ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

M ODULE 6 - DYNAMICS OF S ERIAL AND PARALLEL ROBOTS

Ashitava Ghosal1
1 Department

of Mechanical Engineering
&
Centre for Product Design and Manufacture
Indian Institute of Science
Bangalore 560 012, India
Email: [email protected]

NPTEL, 2010

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

1 / 74

.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

2 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

3 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Position and velocity kinematics cause of motion not considered.

Dynamics motion of links of a robot due to external forces and/or
moments.
Main assumption: All links are rigid no deformation.
Motion of links described by ordinary differential equations (ODEs)
also called equations of motion.
Several methods to derive the equations of motion Newton-Euler,
Lagrangian and Kanes methods most well known.
Newton-Euler obtain linear and angular velocities and accelerations of
each link, free-body diagrams, and Newtons law and Euler equations.
Lagrangian formulation obtain kinetic and potential energy of each
link, obtain the scalar Lagrangian, and take partial and ordinary
derivatives.
Kanes formulation choose generalised coordinates and speeds, obtain
generalised active and inertia forces, and equate the active and inertia
forces.

Each formulation has its advantages and disadvantages.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

4 / 74

I NTRODUCTION
OVERVIEW

Two main problems in robot dynamics:

Direct problem obtain motion of links given the applied external
forces/moments.
Inverse problem obtain joint torques/forces required for a desired
motion of links.

Direct problem involves solution of ODEs Simulation.

Inverse dynamics for sizing of actuators and other components, and
for advanced model based control schemes (see Module 7, Lecture 3).
Computational efficiency of inverse and direct problem is of interest.
Aim is to develop efficient O(N) or O(logN) N is the number of
links (for parallel computing) algorithms for use in protein folding and
in computational biology (see Klepeis et al. 2002).
Dynamics of parallel manipulators complicated by presence of
closed-loops typically give rise to differential-algebraic equations
(DAEs) more difficult to solve.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

5 / 74

I NTRODUCTION
OVERVIEW

Two main problems in robot dynamics:

Direct problem obtain motion of links given the applied external
forces/moments.
Inverse problem obtain joint torques/forces required for a desired
motion of links.

Direct problem involves solution of ODEs Simulation.

Inverse dynamics for sizing of actuators and other components, and
for advanced model based control schemes (see Module 7, Lecture 3).
Computational efficiency of inverse and direct problem is of interest.
Aim is to develop efficient O(N) or O(logN) N is the number of
links (for parallel computing) algorithms for use in protein folding and
in computational biology (see Klepeis et al. 2002).
Dynamics of parallel manipulators complicated by presence of
closed-loops typically give rise to differential-algebraic equations
(DAEs) more difficult to solve.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

5 / 74

I NTRODUCTION
OVERVIEW

Two main problems in robot dynamics:

Direct problem obtain motion of links given the applied external
forces/moments.
Inverse problem obtain joint torques/forces required for a desired
motion of links.

Direct problem involves solution of ODEs Simulation.

Inverse dynamics for sizing of actuators and other components, and
for advanced model based control schemes (see Module 7, Lecture 3).
Computational efficiency of inverse and direct problem is of interest.
Aim is to develop efficient O(N) or O(logN) N is the number of
links (for parallel computing) algorithms for use in protein folding and
in computational biology (see Klepeis et al. 2002).
Dynamics of parallel manipulators complicated by presence of
closed-loops typically give rise to differential-algebraic equations
(DAEs) more difficult to solve.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

5 / 74

I NTRODUCTION
OVERVIEW

Two main problems in robot dynamics:

Direct problem obtain motion of links given the applied external
forces/moments.
Inverse problem obtain joint torques/forces required for a desired
motion of links.

Direct problem involves solution of ODEs Simulation.

Inverse dynamics for sizing of actuators and other components, and
for advanced model based control schemes (see Module 7, Lecture 3).
Computational efficiency of inverse and direct problem is of interest.
Aim is to develop efficient O(N) or O(logN) N is the number of
links (for parallel computing) algorithms for use in protein folding and
in computational biology (see Klepeis et al. 2002).
Dynamics of parallel manipulators complicated by presence of
closed-loops typically give rise to differential-algebraic equations
(DAEs) more difficult to solve.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

5 / 74

I NTRODUCTION
OVERVIEW

Two main problems in robot dynamics:

Direct problem obtain motion of links given the applied external
forces/moments.
Inverse problem obtain joint torques/forces required for a desired
motion of links.

Direct problem involves solution of ODEs Simulation.

Inverse dynamics for sizing of actuators and other components, and
for advanced model based control schemes (see Module 7, Lecture 3).
Computational efficiency of inverse and direct problem is of interest.
Aim is to develop efficient O(N) or O(logN) N is the number of
links (for parallel computing) algorithms for use in protein folding and
in computational biology (see Klepeis et al. 2002).
Dynamics of parallel manipulators complicated by presence of
closed-loops typically give rise to differential-algebraic equations
(DAEs) more difficult to solve.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

5 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

0
Z

Mass m of
the rigid body is

given by V dV , is density.

{0}
dV

Inertia of a rigid body

distribution of mass.
0

0
Y

Inertia tensor 0 [I ]

Ixx
0

[I ] = Ixy
Ixz

in {0}
Ixy
Iyy
Iyz

Ixz
Iyz
Izz

Rigid Body
0
X
Figure 1: Mass and inertia of a rigid body
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

6 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

Elements
of the inertia tensor

Ixx = V (y 2 + z 2 ) dV , Ixy = V xy dV , Ixz = V xz dV

Iyy = V (x 2 + z 2 ) dV , Iyz = V yz dV , Izz = V (x 2 + y 2 ) dV
The inertia tensor is positive definite and symmetric Eigenvalues of
are real and positive.

0 [I ]

Three eigenvalues are the principal moments of inertias and the

associated eigenvectors are the principal axes.
The inertia tensor in {A}, with OA coincident O0 , is given by
A [I ] = A [R]0 [I ]A [R]T .
0
0
To obtain inertia tensor for a link i,
Coordinate system, {Ci }, is chosen at the centre of mass of the link.
{Ci } is parallel {i} (see Module 2, Lecture 2).
Sub-divide complex link into simple shapes and use parallel axis
theorem.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

7 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

Elements
of the inertia tensor

Ixx = V (y 2 + z 2 ) dV , Ixy = V xy dV , Ixz = V xz dV

Iyy = V (x 2 + z 2 ) dV , Iyz = V yz dV , Izz = V (x 2 + y 2 ) dV
The inertia tensor is positive definite and symmetric Eigenvalues of
are real and positive.

0 [I ]

Three eigenvalues are the principal moments of inertias and the

associated eigenvectors are the principal axes.
The inertia tensor in {A}, with OA coincident O0 , is given by
A [I ] = A [R]0 [I ]A [R]T .
0
0
To obtain inertia tensor for a link i,
Coordinate system, {Ci }, is chosen at the centre of mass of the link.
{Ci } is parallel {i} (see Module 2, Lecture 2).
Sub-divide complex link into simple shapes and use parallel axis
theorem.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

7 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

Elements
of the inertia tensor

Ixx = V (y 2 + z 2 ) dV , Ixy = V xy dV , Ixz = V xz dV

Iyy = V (x 2 + z 2 ) dV , Iyz = V yz dV , Izz = V (x 2 + y 2 ) dV
The inertia tensor is positive definite and symmetric Eigenvalues of
are real and positive.

0 [I ]

Three eigenvalues are the principal moments of inertias and the

associated eigenvectors are the principal axes.
The inertia tensor in {A}, with OA coincident O0 , is given by
A [I ] = A [R]0 [I ]A [R]T .
0
0
To obtain inertia tensor for a link i,
Coordinate system, {Ci }, is chosen at the centre of mass of the link.
{Ci } is parallel {i} (see Module 2, Lecture 2).
Sub-divide complex link into simple shapes and use parallel axis
theorem.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

7 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

Elements
of the inertia tensor

Ixx = V (y 2 + z 2 ) dV , Ixy = V xy dV , Ixz = V xz dV

Iyy = V (x 2 + z 2 ) dV , Iyz = V yz dV , Izz = V (x 2 + y 2 ) dV
The inertia tensor is positive definite and symmetric Eigenvalues of
are real and positive.

0 [I ]

Three eigenvalues are the principal moments of inertias and the

associated eigenvectors are the principal axes.
The inertia tensor in {A}, with OA coincident O0 , is given by
A [I ] = A [R]0 [I ]A [R]T .
0
0
To obtain inertia tensor for a link i,
Coordinate system, {Ci }, is chosen at the centre of mass of the link.
{Ci } is parallel {i} (see Module 2, Lecture 2).
Sub-divide complex link into simple shapes and use parallel axis
theorem.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

7 / 74

I NTRODUCTION
M ASS AND I NERTIA OF A L INK

Elements
of the inertia tensor

Ixx = V (y 2 + z 2 ) dV , Ixy = V xy dV , Ixz = V xz dV

Iyy = V (x 2 + z 2 ) dV , Iyz = V yz dV , Izz = V (x 2 + y 2 ) dV
The inertia tensor is positive definite and symmetric Eigenvalues of
are real and positive.

