motion:
motion:
a body The
The
frame momen-
momen-
which
Wheeled
Wheeled robots
robots have
have largely
largely been
been considered as purely tum equation
equation is is in
in this case case given given in in aa body
body frame frame which
which
Wheeled
kinematic robots
systems, have
due largely
to the been considered
considered
simplification in theas purely
purely tum
as motion appears
tum to
equation be is in this
more natural
this casethan given in aa spatial
in a bodyframe, framefor the
which
kinematic
kinematic systems, due
due to
systems, analysis. to the
theThe simplification
simplification in
in the motion
theneeds
motion appears to be more natural than in spatial frame, for the
kinematic
and systems,
controllability due to the simplification
WIP, however, in the motion to appears
latter
appears is to be
rarely
to be more natural
conserved
more natural forthan in
systems
than in a
a spatial
with
spatial frame, for
nonholonomic
frame, for the
the
and
and controllability
controllability analysis.
analysis. The
The WIP,
WIP, however,
however, needs
needs to
to latter is rarely conserved for systems with nonholonomic
be
and stabilized
controllability by dynamic
analysis. effects,
The such
WIP, that
however, the complete
needs to latter
latter is
is
constraints. rarely
rarely The conserved
conserved
derivation for
for ofsystems
systems
the with
with
reduced nonholonomic
nonholonomic
nonholomonic
be stabilized by dynamic effects, such that the complete
be
be stabilized
dynamics
stabilized
dynamics need
need
by
bytoto
dynamic
be
dynamic
be taken
taken
effects,
into
effects,
into
such
such that
account.
account.thatIn In
the
the complete constraints.
complete
mechanical
mechanical
constraints.
constraints.
equations can
The
The
Thebe
derivation
derivation
derivation
done as
of
of the
of
well the
the
using
reduced
reduced
reduced
the
nonholomonic
nonholomonic
nonholomonic
constrained La-
dynamics
dynamics
systems need
need
with to
to be
be
nonholonomic taken
taken into
into
constraints account.
account. the In
In mechanical
mechanical
configuration equations
equations can
can be
be done
done as
as well
well using
using the
the constrained
constrained La-
La-
systems with nonholonomic constraints the grangian
equations and
can a
be so-called
done as Ehresmann
well using the connection
constrained which
La-
systems
space
systems Q with
is a
with nonholonomic
finite dimensional
nonholonomic constraints
smooth
constraints the configuration
manifold,
the configuration
T Q
configurationis the grangian
grangian and
and aa so-called
so-called Ehresmann
Ehresmann connection
connection which
which
space Q is a finite dimensional smooth manifold, T Q is the relates
grangian motionand along
a the
so-called shape directions
Ehresmann with
connection the motion
which
space
tangent
space
tangent
Q is
is a
a finite
Q bundle
bundle finite dimensional
-- the
the velocity
dimensional
velocity
smooth
phase
smooth
phase
manifold,
space -- and
manifold,
space and
T
aa Q
T is
smooth
Q
smooth the relates
is the relates motion along
relates motion
motion along along the
the shape
the shape
directions with
shape directions
directions with with the
the motion
the motion
motion
tangent
tangent bundle
(non-integrable)
bundle -
- the
the velocity
distribution
velocity phase
D
phase ⊂ space
T Q
space -
defines
- and
and a
a smooth
the
smoothcon- 1 The distribution D defines the admissible velocities
(non-integrable)
(non-integrable) distributiondistribution D D ⊂ ⊂ T TQQ defines
defines the the con-
1
con- 111 TheThe distribution
distribution D D defines
defines the the admissible
admissible velocities
velocities
(non-integrable) distribution D ⊂ T Q defines the con- The distribution D defines the admissible velocities
2405-8963 ©
Copyright © 2015,
2015,IFAC
IFAC (International Federation of Automatic Control)
328Hosting by Elsevier Ltd. All rights reserved.
Copyright
Peer review©
Copyright © 2015,
2015, IFAC
IFAC 328
328
Copyright ©under
2015,responsibility
IFAC of International Federation of Automatic
328Control.
10.1016/j.ifacol.2015.05.011
 MATHMOD 2015
February 18 - 20, 2015. Vienna, Austria Sergio Delgado et al. / IFAC-PapersOnLine 48-1 (2015) 328–333 329