0 [I ]

Three eigenvalues are the principal moments of inertias and the

associated eigenvectors are the principal axes.
The inertia tensor in {A}, with OA coincident O0 , is given by
A [I ] = A [R]0 [I ]A [R]T .
0
0
To obtain inertia tensor for a link i,
Coordinate system, {Ci }, is chosen at the centre of mass of the link.
{Ci } is parallel {i} (see Module 2, Lecture 2).
Sub-divide complex link into simple shapes and use parallel axis
theorem.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

7 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

8 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY

Energy base formulation involving kinetic and potential energy

The kinetic energy of link i with mass mi and inertia 0 [I ]i
KEi =

1
1
mi 0 VCi 0 VCi + 0 i 0 [I ]i 0 i
2
2

First and second term from linear velocity of the links centre of mass
and angular velocity of link.
0
VCi and 0 i are the linear and angular velocities of the centre of mass
and link {i}, respectively.
0

VCi = 0i [R]i VCi ,

i = 0i [R]i i

Using above equations

K .Ei =

1
1
mi i VCi i VCi + i i Ci [I ]i i i
2
2
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(1)
.

NPTEL, 2010

9 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY

Energy base formulation involving kinetic and potential energy

The kinetic energy of link i with mass mi and inertia 0 [I ]i
KEi =

1
1
mi 0 VCi 0 VCi + 0 i 0 [I ]i 0 i
2
2

First and second term from linear velocity of the links centre of mass
and angular velocity of link.
0
VCi and 0 i are the linear and angular velocities of the centre of mass
and link {i}, respectively.
0

VCi = 0i [R]i VCi ,

i = 0i [R]i i

Using above equations

K .Ei =

1
1
mi i VCi i VCi + i i Ci [I ]i i i
2
2
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(1)
.

NPTEL, 2010

9 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY

Energy base formulation involving kinetic and potential energy

The kinetic energy of link i with mass mi and inertia 0 [I ]i
KEi =

1
1
mi 0 VCi 0 VCi + 0 i 0 [I ]i 0 i
2
2

First and second term from linear velocity of the links centre of mass
and angular velocity of link.
0
VCi and 0 i are the linear and angular velocities of the centre of mass
and link {i}, respectively.
0

VCi = 0i [R]i VCi ,

i = 0i [R]i i

Using above equations

K .Ei =

1
1
mi i VCi i VCi + i i Ci [I ]i i i
2
2
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(1)
.

NPTEL, 2010

9 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY (C ONTD .)

Velocity propagation formulas for serial manipulators (see Module 5,

Lecture 1)

i
i
i
i
VCi
i

=
=
=

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
joint i is
i1
i1 [R]
i
i
i
Vi + i pCi

joint i is rotary
prismatic

locates the centre of mass of link {i} with respect to Oi .

i : 0 n to obtain kinetic energy of all links in a serial manipulator.
In parallel manipulators, several loops no propagation formulas.
Easier to compute angular and linear velocities using derivatives1
ip

Ci

0
0p
Ci

0 [R]T ,
i = 0i [R]
i

VCi =

d 0
( pCi )
dt

(2)

is the position vector of the centre of mass of link i.

1 In

closed-loop mechanisms or parallel manipulators, one can judiciously use the

propagation formulas for the serial portions.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

10 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY (C ONTD .)

Velocity propagation formulas for serial manipulators (see Module 5,

Lecture 1)

i
i
i
i
VCi
i

=
=
=

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
joint i is
i1
i1 [R]
i
i
i
Vi + i pCi

joint i is rotary
prismatic

locates the centre of mass of link {i} with respect to Oi .

i : 0 n to obtain kinetic energy of all links in a serial manipulator.
In parallel manipulators, several loops no propagation formulas.
Easier to compute angular and linear velocities using derivatives1
ip

Ci

0
0p
Ci

0 [R]T ,
i = 0i [R]
i

VCi =

d 0
( pCi )
dt

(2)

is the position vector of the centre of mass of link i.

1 In

closed-loop mechanisms or parallel manipulators, one can judiciously use the

propagation formulas for the serial portions.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

10 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY (C ONTD .)

Velocity propagation formulas for serial manipulators (see Module 5,

Lecture 1)

i
i
i
i
VCi
i

=
=
=

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
joint i is
i1
i1 [R]
i
i
i
Vi + i pCi

joint i is rotary
prismatic

locates the centre of mass of link {i} with respect to Oi .

i : 0 n to obtain kinetic energy of all links in a serial manipulator.
In parallel manipulators, several loops no propagation formulas.
Easier to compute angular and linear velocities using derivatives1
ip

Ci

0
0p
Ci

0 [R]T ,
i = 0i [R]
i

VCi =

d 0
( pCi )
dt

(2)

is the position vector of the centre of mass of link i.

1 In

closed-loop mechanisms or parallel manipulators, one can judiciously use the

propagation formulas for the serial portions.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

10 / 74

L AGRANGIAN FORMULATION
K INETIC ENERGY (C ONTD .)

Velocity propagation formulas for serial manipulators (see Module 5,

Lecture 1)

i
i
i
i
VCi
i

=
=
=

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
joint i is
i1
i1 [R]
i
i
i
Vi + i pCi

joint i is rotary
prismatic

locates the centre of mass of link {i} with respect to Oi .

i : 0 n to obtain kinetic energy of all links in a serial manipulator.
In parallel manipulators, several loops no propagation formulas.
Easier to compute angular and linear velocities using derivatives1
ip

Ci

0
0p
Ci

0 [R]T ,
i = 0i [R]
i

VCi =

d 0
( pCi )
dt

(2)

is the position vector of the centre of mass of link i.

1 In

closed-loop mechanisms or parallel manipulators, one can judiciously use the

propagation formulas for the serial portions.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

10 / 74

L AGRANGIAN FORMULATION
P OTENTIAL ENERGY

Assumption: Potential energy due to gravity alone2 .

General expression for potential energy due to gravity
PEi = mi 0 g 0 pCi

(3)

g gravity vector of magnitude 9.81m/sec2

0
g along vertical direction denoted by 0 ^
Z axis.
0
pCi location of the centre of mass of link i from the zero or
reference potential energy surface.
Constant value of the reference potential energy does not matter as
derivatives are taken.
2 If

springs or other energy storage devices are present, appropriate modification to

the expression of the potential energy of the link can be done. For example, if torsional
springs are present at joint i, add a term of the form 21 ki i 2 to the expression for the
potential energy.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

11 / 74

L AGRANGIAN FORMULATION
P OTENTIAL ENERGY

Assumption: Potential energy due to gravity alone2 .

General expression for potential energy due to gravity
PEi = mi 0 g 0 pCi

(3)

g gravity vector of magnitude 9.81m/sec2

0
g along vertical direction denoted by 0 ^
Z axis.
0
pCi location of the centre of mass of link i from the zero or
reference potential energy surface.
Constant value of the reference potential energy does not matter as
derivatives are taken.
2 If

springs or other energy storage devices are present, appropriate modification to

the expression of the potential energy of the link can be done. For example, if torsional
springs are present at joint i, add a term of the form 21 ki i 2 to the expression for the
potential energy.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

11 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

From the kinetic and potential energy, define the scalar Lagrangian
N

= (KEi PEi )
L (q, q)

(4)

N is the number of links excluding the fixed link.

In serial manipulators, with R or P joints dim(q) = n = N
The equations of motion3 , are
(
)
d L
L

= Qi
i = 1, 2, ..., n
dt q i
qi

(5)

Qi s are the externally applied generalised forces when only joint

torques or forces are present
Qi = i ,

i = 1, ..., n

(6)

3 The Lagrangian formulation is equivalent to Newtons laws and Eulers equations

from calculus of variation(see Goldstein 1980).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

12 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

From the kinetic and potential energy, define the scalar Lagrangian
N

= (KEi PEi )
L (q, q)

(4)

N is the number of links excluding the fixed link.

In serial manipulators, with R or P joints dim(q) = n = N
The equations of motion3 , are
(
)
d L
L

= Qi
i = 1, 2, ..., n
dt q i
qi

(5)

Qi s are the externally applied generalised forces when only joint

torques or forces are present
Qi = i ,

i = 1, ..., n

(6)

3 The Lagrangian formulation is equivalent to Newtons laws and Eulers equations

from calculus of variation(see Goldstein 1980).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

12 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

From the kinetic and potential energy, define the scalar Lagrangian
N

= (KEi PEi )
L (q, q)

(4)

N is the number of links excluding the fixed link.

In serial manipulators, with R or P joints dim(q) = n = N
The equations of motion3 , are
(
)
d L
L
= Qi
i = 1, 2, ..., n

dt q i
qi

(5)

Qi s are the externally applied generalised forces when only joint

torques or forces are present
Qi = i ,

i = 1, ..., n

(6)

3 The Lagrangian formulation is equivalent to Newtons laws and Eulers equations

from calculus of variation(see Goldstein 1980).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

12 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

From the kinetic and potential energy, define the scalar Lagrangian
N

= (KEi PEi )
L (q, q)

(4)

N is the number of links excluding the fixed link.

In serial manipulators, with R or P joints dim(q) = n = N
The equations of motion3 , are
(
)
d L
L
= Qi
i = 1, 2, ..., n

dt q i
qi

(5)

Qi s are the externally applied generalised forces when only joint

torques or forces are present
Qi = i ,

i = 1, ..., n

(6)

3 The Lagrangian formulation is equivalent to Newtons laws and Eulers equations

from calculus of variation(see Goldstein 1980).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

12 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

After performing the derivatives, equation of motion of a serial

manipulator takes the form
q + G(q) =
[M(q)]
q + [C(q, q)]

(7)

[M(q)] n n mass matrix4 ,

n n matrix and [C(q, q)]
q is an n 1 vector of centripetal
[C(q, q)]
and Coriolis terms contains only quadratic qi qj terms,
G(q) n 1 vector of gravity terms, and
n 1 vector of joint torques or forces.

All serial manipulator equations of motion can be written in the above

form!!
4 For R joints, elements of [M(q)] have units of inertia kg m2 . For P joint, elements
of [M(q)] have units of mass Kg.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

13 / 74

L AGRANGIAN FORMULATION
E QUATIONS OF MOTION

After performing the derivatives, equation of motion of a serial

manipulator takes the form
q + G(q) =
[M(q)]
q + [C(q, q)]

(7)

[M(q)] n n mass matrix4 ,

n n matrix and [C(q, q)]
q is an n 1 vector of centripetal
[C(q, q)]
and Coriolis terms contains only quadratic qi qj terms,
G(q) n 1 vector of gravity terms, and
n 1 vector of joint torques or forces.

All serial manipulator equations of motion can be written in the above

form!!
4 For R joints, elements of [M(q)] have units of inertia kg m2 . For P joint, elements
of [M(q)] have units of mass Kg.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

13 / 74

L AGRANGIAN FORMULATION
P ROPERTIES OF TERMS IN EQUATIONS OF MOTION

Mass matrix, [M(q)], always positive definite and symmetric.

Total kinetic energy of a serial manipulator is
1
KE = q T [M(q)]q
(8)
2
= 0 and zero only when |q|
= 0 [M(q)] is positive
KE 0 for |q|
definite.
Eigenvalues
Inertia cannot be imaginary along any component of q
of [M(q)] must be real [M(q)] must be symmetric.

The centripetal and Coriolis terms can be obtained from the mass
matrix as
(
)
Mij Mik Mkj
1 n
+

Cij =
q k
(9)
2 k=1
qk
qj
qi
The gravity terms can be obtained from the potential energy as
Gi =

(PE )
qi

(10)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

14 / 74

L AGRANGIAN FORMULATION
P ROPERTIES OF TERMS IN EQUATIONS OF MOTION

Mass matrix, [M(q)], always positive definite and symmetric.

Total kinetic energy of a serial manipulator is
1
KE = q T [M(q)]q
(8)
2
= 0 and zero only when |q|
= 0 [M(q)] is positive
KE 0 for |q|
definite.
Eigenvalues
Inertia cannot be imaginary along any component of q
of [M(q)] must be real [M(q)] must be symmetric.

The centripetal and Coriolis terms can be obtained from the mass
matrix as
(
)
Mij Mik Mkj
1 n
Cij =
+

q k
(9)
2 k=1
qk
qj
qi
The gravity terms can be obtained from the potential energy as
Gi =

(PE )
qi

(10)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

14 / 74

L AGRANGIAN FORMULATION
P ROPERTIES OF TERMS IN EQUATIONS OF MOTION

Mass matrix, [M(q)], always positive definite and symmetric.

Total kinetic energy of a serial manipulator is
1
KE = q T [M(q)]q
(8)
2
= 0 and zero only when |q|
= 0 [M(q)] is positive
KE 0 for |q|
definite.
Eigenvalues
Inertia cannot be imaginary along any component of q
of [M(q)] must be real [M(q)] must be symmetric.

The centripetal and Coriolis terms can be obtained from the mass
matrix as
(
)
Mij Mik Mkj
1 n
Cij =
+

q k
(9)
2 k=1
qk
qj
qi
The gravity terms can be obtained from the potential energy as
Gi =

(PE )
qi

(10)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

14 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS

Presence of m loop-closure constraint equations (see Module 4,

Lecture 1)
i (q) = 0, i = 1, 2, ..., m

(11)

q n+m n actuated joint variables, , and m passive joint

variables .
To obtain equations of motion for a system with constraints
Lagrange multipliers (Goldstein 1980, Haug 1989).
Lagrangian written as
m

= L (q, q)
j j (q)
L(q, q)

(12)

j=1

m j s introduced (unknown) Lagrange multipliers.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

15 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS

Presence of m loop-closure constraint equations (see Module 4,

Lecture 1)
i (q) = 0, i = 1, 2, ..., m

(11)

q n+m n actuated joint variables, , and m passive joint

variables .
To obtain equations of motion for a system with constraints
Lagrange multipliers (Goldstein 1980, Haug 1989).
Lagrangian written as
m

= L (q, q)
j j (q)
L(q, q)

(12)

j=1

m j s introduced (unknown) Lagrange multipliers.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

15 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS

Presence of m loop-closure constraint equations (see Module 4,

Lecture 1)
i (q) = 0, i = 1, 2, ..., m

(11)

q n+m n actuated joint variables, , and m passive joint

variables .
To obtain equations of motion for a system with constraints
Lagrange multipliers (Goldstein 1980, Haug 1989).
Lagrangian written as
m

= L (q, q)
j j (q)
L(q, q)

(12)

j=1

m j s introduced (unknown) Lagrange multipliers.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

15 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS

Presence of m loop-closure constraint equations (see Module 4,

Lecture 1)
i (q) = 0, i = 1, 2, ..., m

(11)

q n+m n actuated joint variables, , and m passive joint

variables .
To obtain equations of motion for a system with constraints
Lagrange multipliers (Goldstein 1980, Haug 1989).
Lagrangian written as
m

= L (q, q)
j j (q)
L(q, q)

(12)

j=1

m j s introduced (unknown) Lagrange multipliers.

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

15 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

m constraints, j (q) = 0, are holonomic only functions of q.

For holonomic constraints, equations of motion are
(
)
m
j (q)
d L
L

= i + j
i = 1, 2, ..., n + m
dt q i
qi
qi
j=1

(13)

In matrix form,
q + G(q) = + [(q)]T
[M(q)]
q + [C(q, q)]

(14)

is the m 1 vector of unknown Lagrange multipliers

Constraint matrix [(q)] is obtained from the partial derivatives of m
constraint equations with respect to qi
Concatenation [] = [ [K ] | [K ] ] (see Module 5, Lecture 3 )
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

16 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

m constraints, j (q) = 0, are holonomic only functions of q.

For holonomic constraints, equations of motion are
(
)
m
j (q)
d L
L

= i + j
i = 1, 2, ..., n + m
dt q i
qi
qi
j=1

(13)

In matrix form,
q + G(q) = + [(q)]T
[M(q)]
q + [C(q, q)]

(14)

is the m 1 vector of unknown Lagrange multipliers

Constraint matrix [(q)] is obtained from the partial derivatives of m
constraint equations with respect to qi
Concatenation [] = [ [K ] | [K ] ] (see Module 5, Lecture 3 )
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

16 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

m constraints, j (q) = 0, are holonomic only functions of q.

For holonomic constraints, equations of motion are
(
)
m
j (q)
d L
L

= i + j
i = 1, 2, ..., n + m
dt q i
qi
qi
j=1

(13)

In matrix form,
q + G(q) = + [(q)]T
[M(q)]
q + [C(q, q)]

(14)

is the m 1 vector of unknown Lagrange multipliers

Constraint matrix [(q)] is obtained from the partial derivatives of m
constraint equations with respect to qi
Concatenation [] = [ [K ] | [K ] ] (see Module 5, Lecture 3 )
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

16 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

To determine
Twice differentiate m constraint equations with respect to t

[(q)]
q + [(q)]
q = 0

(15)

is a m (n + m) matrix containing the time derivatives of each of

[]
the elements of [].
Since the mass matrix [M(q)] is always invertible
= [M]1 ( [C]q G) + [M]1 []T
q
in equation (15)
Substituting q
q + [][M]1 ( [C]q G)}
= ([][M]1 []T )1 {[]

(16)

Substitute back into equation (14) equations of motion

q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
f denotes ( [C]q G).
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(17)
.

NPTEL, 2010

17 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

To determine
Twice differentiate m constraint equations with respect to t

[(q)]
q + [(q)]
q = 0

(15)

is a m (n + m) matrix containing the time derivatives of each of

[]
the elements of [].
Since the mass matrix [M(q)] is always invertible
= [M]1 ( [C]q G) + [M]1 []T
q
in equation (15)
Substituting q
q + [][M]1 ( [C]q G)}
= ([][M]1 []T )1 {[]

(16)

Substitute back into equation (14) equations of motion

q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
f denotes ( [C]q G).
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(17)
.

NPTEL, 2010

17 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

Mass matrix [M] is (n + m) (n + m), positive definite and symmetric

matrix
Centripetal/Coriolis terms and the gravity terms are (n + m) 1
vectors.
[(q)]T has units of torque/force constraint forces/torques.
Work done by constraint forces [ [(q)]T ]T q T [(q)]q
[(q)]q = 0 from definition of constraint matrix

Useful to obtain constraint forces/torques for mechanical design of the

joints and links
Most multi-body dynamics software packages (see, for example,
ADAMS 2002) compute and provide the constraint forces and torques.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

18 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

Mass matrix [M] is (n + m) (n + m), positive definite and symmetric

matrix
Centripetal/Coriolis terms and the gravity terms are (n + m) 1
vectors.
[(q)]T has units of torque/force constraint forces/torques.
Work done by constraint forces [ [(q)]T ]T q T [(q)]q
[(q)]q = 0 from definition of constraint matrix

Useful to obtain constraint forces/torques for mechanical design of the

joints and links
Most multi-body dynamics software packages (see, for example,
ADAMS 2002) compute and provide the constraint forces and torques.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

18 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

Mass matrix [M] is (n + m) (n + m), positive definite and symmetric

matrix
Centripetal/Coriolis terms and the gravity terms are (n + m) 1
vectors.
[(q)]T has units of torque/force constraint forces/torques.
Work done by constraint forces [ [(q)]T ]T q T [(q)]q
[(q)]q = 0 from definition of constraint matrix

Useful to obtain constraint forces/torques for mechanical design of the

joints and links
Most multi-body dynamics software packages (see, for example,
ADAMS 2002) compute and provide the constraint forces and torques.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

18 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

Mass matrix [M] is (n + m) (n + m), positive definite and symmetric

matrix
Centripetal/Coriolis terms and the gravity terms are (n + m) 1
vectors.
[(q)]T has units of torque/force constraint forces/torques.
Work done by constraint forces [ [(q)]T ]T q T [(q)]q
[(q)]q = 0 from definition of constraint matrix

Useful to obtain constraint forces/torques for mechanical design of the

joints and links
Most multi-body dynamics software packages (see, for example,
ADAMS 2002) compute and provide the constraint forces and torques.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

18 / 74

L AGRANGIAN FORMULATION
PARALLEL MANIPULATORS (C ONTD .)