along the group directions. The approach is based on On the configuration space Q = G × S, the Lagrangian
taking admissible virtual displacements from the Lagrange is a function L : T G × T S → R and the distribution
d’Alembert principle. Admissible means, that the varia- characterizing the nonholonomic constraints is given by
tions satisfy the constraints (given by the connection). D ⊂ T Q. A curve q(t) on Q is said to satisfy the constraints
This paper follows this modeling tool. Note that we are if q̇(t) ∈ Dq , ∀t. This nonholonomic restriction can also be
not imposing the constraints before taking variations, we given in local coordinates as
are taking variations according to the constraints. ġ + AT ṡ = 0, (1)
Several control laws have been applied to the WIP, mostly where g ∈ G and s ∈ S, and the matrix A describes how
using linearized models as can be seen in Li et al. [2012]. ġ and ṡ are related to each other due to the constraints.
There is still the need to exploit the nonlinear geometric Recall that the equations governing the dynamics of the
structure of the WIP to stabilize and control the system system satisfy the Lagrange-d’Alembert Principle (Fext
using coordinate-free control laws. Nasrallah et al. [2007] denote the external forces)
 
develop a model based on the Euler-Rodrigues parameters T
and analyze the controllability of the WIP moving on δ L(q, q̇)dt + Fext δq dt = 0, (2)
an inclined plane. Pathak et al. [2005] develop a model which is equivalent to
using the Lagrange-d’Alembert equations and check the       
d ∂L ∂L T
strong accessibility condition. The aim of this note is to − − Fext δq dt = 0. (3)
explore the motion of the system in the reduced (shape) dt ∂ q̇ ∂q
space which lead to some net displacement of the mobile Independent from the Lie-group structure, we can intrinsi-
robot (motion in the group space) independently from cally eliminate the Lagrange-multipliers which arise from
the starting point. Additionally, we present the equations the constraint forces and write the reduced equations of
of motion in more suitable coordinates 2 for control or motion using the Ehresmann connection (1), which is
trajectory planning purposes: Since the shape space of nothing but a way to split the tangent space into a hori-
the WIP is not fully actuated, the control task becomes zontal (tangent to the shape space) and a vertical (tangent
difficult in these coordinates. The choice of the model can to the group space) part 3 . The curves q(t) solving the
be done depending on which better suits the task. equations of motion need to satisfy the constraints. Thus,
the variations δq = (δs, δg) are of the form δg + AT δs = 0
Notation: Contrary to most of the literature, we use (see Bloch et al. [1996]). We assume, that the external
here the matrix/vector representation instead of the index forces are input torques τ and only act on the shape
convention. Readers are encouraged to read the referred T
variables, i. e., Fext δq = τ T δs. This assumption is valid,
literature for the wide-used index convention. Further, we since we will consider group space motion only as a result
use the following simplified notation for the transposed of a change in the shape variables, and we neglect friction
 ∂ T
Jacobian: ∂xT = ∂x . forces. Equation (3) takes the following form
      