Mass matrix [M] is (n + m) (n + m), positive definite and symmetric

matrix
Centripetal/Coriolis terms and the gravity terms are (n + m) 1
vectors.
[(q)]T has units of torque/force constraint forces/torques.
Work done by constraint forces [ [(q)]T ]T q T [(q)]q
[(q)]q = 0 from definition of constraint matrix

Useful to obtain constraint forces/torques for mechanical design of the

joints and links
Most multi-body dynamics software packages (see, for example,
ADAMS 2002) compute and provide the constraint forces and torques.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

18 / 74

L AGRANGIAN FORMULATION
N ON - HOLONOMIC CONSTRAINTS

In mobile robots and few other mechanical systems, constraints are

non-holonomic, non-integrable constraints
Constraints can also contain explicit functions of time.
Lagrangian formulation for such systems
General constraints in the so-called Pfaffian form
(t) + [(q)]q = 0

(18)

Differentiate to get
q + (t)

[]
q + []
=0

Equations of motion
q}

(19)
[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + (t)
+ []

given by
q + [][M]1 ( [C]q G)}

= ([][M]1 []T )1 {(t)

+ []
(20)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

19 / 74

L AGRANGIAN FORMULATION
N ON - HOLONOMIC CONSTRAINTS

In mobile robots and few other mechanical systems, constraints are

non-holonomic, non-integrable constraints
Constraints can also contain explicit functions of time.
Lagrangian formulation for such systems
General constraints in the so-called Pfaffian form
(t) + [(q)]q = 0

(18)

Differentiate to get
q + (t)

[]
q + []
=0

Equations of motion
q}

(19)
[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + (t)
+ []

given by
q + [][M]1 ( [C]q G)}

= ([][M]1 []T )1 {(t)

+ []
(20)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

19 / 74

L AGRANGIAN FORMULATION
N ON - HOLONOMIC CONSTRAINTS

In mobile robots and few other mechanical systems, constraints are

non-holonomic, non-integrable constraints
Constraints can also contain explicit functions of time.
Lagrangian formulation for such systems
General constraints in the so-called Pfaffian form
(t) + [(q)]q = 0

(18)

Differentiate to get
q + (t)

[]
q + []
=0

Equations of motion
q}

(19)
[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + (t)
+ []

given by
q + [][M]1 ( [C]q G)}

= ([][M]1 []T )1 {(t)

+ []
(20)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

19 / 74

L AGRANGIAN FORMULATION
N ON - HOLONOMIC CONSTRAINTS

In mobile robots and few other mechanical systems, constraints are

non-holonomic, non-integrable constraints
Constraints can also contain explicit functions of time.
Lagrangian formulation for such systems
General constraints in the so-called Pfaffian form
(t) + [(q)]q = 0

(18)

Differentiate to get
q + (t)

[]
q + []
=0

Equations of motion
q}

(19)
[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + (t)
+ []

given by
q + [][M]1 ( [C]q G)}

= ([][M]1 []T )1 {(t)

+ []
(20)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

19 / 74

L AGRANGIAN FORMULATION
N ON - HOLONOMIC CONSTRAINTS

In mobile robots and few other mechanical systems, constraints are

non-holonomic, non-integrable constraints
Constraints can also contain explicit functions of time.
Lagrangian formulation for such systems
General constraints in the so-called Pfaffian form
(t) + [(q)]q = 0

(18)

Differentiate to get
q + (t)

[]
q + []
=0

Equations of motion
q}

(19)
[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + (t)
+ []

given by
q + [][M]1 ( [C]q G)}

= ([][M]1 []T )1 {(t)

+ []
(20)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

19 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

L AGRANGIAN FORMULATION
S UMMARY

Equations of motion for serial, parallel manipulators and multi-body

system (even with non-holonomic constraints) can be obtained using
Lagrangian formulation.
Equations of motion obtained using Lagrangian formulation does not
contain friction or any other dissipative term.
Friction accommodated in ad-hoc manner in the right-hand side at
appropriate place.
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)

(21)

looks similar to centripetal/Coriolis term but

Friction term, F(q, q),
actually very different!!
Typical friction = sum of a constant (Coulomb friction) + term
proportional to q (viscous damping).
Equations of motion does not contain effects of flexibility, backlash,
and other unmodeled dynamics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

20 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE

Equations of motion in joint space function of q and q
Equations of motion in terms of position and orientation of
end-effector (Khatib 1987).
Useful for Cartesian space motion and force control.
Equations of motion given as
+ GX (q)
F = [MX (q)]X + CX (q, q)

(22)

F is a 6 1 entity of forces and moments acting on the end-effector,

X is 6 1 entity representing the position and orientation of the
end-effector.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

21 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE

Equations of motion in joint space function of q and q
Equations of motion in terms of position and orientation of
end-effector (Khatib 1987).
Useful for Cartesian space motion and force control.
Equations of motion given as
+ GX (q)
F = [MX (q)]X + CX (q, q)

(22)

F is a 6 1 entity of forces and moments acting on the end-effector,

X is 6 1 entity representing the position and orientation of the
end-effector.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

21 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE

Equations of motion in joint space function of q and q
Equations of motion in terms of position and orientation of
end-effector (Khatib 1987).
Useful for Cartesian space motion and force control.
Equations of motion given as
+ GX (q)
F = [MX (q)]X + CX (q, q)

(22)

F is a 6 1 entity of forces and moments acting on the end-effector,

X is 6 1 entity representing the position and orientation of the
end-effector.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

21 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE

Equations of motion in joint space function of q and q
Equations of motion in terms of position and orientation of
end-effector (Khatib 1987).
Useful for Cartesian space motion and force control.
Equations of motion given as
+ GX (q)
F = [MX (q)]X + CX (q, q)

(22)

F is a 6 1 entity of forces and moments acting on the end-effector,

X is 6 1 entity representing the position and orientation of the
end-effector.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

21 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE (C ONTD .)

and GX (q) analogous to mass matrix,
[MX (q)], CX (q, q),
Coriolis/centripetal and gravity term, respectively.
Relationships between Cartesian and joint space terms

= [J(q)]T F
[MX (q)] = [J(q)]T [M(q)][J(q)]1
q)
= [J(q)]T (C(q, q)
[M(q)][J(q)]1 [J(q)]

CX (q, q)
T
GX (q) = [J(q)] G(q)
(23)
[J(q)]T denotes the inverse of [J(q)]T ,
[J(q)], F and X are in the same coordinate system.

q, q can be (at least conceptually) replaced with the Cartesian

variables inverse kinematics and inverse Jacobian.
Difficult (if not impossible) for practical manipulators!
Fortunately replacement not required for Cartesian space motion and
force control!!
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

22 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE (C ONTD .)

and GX (q) analogous to mass matrix,
[MX (q)], CX (q, q),
Coriolis/centripetal and gravity term, respectively.
Relationships between Cartesian and joint space terms

= [J(q)]T F
[MX (q)] = [J(q)]T [M(q)][J(q)]1
q)
= [J(q)]T (C(q, q)
[M(q)][J(q)]1 [J(q)]

CX (q, q)
T
GX (q) = [J(q)] G(q)
(23)
[J(q)]T denotes the inverse of [J(q)]T ,
[J(q)], F and X are in the same coordinate system.

q, q can be (at least conceptually) replaced with the Cartesian

variables inverse kinematics and inverse Jacobian.
Difficult (if not impossible) for practical manipulators!
Fortunately replacement not required for Cartesian space motion and
force control!!
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

22 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE (C ONTD .)

and GX (q) analogous to mass matrix,
[MX (q)], CX (q, q),
Coriolis/centripetal and gravity term, respectively.
Relationships between Cartesian and joint space terms

= [J(q)]T F
[MX (q)] = [J(q)]T [M(q)][J(q)]1
q)
= [J(q)]T (C(q, q)
[M(q)][J(q)]1 [J(q)]

CX (q, q)
T
GX (q) = [J(q)] G(q)
(23)
[J(q)]T denotes the inverse of [J(q)]T ,
[J(q)], F and X are in the same coordinate system.

q, q can be (at least conceptually) replaced with the Cartesian

variables inverse kinematics and inverse Jacobian.
Difficult (if not impossible) for practical manipulators!
Fortunately replacement not required for Cartesian space motion and
force control!!
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

22 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE (C ONTD .)

and GX (q) analogous to mass matrix,
[MX (q)], CX (q, q),
Coriolis/centripetal and gravity term, respectively.
Relationships between Cartesian and joint space terms

= [J(q)]T F
[MX (q)] = [J(q)]T [M(q)][J(q)]1
q)
= [J(q)]T (C(q, q)
[M(q)][J(q)]1 [J(q)]

CX (q, q)
T
GX (q) = [J(q)] G(q)
(23)
[J(q)]T denotes the inverse of [J(q)]T ,
[J(q)], F and X are in the same coordinate system.

q, q can be (at least conceptually) replaced with the Cartesian

variables inverse kinematics and inverse Jacobian.
Difficult (if not impossible) for practical manipulators!
Fortunately replacement not required for Cartesian space motion and
force control!!
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

22 / 74

E QUATIONS OF M OTION IN C ARTESIAN S PACE (C ONTD .)

and GX (q) analogous to mass matrix,
[MX (q)], CX (q, q),
Coriolis/centripetal and gravity term, respectively.
Relationships between Cartesian and joint space terms

= [J(q)]T F
[MX (q)] = [J(q)]T [M(q)][J(q)]1
q)
= [J(q)]T (C(q, q)
[M(q)][J(q)]1 [J(q)]

CX (q, q)
T
GX (q) = [J(q)] G(q)
(23)
[J(q)]T denotes the inverse of [J(q)]T ,
[J(q)], F and X are in the same coordinate system.

q, q can be (at least conceptually) replaced with the Cartesian

variables inverse kinematics and inverse Jacobian.
Difficult (if not impossible) for practical manipulators!
Fortunately replacement not required for Cartesian space motion and
force control!!
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

22 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

23 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR

Simplest possible serial manipulator

Two moving links, 2 joint variables 1 and 2
Joint torques 1 and 2 .
0)
Gravity along ( 0 Y
axis.

Lo cation of cg of Link 2

g
Y

(mi , li , ri , Ii ), i = 1, 2
denote mass, length, CG
location and Izz
component of inertia
matrix, respectively.

Link 2
(m2 , l 2 , r2 , I 2 )

Lo cation of cg of Link 1
O2

{ 0}

2
Link 1
(m1 , l 1 , r 1 , I 1 )

2
1

Planar case only Izz

relevant.

X 0

O1

Figure 2: A 2R manipulator
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

24 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Velocity propagation to find linear and angular velocities

{0} fixed 0 0 =0 V0 = 0.
i=1

1 = (0 0 1 )T
1
V1 = 0
1
VC1 = 0 + (0 0 1 )T (r1 0 0)T = (0 r1 1 0)T
1

i=2

2 = (0 0 1 + 2 )T

0
c2 s 2 0
l1 s2 1
2
V2 = s2 c2 0 l1 1 = l1 c2 1
0
0 1
0
0

VC2

V2 +2 2 (r2 0 0)T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

25 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Velocity propagation to find linear and angular velocities

{0} fixed 0 0 =0 V0 = 0.
i=1

1 = (0 0 1 )T
1
V1 = 0
1
VC1 = 0 + (0 0 1 )T (r1 0 0)T = (0 r1 1 0)T
1

i=2

2 = (0 0 1 + 2 )T

0
c2 s 2 0
l1 s2 1
2
V2 = s2 c2 0 l1 1 = l1 c2 1
0
0 1
0
0

VC2

V2 +2 2 (r2 0 0)T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

25 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Velocity propagation to find linear and angular velocities

{0} fixed 0 0 =0 V0 = 0.
i=1

1 = (0 0 1 )T
1
V1 = 0
1
VC1 = 0 + (0 0 1 )T (r1 0 0)T = (0 r1 1 0)T
1

i=2

2 = (0 0 1 + 2 )T

0
c2 s 2 0
l1 s2 1
2
V2 = s2 c2 0 l1 1 = l1 c2 1
0
0 1
0
0

VC2

V2 +2 2 (r2 0 0)T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

25 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Velocity propagation to find linear and angular velocities

{0} fixed 0 0 =0 V0 = 0.
i=1

1 = (0 0 1 )T
1
V1 = 0
1
VC1 = 0 + (0 0 1 )T (r1 0 0)T = (0 r1 1 0)T
1

i=2

2 = (0 0 1 + 2 )T

0
c2 s 2 0
l1 s2 1
2
V2 = s2 c2 0 l1 1 = l1 c2 1
0
0 1
0
0

VC2

V2 +2 2 (r2 0 0)T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

25 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Total kinetic energy

KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 c2 1 (1 + 2 )) (24)
2

Link 1 first two terms; Link 2 second two terms.

Total potential energy
PE = m1 gr1 s1 + m2 g (l1 s1 + r2 s12 )

(25)

Lagrangian for planar 2R manipulator

= KE PE ,
L (, )

= (1 , 2 )T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(26)
.

NPTEL, 2010

26 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Total kinetic energy

KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 c2 1 (1 + 2 )) (24)
2

Link 1 first two terms; Link 2 second two terms.

Total potential energy
PE = m1 gr1 s1 + m2 g (l1 s1 + r2 s12 )

(25)

Lagrangian for planar 2R manipulator

= KE PE ,
L (, )

= (1 , 2 )T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(26)
.

NPTEL, 2010

26 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Total kinetic energy

KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 c2 1 (1 + 2 )) (24)
2

Link 1 first two terms; Link 2 second two terms.

Total potential energy
PE = m1 gr1 s1 + m2 g (l1 s1 + r2 s12 )

(25)

Lagrangian for planar 2R manipulator

= KE PE ,
L (, )

= (1 , 2 )T
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(26)
.

NPTEL, 2010

26 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Partial derivatives of L with respect to i , i = 1, 2

L
1
L
2

= m1 gr1 c1 m2 g (l1 c1 + r2 c12 )

= m2 l1 r2 s2 1 (1 + 2 ) m2 gr2 c12

Partial derivatives of L with respect to i , i = 1, 2

L
1

= (I1 + I2 + m1 r12 + m2 l12 + m2 r22 + 2m2 l1 r2 c2 )1

+(I2 + m2 r22 + m2 l1 r2 c2 )2

L
2

= (I2 + m2 r22 + m2 l1 r2 c2 )1 + (I2 + m2 r22 )2

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

27 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Partial derivatives of L with respect to i , i = 1, 2

L
1
L
2

= m1 gr1 c1 m2 g (l1 c1 + r2 c12 )

= m2 l1 r2 s2 1 (1 + 2 ) m2 gr2 c12

Partial derivatives of L with respect to i , i = 1, 2

L
1

= (I1 + I2 + m1 r12 + m2 l12 + m2 r22 + 2m2 l1 r2 c2 )1

+(I2 + m2 r22 + m2 l1 r2 c2 )2

L
2

= (I2 + m2 r22 + m2 l1 r2 c2 )1 + (I2 + m2 r22 )2

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

27 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Derivatives of

L
i

with respect to t

d L
(
) = 1 (I1 + I2 + m1 r12 + m2 r22 + m2 l12 + 2m2 l1 r2 c2 )
dt 1
+2 (I2 + m2 r22 + m2 l1 r2 c2 ) m2 l1 r2 s2 2 (21 + 2 )
d L
(
) = 1 (I2 + m2 r22 + m2 l1 r2 c2 )
dt 2
+2 (I2 + m2 r22 ) m2 l1 r2 s2 1 2
Assemble terms, collect and simplify
1

1 (I1 + I2 + m2 l12 + m1 r12 + m2 r22 + 2m2 l1 r2 c2 )

+2 (I2 + m2 r22 + m2 l1 r2 c2 )
m2 l1 r2 s2 (21 + 2 )2 + m2 g (l1 c1 + r2 c12 ) + m1 gr1 c1
= 1 (I2 + m2 r22 + m2 l1 r2 c2 ) + 2 (I2 + m2 r22 ) + m2 l1 r2 s2 12 + m2 r2 gc12
=

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

28 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR (C ONTD .)

Derivatives of

L
i

with respect to t

d L
(
) = 1 (I1 + I2 + m1 r12 + m2 r22 + m2 l12 + 2m2 l1 r2 c2 )
dt 1
+2 (I2 + m2 r22 + m2 l1 r2 c2 ) m2 l1 r2 s2 2 (21 + 2 )
d L
(
) = 1 (I2 + m2 r22 + m2 l1 r2 c2 )
dt 2
+2 (I2 + m2 r22 ) m2 l1 r2 s2 1 2
Assemble terms, collect and simplify
1

1 (I1 + I2 + m2 l12 + m1 r12 + m2 r22 + 2m2 l1 r2 c2 )

+2 (I2 + m2 r22 + m2 l1 r2 c2 )
m2 l1 r2 s2 (21 + 2 )2 + m2 g (l1 c1 + r2 c12 ) + m1 gr1 c1
= 1 (I2 + m2 r22 + m2 l1 r2 c2 ) + 2 (I2 + m2 r22 ) + m2 l1 r2 s2 12 + m2 r2 gc12
=

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

28 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR E QUATIONS OF M OTION

Equations of motion two nonlinear ODEs in standard form

(
[

1
2

)
=

](
I1 + I2 + m2 l12 + m1 r12 + m2 r22 + 2m2 l1 r2 c2 I2 + m2 r22 + m2 l1 r2 c2
I2 + m2 r22 + m2 l1 r2 c2
I2 + m2 r22
(
) (
)
m2 l1 r2 s2 (21 + 2 )2
m2 g (l1 c1 + r2 c12 ) + m1 gr1 c1
+
+
m2 r2 gc12
m2 l1 r2 s2 12

1
2

(27)

In the above matrix equation

2 2 matrix is the mass matrix,
2 1 vector with 1 2 is the centripetal/Coriolis term, and
2 1 vector with g is the gravity term.
As mentioned no friction or dissipative terms in equations of motion.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

29 / 74

E XAMPLES
P LANAR 2R M ANIPULATOR E QUATIONS OF M OTION

Equations of motion two nonlinear ODEs in standard form

(
[

1
2

)
=

](
I1 + I2 + m2 l12 + m1 r12 + m2 r22 + 2m2 l1 r2 c2 I2 + m2 r22 + m2 l1 r2 c2
I2 + m2 r22 + m2 l1 r2 c2
I2 + m2 r22
(
) (
)
m2 l1 r2 s2 (21 + 2 )2
m2 g (l1 c1 + r2 c12 ) + m1 gr1 c1
+
+
m2 r2 gc12
m2 l1 r2 s2 12

1
2

(27)

In the above matrix equation

2 2 matrix is the mass matrix,
2 1 vector with 1 2 is the centripetal/Coriolis term, and
2 1 vector with g is the gravity term.
As mentioned no friction or dissipative terms in equations of motion.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

29 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM

Simplest possible one degree-of-freedom closed-loop mechanism!