d ∂L ∂L T
− − τ δs dt
2. EQUATIONS OF MOTION IN SHAPE SPACE dt ∂ ṡ ∂s
     
d ∂L ∂L
Consider the configuration space Q = G × S, where S − − AT δs dt = 0. (4)
dt ∂ ġ ∂g
denotes the shape space and G denotes the group space: Q
is a trivial principal bundle with fibers G over a base man- To eliminate the group velocities ġ, define the constrained
ifold S. The shape space, as the name suggests, denotes Lagrangian
the space of the possible shapes of the system. As stated Lc (s, g, ṡ) = L(s, g, ṡ, −AT ṡ). (5)
by Ostrowski [1999], this division is natural in mechanisms The following relationships hold
that locomote, like mobile robots, where position changes ∂Lc ∂L ∂L ∂ ġ ∂L ∂L T
are generated by (mostly cyclic) changes in the shape. = + = − A (6)
∂ ṡ ∂ ṡ ∂ ġ ∂ ṡ ∂ ṡ ∂ ġ
See for example the oscillations of a snake-board which
create the forward motion, or the rotation of the wheels of ∂Lc ∂L ∂L ∂ ġ ∂L ∂L ∂(AT ṡ)
= + = − (7)
a mobile platform resulting in a platforms displacement ∂s ∂s ∂ ġ ∂s ∂s ∂ ġ ∂s
due to the rolling-without-slipping interaction with the ∂Lc ∂L ∂L ∂ ġ ∂L ∂L ∂(AT ṡ)
environment. The internal shapes of the WIP are solely = + = − . (8)
∂g ∂g ∂ ġ ∂g ∂g ∂ ġ ∂g
defined by the relative angles of the wheels with respect to According to (4), and using the mentioned relationships
the body. And since the gravity acts on the WIP depending (6) - (8), the equations of motion in terms of the con-
on the tilting angle (the gravity breaks the symmetry), and strained Lagrangian Lc are given by
it is crucial for the stability of the system, the tilting angle
d  T 
is also considered as a shape variable (more on that later). ∂ Lc − ∂sT Lc + A ∂gT Lc = τ − B ∂ġT L, (9)
Note that the net motion resulting from a change in shape dt ṡ
is independent from the initial position (we assume, that where
the WIP is moving on a horizontal plane). Mathematically d  T  d  T 
B ∂ġT L = A ∂ġ L − A ∂ L
speaking is this nothing but an invariance (symmetry) dt T T dt ġ 
of the Lagrangian under a change in position (group) + A ∂g (A ṡ) − ∂sT (AT ṡ) ∂ġT L
coordinates. We are therefore interested in the reduced ⇒ B = Ȧ − ∂sT (AT ṡ) + A ∂gT (AT ṡ). (10)
equations of motion in shape space variables.
3 The reader is referred to the references for detailed information
2 The same equations of motion can be found in Pathak et al. [2005] regarding Ehresmann connections

329
MATHMOD 2015
330
February 18 - 20, 2015. Vienna, Austria Sergio Delgado et al. / IFAC-PapersOnLine 48-1 (2015) 328–333
The equations of motion can be further simplified if there
additionally exists a symmetry with respect to the group
variables, meaning that the constrained Lagrangian is
independent from g and thus ∂gT Lc = 0. In that case the
equations of motion are
d  T 
∂ Lc − ∂sT Lc = τ − B ∂ġT L. (11)
dt ṡ
Note that these are not the Euler-Lagrange equations for
the constrained Lagrangian Lc (s, ṡ) - that would have been
the case for holonomic (integrable) constraints. The term
B ∂ġT L on the right hand side of (11) would be missing if we
had imposed the constraints before taking the variations
as shown in Bloch [2003]. This ”forcing” term is written
in terms of the curvature of the Ehresmann connection (1)
as stated in Bloch et al. [1996], and can be interpreted as
additional gyroscopic forces.
3. THE WHEELED INVERTED PENDULUM
Figure 1 shows a simple scheme of the WIP. It is basically
a body of mass mB (center of mass at a distance b from
the wheels rotation axes) mounted on two wheels of radius
r. The distance between the wheels is 2d and their mass
is denoted by mW . The wheels are directly attached to
the body and can rotate independently. Since they are
actuated by motors sitting on the body itself, a tilting
motion will automatically rotate the wheels by the tilting
angle if the wheels are blocked. The body needs to be
stabilized in the upper position through a back and forth
motion of the system similar to the inverted pendulum
on a cart. In this section we introduce the configuration
variables and move on to defining the velocities, the
Lagrangian and the constraints.
α α
mB , JB , b
Iz
Sz
r, mW , JW Bz
Sy τ2 , φ2
Ix By
Iy
Sx
Bx
2d θ coordinate frame 4 (position of the shaft) given as
Fig. 1. The Wheeled Inverted Pendulum
We will mainly make use of three different coordinate
systems: The inertial I-System, the shaft S-System, which
has been rotated around the I z-axis by the yaw angle θ and
will be used only for visualizing purposes, and the body
fixed B-System, which is attached to the pendulum’s body.
For completeness, we also introduce the Wj - frames (for
j = 1, 2), which are fixed to the wheels. The following
notation has been adopted. I (∗) - inertial frame, S (∗) -
shaft frame, B (∗) - body (pendulum) frame, Wj (∗) - wheel
j frame. The triples (i ê1 , i ê2 , i ê3 ) for i = {I, S, B, Wj }
denote the unit vectors of the respective coordinate sys-
tem. The set of generalized coordinates describing the WIP
consists of
(1) Coordinates of the origin of the body-fixed coordinate
system in the horizontal plane (x, y ∈ R2 )
(2) Heading angle around the I z-axis (θ ∈ S1 )
(3) Tilting angle around the S y-axis (α ∈ S1 )
(4) Relative rotation angle of each of the wheels with
respect to the body around the Wj y-axis, which
coincides with the B y-axis (φ1 ∈ S1 and φ2 ∈ S1 )
The configuration space Q = G × S of the system is thus
(R2 × S1 ) × (S1 × S1 × S1 ). Let X be a point on the
pendulum, then the relationship between the coordinates
of X in different frames is given by
I X = RIS (θ)RSB (α) B X; S X = RSB (α) B X,
where RIS (θ) is the orientation of the shaft with respect
to the inertial frame, and RSB (α) is the orientation of the
pendulum with respect to the shaft given by
 