Three moving links 1 actuated, i , i = 1, 2, 3 passive, 1 actuating
torque.
3
Link 2
(m2 , l2 , r2 , I2 )

O2

Geometry and inertial

parameters of links
(mi , li , ri , Ii ) for i = 1, 2, 3.

O3

{L}

L
Y

Link 3
(m3 , l3 , r3 , I3 )
R
Y

Link 1
(m1 , l1 , r1 , I1 )

{R}

Location of cg of links
1

L
X

Only Izz component of the

inertia matrix of each link is
relevant.

l0
R
X

OL , O1

OR

Figure 3: A planar four-bar mechanism

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

30 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION

Break four-bar mechanism at O3 planar 2R + planar 1R

KE of planar 2R similar 2 replaced by 2
KE of 1R 12 m3 r32 12 + 12 I3 12
Total kinetic energy
KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
2 2
2
m2 (l1 1 + r2 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 )) +
2
1
1
m3 r32 12 + I3 12
(28)
2
2

Total potential energy planar 2R + planar 1R

PE = m1 gr1 sin(1 ) + m2 g (l1 sin(1 ) + r2 sin(1 + 2 )) + m3 gr3 sin(1 )
(29)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

31 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION

Break four-bar mechanism at O3 planar 2R + planar 1R

KE of planar 2R similar 2 replaced by 2
KE of 1R 12 m3 r32 12 + 12 I3 12
Total kinetic energy
KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
2 2
2
m2 (l1 1 + r2 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 )) +
2
1
1
m3 r32 12 + I3 12
(28)
2
2

Total potential energy planar 2R + planar 1R

PE = m1 gr1 sin(1 ) + m2 g (l1 sin(1 ) + r2 sin(1 + 2 )) + m3 gr3 sin(1 )
(29)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

31 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION

Break four-bar mechanism at O3 planar 2R + planar 1R

KE of planar 2R similar 2 replaced by 2
KE of 1R 12 m3 r32 12 + 12 I3 12
Total kinetic energy
KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 )) +
2
1
1
m3 r32 12 + I3 12
(28)
2
2

Total potential energy planar 2R + planar 1R

PE = m1 gr1 sin(1 ) + m2 g (l1 sin(1 ) + r2 sin(1 + 2 )) + m3 gr3 sin(1 )
(29)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

31 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION

Break four-bar mechanism at O3 planar 2R + planar 1R

KE of planar 2R similar 2 replaced by 2
KE of 1R 12 m3 r32 12 + 12 I3 12
Total kinetic energy
KE

1
1
1
m1 (r1 1 )2 + I1 12 + I2 (1 + 2 )2 +
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 )) +
2
1
1
m3 r32 12 + I3 12
(28)
2
2

Total potential energy planar 2R + planar 1R

PE = m1 gr1 sin(1 ) + m2 g (l1 sin(1 ) + r2 sin(1 + 2 )) + m3 gr3 sin(1 )
(29)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

31 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

Lagrangian for the planar 2R + planar 1R mechanisms

=
L (q, q)
1
1 2 1
1
1
I1 1 + I2 (1 + 2 )2 + I3 12 + m1 r12 12 + m3 r32 12
2
2
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 ))
2
m1 gr1 sin(1 ) m2 g (l1 sin(1 + r2 sin(1 + 2 ))
m3 gr3 sin(1 )

(30)

Constraint equations of the planar four-bar (see Module 4, Lecture 1)

l1 cos(1 ) + l2 cos(1 + 2 ) = l0 + l3 cos(1 )
l1 sin(1 ) + l2 sin(1 + 2 ) = l3 sin(1 )

(31)

Perform partial derivatives with respect to q and q and time

derivatives (see Lagrangian formulation)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

32 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

Lagrangian for the planar 2R + planar 1R mechanisms

=
L (q, q)
1 2 1
1
1
1
I1 1 + I2 (1 + 2 )2 + I3 12 + m1 r12 12 + m3 r32 12
2
2
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 ))
2
m1 gr1 sin(1 ) m2 g (l1 sin(1 + r2 sin(1 + 2 ))
m3 gr3 sin(1 )

(30)

Constraint equations of the planar four-bar (see Module 4, Lecture 1)

l1 cos(1 ) + l2 cos(1 + 2 ) = l0 + l3 cos(1 )
l1 sin(1 ) + l2 sin(1 + 2 ) = l3 sin(1 )

(31)

Perform partial derivatives with respect to q and q and time

derivatives (see Lagrangian formulation)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

32 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

Lagrangian for the planar 2R + planar 1R mechanisms

=
L (q, q)
1 2 1
1
1
1
I1 1 + I2 (1 + 2 )2 + I3 12 + m1 r12 12 + m3 r32 12
2
2
2
2
2
1
m2 (l12 12 + r22 (1 + 2 )2 + 2l1 r2 cos(2 )1 (1 + 2 ))
2
m1 gr1 sin(1 ) m2 g (l1 sin(1 + r2 sin(1 + 2 ))
m3 gr3 sin(1 )

(30)

Constraint equations of the planar four-bar (see Module 4, Lecture 1)

l1 cos(1 ) + l2 cos(1 + 2 ) = l0 + l3 cos(1 )
l1 sin(1 ) + l2 sin(1 + 2 ) = l3 sin(1 )

(31)

Perform partial derivatives with respect to q and q and time

derivatives (see Lagrangian formulation)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

32 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

3 3 mass matrix [M(q)]

I2 + m2 r2 2 + I1 + m2 l1 2 + 2 m2 l1 r2 cos(2 ) + m1 r1 2 , I2 + m2 r2 2 + m2 l1 r2 cos(2 ) , 0

I2 + m2 r2 2 + m2 l1 r2 cos(2 ) , I2 + m2 r2 2 , 0
0 , 0 , m3 r3 2 + I3

3 3 Coriolis/Centripetal terms [C(q, q)]

m2 l1 r2 sin(2 ) 2 m2 l1 r2 sin(2 ) 1 m2 l1 r2 sin(2 ) 2

m2 l1 r2 sin(2 ) 1
0
0
0

0
0
0

3 1 vector of gravity terms G(q)

m1 g r1 cos(1 ) + m2 g (l1 cos(1 ) + r2 cos(1 + 2 ))

m2 g r2 cos(1 + 2 )
m3 g r3 cos(1 )
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

33 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

Equations of motion of the planar 2R + 1R mechanisms

1 = (m2 r2 cos(1 + 2 ) + m1 r1 cos(1 ) + m2 l1 cos(1 ))g

+(I2 + m2 r2 2 + I1 + m2 l1 2 + 2 m2 l1 r2 cos(2 ) + m1 r1 2 ) 1
+(I2 + m2 r2 2 + m2 l1 r2 cos(2 )) 2 m2 l1 r2 sin(2 ) (2 )2
2 m2 l1 r2 sin(2 ) 1 2
2 = m2 g r2 cos(1 + 2 ) + (I2 + m2 r2 2 + m2 l1 r2 cos(2 )) 1
+(I2 + m2 r2 2 ) 2 + m2 l1 r2 sin(2 ) 12
3 = m3 g r3 cos(1 ) + (m3 r3 2 + I3 ) 1

(32)

Three non-linear ordinary differential equations.

Constraint equations not yet taken in to account Third equation
not related to the other two!
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

34 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

Equations of motion of the planar 2R + 1R mechanisms

1 = (m2 r2 cos(1 + 2 ) + m1 r1 cos(1 ) + m2 l1 cos(1 ))g

+(I2 + m2 r2 2 + I1 + m2 l1 2 + 2 m2 l1 r2 cos(2 ) + m1 r1 2 ) 1
+(I2 + m2 r2 2 + m2 l1 r2 cos(2 )) 2 m2 l1 r2 sin(2 ) (2 )2
2 m2 l1 r2 sin(2 ) 1 2
2 = m2 g r2 cos(1 + 2 ) + (I2 + m2 r2 2 + m2 l1 r2 cos(2 )) 1
+(I2 + m2 r2 2 ) 2 + m2 l1 r2 sin(2 ) 12
3 = m3 g r3 cos(1 ) + (m3 r3 2 + I3 ) 1

(32)

Three non-linear ordinary differential equations.

Constraint equations not yet taken in to account Third equation
not related to the other two!
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

34 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

2 3 constraint matrix [(q)]

]
[
l1 sin(1 ) l2 sin(1 + 2 ) l2 sin(1 + 2 ) l3 sin(1 )
l1 cos(1 ) + l2 cos(1 + 2 ) l2 cos(1 + 2 ) l3 cos(1 )
Obtain derivative of constraint equations

[(q)]
q + [(q)]
q = 0
from equation of motion
Obtain q
= [M]1 ( [C]q G) + [M]1 []T
q
in derivative of constraint equation and solve for
Substitute q
Substitute to obtain equations of motion of planar four-bar
mechanism
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
where f = ( [C]q G) and q = (1 , 2 , 1 )T .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(33)
.