cos θ − sin θ 0
RIS (θ) = Rθ = sin θ cos θ 0 , (12)
0 0 1
 
cos α 0 sin α
RSB (α) = Rα = 0 1 0 . (13)
− sin α 0 cos α
Analogously, let X be a point on the left wheel - wheel 1
(the right wheel case works the same way). The orientation
of the wheel with respect to the pendulum is given by
 
cos φ1 0 sin φ1
RW1 (φ1 ) = Rφ1 = 0 1 0 , (14)
− sin φ1 0 cos φ1
such that the coordinate representation of X in the inertial
and the shaft frame reads
I X = Rθ Rα Rφ1 1 X , S X = Rα Rφ1 1 X.
Vectors represented in one of the systems (sys) can be
transformed into a different coordinate system by the
following rule:
θ α φ1/2
I-sys −−−−
→ S-sys −−−−→ B-sys −−−−→ W1/2 -sys.
z-axis y-axis y-axis
Note that φ1 and φ2 are the relative angles of the left and
right wheel with respect to the body. This definition seems
natural: Since the wheels sit on the body, we can measure
τ1 , φ1
the relative angle of rotation with respect to the body. The
absolute wheel rotation angles are given by ϕ1 = α + φ1
and ϕ2 = α + φ2 . Now, let rS be the origin of the B-
rS = x I ê1 + y I ê2 + r I ê3 , (15)
where r is the wheel radius. A point X on the pendulum
(body/bar) will hence be given in inertial coordinates as
(x)b = Rθ Rα X + rS . (16)
For X being a point on the left (right) wheel expressed in
body fixed coordinates, its inertial position is given as
(x)w1 = Rθ Rα (Rφ1 X + d B ê2 ) + rS , (17)
w2
(x) = Rθ Rα (Rφ2 X − d B ê2 ) + rS . (18)
3.1 Velocities
Differentiating (15) we get the translational velocity of the
origin of the B-frame
I ṙS = ẋI ê1 + ẏI ê2 . (19)
4 Note that the origin of both, the S- and the B-coordinate frame
coincide
330
MATHMOD 2015

February 18 - 20, 2015. Vienna, Austria Sergio Delgado et al. / IFAC-PapersOnLine 48-1 (2015) 328–333 331

   
Now, we differentiate (16), (17), and (18) to calculate the JBxx 0 0 JWxx 0 0
inertial velocity of a point on the pendulum, and on the JB = 0 JByy 0 , JW = 0 JWyy 0 .
wheels, which is given as 0 0 JBzz 0 0 JWxx
(ẋ)b = (Ṙθ Rα + Rθ Ṙα ) X + ṙS (20) (30)
The kinetic energy of the body is given as
(ẋ)w1 = (Ṙθ Rα + Rθ Ṙα )(Rφ1 X + d B ê2 ) 1 1
TB = mB �ṙS +b(Ṙθ Rα +Rθ Ṙα )B ê3 �2 + ωbT JB ωb , (31)
+ Rθ Rα Ṙφ1 X + ṙS , (21) 2 2
w2
(ẋ) = (Ṙθ Rα + Rθ Ṙα )(Rφ2 X − d B ê2 ) where ωb = B ωb is the absolute rotation of the body in
body coordinates
+ Rθ Rα Ṙφ2 X + ṙS . (22) T
ωb = Rα (
ωθ )Rα + ω α .
T
Since all rotation matrices satisfy R(t) R(t) = I, ∀t, by Let ωwj = ωb + ωφj be the absolute rotation of the wheel j
differentiating with respect to time we get the relation in its wheel fixed frame Wj . From (21) and (22), and with
RT Ṙ + ṘT R = 0, meaning that RT Ṙ is skew symmetric.
the fact, that RαB ê2 = S ê2 , and ṘαB ê2 = 0, the kinetic
The matrix ω̂ = RT Ṙ, defined as energy of the wheels takes the form
 