NPTEL, 2010

35 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

2 3 constraint matrix [(q)]

[
]
l1 sin(1 ) l2 sin(1 + 2 ) l2 sin(1 + 2 ) l3 sin(1 )
l1 cos(1 ) + l2 cos(1 + 2 ) l2 cos(1 + 2 ) l3 cos(1 )
Obtain derivative of constraint equations

[(q)]
q + [(q)]
q = 0
from equation of motion
Obtain q
= [M]1 ( [C]q G) + [M]1 []T
q
in derivative of constraint equation and solve for
Substitute q
Substitute to obtain equations of motion of planar four-bar
mechanism
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
where f = ( [C]q G) and q = (1 , 2 , 1 )T .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(33)
.

NPTEL, 2010

35 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

2 3 constraint matrix [(q)]

[
]
l1 sin(1 ) l2 sin(1 + 2 ) l2 sin(1 + 2 ) l3 sin(1 )
l1 cos(1 ) + l2 cos(1 + 2 ) l2 cos(1 + 2 ) l3 cos(1 )
Obtain derivative of constraint equations

[(q)]
q + [(q)]
q = 0
from equation of motion
Obtain q
= [M]1 ( [C]q G) + [M]1 []T
q
in derivative of constraint equation and solve for
Substitute q
Substitute to obtain equations of motion of planar four-bar
mechanism
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
where f = ( [C]q G) and q = (1 , 2 , 1 )T .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(33)
.

NPTEL, 2010

35 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

2 3 constraint matrix [(q)]

[
]
l1 sin(1 ) l2 sin(1 + 2 ) l2 sin(1 + 2 ) l3 sin(1 )
l1 cos(1 ) + l2 cos(1 + 2 ) l2 cos(1 + 2 ) l3 cos(1 )
Obtain derivative of constraint equations

[(q)]
q + [(q)]
q = 0
from equation of motion
Obtain q
= [M]1 ( [C]q G) + [M]1 []T
q
in derivative of constraint equation and solve for
Substitute q
Substitute to obtain equations of motion of planar four-bar
mechanism
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
where f = ( [C]q G) and q = (1 , 2 , 1 )T .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(33)
.

NPTEL, 2010

35 / 74

E XAMPLES
P LANAR FOUR - BAR M ECHANISM E QUATIONS OF M OTION (C ONTD .)

2 3 constraint matrix [(q)]

[
]
l1 sin(1 ) l2 sin(1 + 2 ) l2 sin(1 + 2 ) l3 sin(1 )
l1 cos(1 ) + l2 cos(1 + 2 ) l2 cos(1 + 2 ) l3 cos(1 )
Obtain derivative of constraint equations

[(q)]
q + [(q)]
q = 0
from equation of motion
Obtain q
= [M]1 ( [C]q G) + [M]1 []T
q
in derivative of constraint equation and solve for
Substitute q
Substitute to obtain equations of motion of planar four-bar
mechanism
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
where f = ( [C]q G) and q = (1 , 2 , 1 )T .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(33)
.

NPTEL, 2010

35 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

36 / 74

I NTRODUCTION

Two problems in dynamics of robots

Inverse dynamics given D-H and inertial parameters and a trajectory
as a function of time, find joint torques Obtain (t) from known

(t).
q(t), q(t)
and q
Direct problem given the kinematic and inertial parameters and the
joint torques as functions of time, find the trajectory of the
manipulator Obtain q(t) from known (t).

Inverse dynamics is required for sizing of actuators and model based

control.
Direct problem solution is required for simulation.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

37 / 74

I NTRODUCTION

Two problems in dynamics of robots

Inverse dynamics given D-H and inertial parameters and a trajectory
as a function of time, find joint torques Obtain (t) from known

(t).
q(t), q(t)
and q
Direct problem given the kinematic and inertial parameters and the
joint torques as functions of time, find the trajectory of the
manipulator Obtain q(t) from known (t).

Inverse dynamics is required for sizing of actuators and model based

control.
Direct problem solution is required for simulation.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

37 / 74

I NTRODUCTION

Two problems in dynamics of robots

Inverse dynamics given D-H and inertial parameters and a trajectory
as a function of time, find joint torques Obtain (t) from known

(t).
q(t), q(t)
and q
Direct problem given the kinematic and inertial parameters and the
joint torques as functions of time, find the trajectory of the
manipulator Obtain q(t) from known (t).

Inverse dynamics is required for sizing of actuators and model based

control.
Direct problem solution is required for simulation.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

37 / 74

I NVERSE DYNAMICS OF ROBOTS

Inverse dynamics problem is very simple!

(t) in the right-hand side of equations of

Substitute qt, q(t)
and q
motion
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Obtain the left-hand side (t).
Can be done for any robot once the equations of motion are known.
Can be efficiently done in O(N) steps using recursive algorithms.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

38 / 74

I NVERSE DYNAMICS OF ROBOTS

Inverse dynamics problem is very simple!

(t) in the right-hand side of equations of

Substitute qt, q(t)
and q
motion
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Obtain the left-hand side (t).
Can be done for any robot once the equations of motion are known.
Can be efficiently done in O(N) steps using recursive algorithms.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

38 / 74

I NVERSE DYNAMICS OF ROBOTS

Inverse dynamics problem is very simple!

(t) in the right-hand side of equations of

Substitute qt, q(t)
and q
motion
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Obtain the left-hand side (t).
Can be done for any robot once the equations of motion are known.
Can be efficiently done in O(N) steps using recursive algorithms.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

38 / 74

I NVERSE DYNAMICS OF ROBOTS

Inverse dynamics problem is very simple!

(t) in the right-hand side of equations of

Substitute qt, q(t)
and q
motion
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Obtain the left-hand side (t).
Can be done for any robot once the equations of motion are known.
Can be efficiently done in O(N) steps using recursive algorithms.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

38 / 74

I NVERSE DYNAMICS OF ROBOTS

Inverse dynamics problem is very simple!

(t) in the right-hand side of equations of

Substitute qt, q(t)
and q
motion
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Obtain the left-hand side (t).
Can be done for any robot once the equations of motion are known.
Can be efficiently done in O(N) steps using recursive algorithms.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

38 / 74

I NVERSE DYNAMICS OF ROBOTS

P LANAR 2R E XAMPLE

Lo cation of cg of Link 2

g
Y

Link 2
(m2 , l 2 , r2 , I 2 )

Lo cation of cg of Link 1
O2

{ 0}

2
Link 1
(m1 , l 1 , r 1 , I 1 )

Link
1

X 0

O1

Figure 4: A 2R manipulator

1
2

Length
(m)
1.0
1.0

Mass
(kg)
12.456
12.456
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

C.G.
(m)
0.773
0.583
.

Inertia
(kg m2 )
1.042
1.042
.

NPTEL, 2010

39 / 74

I NVERSE DYNAMICS OF ROBOTS

P LANAR 2R E XAMPLE (C ONTD .)
1.4

Circle traced by end-point

1.2

Mass, inertia and

geometry as in Table

Chosen circular trajectory

l 2 =1
0.8

0.6

= a + r cos( )

= b + r sin( )

0.4

l 1 =1

r = 0.2, a = 1.2, b = 1.2

0.2

0 2 in 10 seconds
0
0

0.2

0.4

0.6

0.8

1.2

1.4

Figure 5: A 2R manipulator executing circular trajectory

Planar 2R inverse dynamics trajectory movie

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

40 / 74

I NVERSE DYNAMICS OF ROBOTS

P LANAR 2R E XAMPLE (C ONTD .)
Plot of Joints Angles for a 2R Manipulator
2

1.5

Angle(rad)

0.5

0.5

1.5

Time(Seconds)

10

Figure 6: Computed 1 and 2 using inverse kinematics

Plot of Joint Angular Velocities for a 2R Manipulator
0.3

0.2

Angular Vel.(rad/sec 2)

0.1

0.1

0.2

2
0.3

0.4

10

Time(Seconds)

Figure 7: Computed 1 and 2 using inverse Jacobian

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

41 / 74

I NVERSE DYNAMICS OF ROBOTS

P LANAR 2R E XAMPLE (C ONTD .)
Plot of Joint Angular Accelerations for a 2R Manipulator
0.4

0.3

Angular Acc.(rad/sec 2 )

0.2

0.1

0.1

0.2

0.3

0.4

Time(seconds)

10

Figure 8: Computed 1 and 2

Plot of Joint Torques for a 2R Manipulator
180

Torque(Nm)

160

140

120

100

80

60

Time(Seconds)

Figure 9: Computed 1 (t) and 2 (t)

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

10

NPTEL, 2010

42 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Simulation given external torque/forces obtain motion of robot.
General form of equations of motion of a n degree-of-freedom robot
+ G(q) + F(q, q)

= [M(q)]
q + C(q, q)
Simulation given (t) find q(t) by solving equations of motion.
n coupled, non-linear, second-order, ordinary differential equations
(ODEs).
Cannot be solved analytically except for simplest cases.