0 −ω3 ω2 1 1 T
 = ω3 0 −ω1
ω (23) TW1 = mW �ṙS + dṘθ S ê2 �2 + ωw JW ωw1 (32)
2 2 1
−ω2 ω1 0 1 1 T
TW2 = mW �ṙS − dṘθ S ê2 �2 + ωw JW ωw2 . (33)
denotes the relative angular velocity of the body with 2 2 2
respect to its body fixed coordinate frame and expressed The potential energy is given by V = mB gb cos α. The
in the body frame. The body velocities can therefore be Lagrangian is simply L = TB + TW1 + TW2 − V .
given as
(ẋ)b = Rθ Rα (Rα
T
(
ωθ )Rα + ω
α ) X + ṙS , (24) 3.3 Constraints
w1 T
(ẋ) = Rθ Rα (Rα (
ωθ )Rα+ωα )(Rφ1 X + d B ê2 )
Since the wheels roll without slipping on the plane, the
+ Rθ Rα Rφ1 (
ωφ1 ) X + ṙS , (25) velocity of the center points of the wheels
(ẋ)w2 = Rθ Rα (Rα
T
(
ωθ )Rα + ωα )(Rφ2 X − d B ê2 ) (rc )1 = rS + dRθ Rα B ê2 (34)
+ Rθ Rα Rφ2 (
ωφ2 ) X + ṙS . (26) (rc )2 = rS − dRθ Rα B ê2 (35)
is solely given by their (absolute) rotation
3.2 Lagrangian (ṙc )1 = rRθ Rα ( T T
ωw1 )Rα Rθ I ê3 (36)
T T
(ṙc )2 = rRθ Rα (
ωw2 )Rα Rθ I ê3 . (37)
The kinetic energy T is the sum of the kinetic energy terms
of each of the bodies (pendulum and wheels) Differentiating (34) and (35), and comparing them to the
  equations above yields the rolling constraints as a subset
1 1 D of T Q
T = �(ẋ)b �2 ρ(X)dV + �(ẋ)w1 �2 ρ(X)dV
2 B 2 W1 T T
 ṙS + d(Ṙθ Rα + Rθ Ṙα ) B ê2 = rRθ Rα (
ωw1 )Rα Rθ I ê3
1 (38)
+ �(ẋ)w2 �2 ρ(X)dV. (27)
2 W2 T T
ṙS − d(Ṙθ Rα + Rθ Ṙα ) B ê2 = rRθ Rα (
ωw2 )Rα Rθ I ê3 .
The potential energy due to the gravity is given by (39)
V = mb gb (I ê3 T Rθ Rα B ê3 ) = mb gb �I ê3 , Rθ Rα B ê3 �,
(28) 4. GROUP ACTION, INVARIANCE OF THE
where b B ê3 denotes the vector from the origin of the body LAGRANGIAN AND DISTRIBUTION, AND
frame to the body’s center of mass, expressed in body fixed EQUATIONS OF MOTION
coordinates, and g the gravity constant. The Lagrangian
is defined as L = T − V . Using (20), (21), (22), and (28), The left (or right) action of a Lie group G on a smooth
L can be written as manifold M is a mapping Φ : G × M → M . Assuming
 the action of G is free and proper (Φ is simple, and
1
L= �(Ṙθ Rα + Rθ Ṙα ) X + ṙS �2 ρ(X)dV therefore M/G is a smooth manifold and the mapping
2 B
 π : M → M/G is a submersion), the Lagrangian L is
1 said to be invariant under the group action if L remains
+ �(Ṙθ Rα + Rθ Ṙα )(Rφ1 X + d B ê2 )
2 W1 invariant under the induced action of G on T M . For more
+ Rθ Rα Ṙφ1 X + ṙS �2 ρ(X)dV details on this topic see, e.,g., Marsden and Ratiu [1994],
 or Holm et al. [2009]. Consider the configuration space
1
+ �(Ṙθ Rα + Rθ Ṙα )(Rφ2 X − d B ê2 ) Q as a submanifold of the space Q̃ = (R3 × S1 ) × (S1 ×
2 W2
S1 × S1 ) = G̃ × S, where S denotes the shape space and
+ Rθ Rα Ṙφ2 X + ṙS �2 ρ(X)dV consists of the tilt angle α, and the relative wheel angles
− mb gb �I ê3 , Rθ Rα B ê3 �. (29) φ1 and φ2 . On this extended space, the Lagrangian is a
Assume that the moment of inertia of body and the function L̃ : T Q̃ → R and the distribution characterizing
wheels have the following diagonal form (in body-fixed the constraints is given by D̃ ⊂ T Q̃. Note that Q̃ is a trivial
coordinates) principal bundle with fibers G̃ over a base manifold S. If

331
MATHMOD 2015
332
February 18 - 20, 2015. Vienna, Austria Sergio Delgado et al. / IFAC-PapersOnLine 48-1 (2015) 328–333

the Lagrangian L̃ and the Distribution D̃ are invariant as D = { (s, g, ṡ, ġ) ∈ T Q | ġ + AT ṡ = 0 }, for the
under the action of G̃, the dynamics can be reduced group variables g = (x, y, θ)T , and shape variables s =
to the quotient space Q̃/G̃ (the set of orbits), which is (α, φ1 , φ2 )T . The constraints are satisfied in that specific
diffeomorphic to g̃ × S. The left action of the Lie group 5 set of local coordinates for
� �
G̃ = {(s, R̄) ∈ R3 × S1 |R̄ê3 = ê3 }. (40) 1 −2 r cos θ −2 r sin θ r0
A= −r cos θ −r sin θ /d . (47)
on the manifold Q̃ is given by 2 −r cos θ −r sin θ −r/d
Φ(R̄,s) :((rS , Rθ ), (Rα , Rφ1 , Rφ2 )) We can get the constrained Lagrangian Lc by replacing
→ ((R̄rS + s, R̄Rθ ), (Rα , Rφ1 , Rφ2 )). (41) ġ = −AT ṡ
The left action on the tangent-lifted coordinates of the in (29) according to (47). Due to the symmetry of the La-
manifold Q̃ is grangian and the distribution, Lc = Lc (s, ṡ) is additionally
independent from g. The governing equations describing
T Φ(R̄,s) :((ṙS , Ṙθ ), (Ṙα , Ṙφ1 , Ṙφ2 )) the dynamics of the system are thus given by (11). The
→ ((R̄ṙS , R̄Ṙθ ), (Ṙα , Ṙφ1 , Ṙφ2 )). (42) input is simply the torque applied to the wheels
T
Claim: The Lagrangian L = L̃|Q and the distribution D τ = (0 τ1 τ2 ) . (48)
are invariant under the action of the group ∂ 2 Lc
The mass matrix for this system is defined as Mc = ∂ ṡ∂ ṡ ,
G̃ = {(s, R̄) ∈ R3 × S1 |R̄ê3 = ê3 }. (43) which explicitly results in
� �
Mc11 Mc12 Mc13
Proof. Under the left action of G̃, the Lagrangian (29) is Mc = Mc12 Mc22 Mc23 , (49)
given by Mc13 Mc23 Mc33
�
1 where
L= �R̄(Ṙθ Rα + Rθ Ṙα ) X + R̄ṙS �2 ρ(X)dV
2 B
� Mc11 = c3 + r2 c1 + 2rc2 cos α
1 r r2
+ �R̄(Ṙθ Rα + Rθ Ṙα )(Rφ1 X + d B ê2 ) Mc12 = Mc13 = c2 cos α + c1
2 W1 2 2
+ R̄Rθ Rα Ṙφ1 X + R̄ṙS �2 ρ(X)dV r2 r2
� Mc23 = c1 − 2 (c4 sin2 α + c5 )
1 4 4d
+ �R̄(Ṙθ Rα + Rθ Ṙα )(Rφ2 X − d B ê2 ) r2 r2
2 W2 Mc22 = Mc33 = c1 + 2 (c4 sin2 α + c5 ),
+ R̄Rθ Rα Ṙφ2 X + R̄ṙS �2 ρ(X)dV 4 4d
and with
− mb gb �I ê3 , R̄Rθ Rα B ê3 � (44) 1
� c1 = mB + 2mW + 2 2 JWyy
1 r
= �(Ṙθ Rα + Rθ Ṙα ) X + ṙS �2 ρ(X)dV
2 B c2 = mB b, c3 = mB b2 + JByy
�
1 c4 = mB b2 + JBxx − JBzz
+ �(Ṙθ Rα + Rθ Ṙα )(Rφ1 X + d B ê2 )
2 W1 d2
c5 = 2JWxx + JBzz + 2mW d2 + 2 JW .
+ Rθ Rα Ṙφ1 X + ṙS �2 ρ(X)dV r2 yy
�
1 The matrix of the Coriolis and centrifugal forces Cc can
+ �(Ṙθ Rα + Rθ Ṙα )(Rφ2 X − d B ê2 ) be derived from the Christoffel symbols (Bloch [2003]) or
2 W2
by using the following relations:
+ Rθ Rα Ṙφ2 X + ṙS �2 ρ(X)dV 1
− mb gb �I ê3 , R̄Rθ Rα B ê3 �, (45) Cc ṡ = Ṁc ṡ − ∂sT (ṡT Mc ṡ) (50)
2
T T
where it has been used the fact that R̄ R̄ = I. The Ṁc = Cc + Cc . (51)
Lagrangian coincides with (29), since R̄T I ê3 = I ê3 . It can be explicitly written as
 
The distribution D = D̃|Q is described by the equations −r c2 α̇ sin α −δ φ̇1−2 δ φ̇1−2
(38) and (39). Since Dq ⊂ Tq Q, under left group action of Cc = −r/2 c2 α̇ sin α + δ φ̇1−2 δ α̇ −δ α̇  ,
−r/2 c2 α̇ sin α − δ φ̇
G̃ the distribution D becomes 1−2 −δ α̇ δ α̇
2
R̄ṙS = ∓ dR̄(Ṙθ Rα + Rθ Ṙα ) B ê2 (46) r
where δ = 4d 2 c4 sin α cos α, φ̇1−2 = φ̇1 − φ̇2 . The forcing
T T
+ rR̄Rθ Rα (Rα R̄ R̄Ṙα + RφT1/2 Ṙφ1/2 )Rα
T T T
Rθ R̄ I ê3 term on the right hand side is
−B ∂ġT L| = Jc ṡ,
⇒ ṙS = ∓ d(Ṙθ Rα + Rθ Ṙα ) B ê2 ġ=−Aṡ
� �
T
+ rRθ Rα (Rα Ṙα + RφT1/2 Ṙφ1/2 )Rα
T T
Rθ I ê3 .  0 −β β r3
where Jc = β 0 β , β = 2 c2 (φ̇2 − φ̇1 ) sin α.
−β −β 0 4d
4.1 Constrained equations of motion of the WIP
The term corresponding to the gravitational forces is
simply the gradient of the potential
Since a curve q(t) satisfies the constraints if q̇(t) ∈ Dq , ∀t, � �
we can express the distribution given by (38) and (39) −c2 g sin α
∇s V = 0 . (52)
5 Translation, and rotation about the z-axis 0

332
 MATHMOD 2015
February 18 - 20, 2015. Vienna, Austria Sergio Delgado et al. / IFAC-PapersOnLine 48-1 (2015) 328–333 333

The equations of motion can be now written as the by Pathak et al. [2005] using the Lagrange-d’Alembert
equations of motion in the reduced (shape) space and the equations, since the assumptions made for developing the
reconstruction equation 6 model are identical.
Mc s̈ + Cc ṡ + ∇s V = τ + Jc ṡ (53)
5. CONCLUSION
ġ = −AT ṡ. (54)
In this paper we have derived the reduced equations of
4.2 Change of coordinates motion for the WIP from the Lagrange-d’Alembert Prin-
ciple and identifying inherent symmetries in the system.
Since the shape variables are not fully actuated, separat- Our immediate goal is to identify the Lie-Poisson structure
ing the equations of motion into dynamics of the shape for this nonholonomic system and employ this feature
variables, and reconstruction equations does not simplify in arriving at a suitable control law for the objective of
the control problem, nor any controllability analysis or stabilization and, later on, trajectory tracking.
trajectory planing. Thus, we want to get a functional
relationship between the inputs and the group variables ACKNOWLEDGEMENTS
instead. We also need to consider the tilting angle α, for
this variable is critical for the stability of the WIP. Let us A part of this work was performed during the first author’s
consider the �relation visit to the Systems and Control group, IIT Bombay.
r � r � � The first author thanks the TUM Graduate School for
v= φ̇1 + φ̇2 + 2α̇ , θ̇ = φ̇2 − φ̇1 supporting this work.
2 2d
to get the equations of motion in terms of the forward
REFERENCES
acceleration v̇ of the WIP and the yawing rotation θ̇. Let us
introduce the velocities ξ˙ = (v, α̇, θ̇)T , such that ξ̇ = T −1 ṡ, Baloh, M. and Parent, M. (2003). Modeling and model
for the constant matrix verification of an intelligent self-balancing two-wheeled
� � vehicle for an autonomous urban transportation system.
0 1 0
T = /r −1 /r .
1 −d (55) In Proc. Conf. Comp. Intellig., Robot. Autonom. Syst.
1/r −1 d/r Bloch, A.M. (2003). Nonholonomic mechanics and con-
trol. Interdisciplinary applied mathematics: Systems
In the new coordinates ξ, the equations of motion are and control. Springer.
M ξ̈ + C ξ˙ + ∇ξ V = τ̃ + J ξ˙ (56) Bloch, A.M., Krishnaprasad, P.S., Marsden, J.E., and
ġ = −AT T ξ̇. (57) Murray, R.M. (1996). Nonholonomic mechanical sys-
tems with symmetry. Archive for Rational Mechanics
The Mass matrix considerably simplifies to and Analysis, 136(1), 21–99.
  Bloch, A.M., Marsden, J.E., and Zenkov, D.V. (2009).
c1 c2 cos α 0
Quasivelocities and symmetries in non-holonomic sys-
M = T T Mc T = c2 cos α c3 0  , (58)
2 tems. Dynamical Systems, 24(2), 187–222.
0 0 c4 sin α + c5
Brockett, R.W. (1983). Asymptotic Stability and Feedback
such that the matrix of the Coriolis and centrifugal forces Stabilization. Defense Technical Information Center.
becomes   Gajbhiye, S. and Banavar, R.N. (2012). The Euler-
0 −c2 α̇ sin α 0 Poincaré equations for a spherical robot actuated by a
C = 0 0 −c4 θ̇ sin α cos α , (59) pendulum. In Proc. 4th IFAC Workshop on Lagrangian
0 c4 θ̇ sin α cos α c4 α̇ sin α cos α and Hamiltonian Methods for Non Linear Control.
the term corresponding to the gravitational forces is Grasser, F., D’Arrigo, A., Colombi, S., and Rufer, A.C.
� � (2002). Joe: a mobile, inverted pendulum. IEEE
0 Transactions on Industrial Electronics, 49(1), 107–114.
∇ξ V = −c2 g sin α , (60) Holm, D.D., Schmah, T., and Stoica, C. (2009). Geometric
0 Mechanics and Symmetry. Oxford University Press Inc.,
and the forcing term on the right hand side is New York.
 
0 0 c2 θ̇ sin α Li, Z., Yang, C., and Fan, L. (2012). Advanced Control of
J ξ̇ =  0 0 0 ˙
 ξ. (61) Wheeled Inverted Pendulum Systems. Springer.
−c2 θ̇ sin α 0 0 Marsden, J.E. and Ratiu, T.S. (1994). Introduction to
Mechanics and Symmetry. Springer-Verlag, New York.
In the new coordinates ξ, the systems input is given by Nasrallah, D., Michalska, H., and Angeles, J. (2007). Con-
�1 �
/r (τ1 + τ2 ) trollability and posture control of a wheeled pendulum
T
τ̃ = T τ = −(τ1 + τ2 ) , (62) moving on an inclined plane. Robotics, IEEE Transac-
d/r (τ2 − τ1 )
tions on, 23(3), 564–577.
and the reconstruction equation ġ = −AT T ξ̇ for the Ostrowski, J.P. (1999). Computing reduced equations for
position is simply robotic systems with constraints and symmetries. IEEE
T. Robotics and Automation, 15(1), 111–123.
ẋ = v cos θ ẏ = v sin θ. Pathak, K., Franch, J., and Agrawal, S. (2005). Velocity
Note that the new coordinates ξ consist of the length of and position control of a wheeled inverted pendulum by
the path, and the tilting and yaw angles, α and θ, respec- partial feedback linearization. Robotics, IEEE Transac-
tively. The equations of motion (56) are the same derived tions on, 21(3), 505–513.
6 There is no momentum equation for this system. Segway (2015, Jan). [online]. http://www.segway.com.

333