R
Numerical solution of the ODEs Use of software such as Matlab
and in-built integration routine such as ODE45.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

43 / 74

S IMULATION OF E QUATIONS OF M OTION

Input to ODE45 (or other routines) required to be in state-space form.
Conversion to state-space form
(1) Mass matrix [M(q)] is invertible,
= [M(q)]1 [ C(q, q)
G(q) F(q, q)]

q
(2) Define X 2n (X1 , ..., Xn )T = (q1 , ..., qn )T and
(Xn+1 , ...., X2n )T = (q 1 , ...., q n )T .
(3) Rewrite n second-order ODEs as 2n first-order ODEs
X 1 = Xn+1 , X 2 = Xn+2 , ..., X n = X2n

Xn+1
.
1

. = [M(X)] [ C(X) G(X) F(X)]

X 2n

(34)

State-space form of equations of motion

= g(X, )
X
with initial conditions X(0).
A SHITAVA G HOSAL (IIS C )

(35)
.

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

44 / 74

S IMULATION OF E QUATIONS OF M OTION

Input to ODE45 (or other routines) required to be in state-space form.
Conversion to state-space form
(1) Mass matrix [M(q)] is invertible,
= [M(q)]1 [ C(q, q)
G(q) F(q, q)]

q
(2) Define X 2n (X1 , ..., Xn )T = (q1 , ..., qn )T and
(Xn+1 , ...., X2n )T = (q 1 , ...., q n )T .
(3) Rewrite n second-order ODEs as 2n first-order ODEs
X 1 = Xn+1 , X 2 = Xn+2 , ..., X n = X2n

Xn+1
.
1

. = [M(X)] [ C(X) G(X) F(X)]

X 2n

(34)

State-space form of equations of motion

= g(X, )
X
with initial conditions X(0).
A SHITAVA G HOSAL (IIS C )

(35)
.

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

44 / 74

S IMULATION OF E QUATIONS OF M OTION

Input to ODE45 (or other routines) required to be in state-space form.
Conversion to state-space form
(1) Mass matrix [M(q)] is invertible,
= [M(q)]1 [ C(q, q)
G(q) F(q, q)]

q
(2) Define X 2n (X1 , ..., Xn )T = (q1 , ..., qn )T and
(Xn+1 , ...., X2n )T = (q 1 , ...., q n )T .
(3) Rewrite n second-order ODEs as 2n first-order ODEs
X 1 = Xn+1 , X 2 = Xn+2 , ..., X n = X2n

Xn+1
.
1

. = [M(X)] [ C(X) G(X) F(X)]

X 2n

(34)

State-space form of equations of motion

= g(X, )
X
with initial conditions X(0).
A SHITAVA G HOSAL (IIS C )

(35)
.

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

44 / 74

S IMULATION OF E QUATIONS OF M OTION

For parallel manipulator and closed-loop mechanisms, equations of
motion
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
f denotes ( C G F).
Obtain the 2(n + m) first-order state equations as
X 1 = Xn+m+1 ,

X 2 = Xn+m+2 , ..., X n+m = X2(n+m)

Xn+m+1

.
= [M]1 (f []T ([][M]1 []T )1 {[][M]1 f

X2(n+m)
X 1 , ..., X n+m )T })
+[](
(36)
f denotes ( C G F).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

45 / 74

S IMULATION OF E QUATIONS OF M OTION

For parallel manipulator and closed-loop mechanisms, equations of
motion
q}

[M]
q = f []T ([][M]1 []T )1 {[][M]1 f + []
f denotes ( C G F).
Obtain the 2(n + m) first-order state equations as
X 1 = Xn+m+1 ,

X 2 = Xn+m+2 , ..., X n+m = X2(n+m)

Xn+m+1

= [M]1 (f []T ([][M]1 []T )1 {[][M]1 f

X2(n+m)
X 1 , ..., X n+m )T })
+[](
(36)
f denotes ( C G F).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

45 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

The nature of ODEs in equations (36) are different than ODEs

obtained for serial manipulators.
m loop-closure (holonomic) constraints must be satisfied
differential-algebraic equations or DAEs.
DAEs are inherently stiff5 use stiff-solvers such as ODE15S or
R
ODE23S in Matlab
.
Stiff solvers use implicit schemes and are slower than non-stiff solvers.
For simple problems (such as a 4-bar mechanism), ODE45 is good
enough.
5 A system of ODEs is said to be stiff if the time constants of the individual ODEs
are orders of magnitude different more than 1000:1. For stiff ODEs, the time step in
integration is determined by the smallest time constants and hence a set of stiff ODEs
can take very long to integrate. DAEs can be thought of as infinitely stiff since the
algebraic constraints have zero time constant.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

46 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

The nature of ODEs in equations (36) are different than ODEs

obtained for serial manipulators.
m loop-closure (holonomic) constraints must be satisfied
differential-algebraic equations or DAEs.
DAEs are inherently stiff5 use stiff-solvers such as ODE15S or
R
ODE23S in Matlab
.
Stiff solvers use implicit schemes and are slower than non-stiff solvers.
For simple problems (such as a 4-bar mechanism), ODE45 is good
enough.
5 A system of ODEs is said to be stiff if the time constants of the individual ODEs
are orders of magnitude different more than 1000:1. For stiff ODEs, the time step in
integration is determined by the smallest time constants and hence a set of stiff ODEs
can take very long to integrate. DAEs can be thought of as infinitely stiff since the
algebraic constraints have zero time constant.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

46 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

The nature of ODEs in equations (36) are different than ODEs

obtained for serial manipulators.
m loop-closure (holonomic) constraints must be satisfied
differential-algebraic equations or DAEs.
DAEs are inherently stiff5 use stiff-solvers such as ODE15S or
R
ODE23S in Matlab
.
Stiff solvers use implicit schemes and are slower than non-stiff solvers.
For simple problems (such as a 4-bar mechanism), ODE45 is good
enough.
5 A system of ODEs is said to be stiff if the time constants of the individual ODEs
are orders of magnitude different more than 1000:1. For stiff ODEs, the time step in
integration is determined by the smallest time constants and hence a set of stiff ODEs
can take very long to integrate. DAEs can be thought of as infinitely stiff since the
algebraic constraints have zero time constant.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

46 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

The nature of ODEs in equations (36) are different than ODEs

obtained for serial manipulators.
m loop-closure (holonomic) constraints must be satisfied
differential-algebraic equations or DAEs.
DAEs are inherently stiff5 use stiff-solvers such as ODE15S or
R
ODE23S in Matlab
.
Stiff solvers use implicit schemes and are slower than non-stiff solvers.
For simple problems (such as a 4-bar mechanism), ODE45 is good
enough.
5 A system of ODEs is said to be stiff if the time constants of the individual ODEs
are orders of magnitude different more than 1000:1. For stiff ODEs, the time step in
integration is determined by the smallest time constants and hence a set of stiff ODEs
can take very long to integrate. DAEs can be thought of as infinitely stiff since the
algebraic constraints have zero time constant.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

46 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

The nature of ODEs in equations (36) are different than ODEs

obtained for serial manipulators.
m loop-closure (holonomic) constraints must be satisfied
differential-algebraic equations or DAEs.
DAEs are inherently stiff5 use stiff-solvers such as ODE15S or
R
ODE23S in Matlab
.
Stiff solvers use implicit schemes and are slower than non-stiff solvers.
For simple problems (such as a 4-bar mechanism), ODE45 is good
enough.
5 A system of ODEs is said to be stiff if the time constants of the individual ODEs
are orders of magnitude different more than 1000:1. For stiff ODEs, the time step in
integration is determined by the smallest time constants and hence a set of stiff ODEs
can take very long to integrate. DAEs can be thought of as infinitely stiff since the
algebraic constraints have zero time constant.
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

46 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

Equations of motion involve second derivative of m constraint

equations (q, t) = 0.
Small numerical errors in acceleration (
q) due to integration results in
increasing errors in q and q.
Baumgarte stabilization (Baumgarte 1983) Replace second

derivative constraint equation [(q)]

q + [(q)]
q + (t)
= 0 with
q + (t))

+ 2 (q, t) = 0
([]
q + []
+ 2 ((t) + [(q)]q)

and are constants.

Similar to a spring-mass-damper system6 , for proper choice of and

, limt {q(t), q(t)}

0
Not clear how to choose and !
6 The

equations of motion of an unforced mass-spring-damper system is written as

x + 2 n x + n 2 x = 0. is similar to the natural frequency n and is similar to
the product of natural frequency and damping n .
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

47 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

Equations of motion involve second derivative of m constraint

equations (q, t) = 0.
Small numerical errors in acceleration (
q) due to integration results in
increasing errors in q and q.
Baumgarte stabilization (Baumgarte 1983) Replace second

derivative constraint equation [(q)]

q + [(q)]
q + (t)
= 0 with
q + (t))

+ 2 (q, t) = 0
([]
q + []
+ 2 ((t) + [(q)]q)

and are constants.

Similar to a spring-mass-damper system6 , for proper choice of and

, limt {q(t), q(t)}

0
Not clear how to choose and !
6 The

equations of motion of an unforced mass-spring-damper system is written as

x + 2 n x + n 2 x = 0. is similar to the natural frequency n and is similar to
the product of natural frequency and damping n .
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

47 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

Equations of motion involve second derivative of m constraint

equations (q, t) = 0.
Small numerical errors in acceleration (
q) due to integration results in
increasing errors in q and q.
Baumgarte stabilization (Baumgarte 1983) Replace second

derivative constraint equation [(q)]

q + [(q)]
q + (t)
= 0 with
q + (t))

+ 2 (q, t) = 0
([]
q + []
+ 2 ((t) + [(q)]q)

and are constants.

Similar to a spring-mass-damper system6 , for proper choice of and

, limt {q(t), q(t)}

0
Not clear how to choose and !
6 The

equations of motion of an unforced mass-spring-damper system is written as

x + 2 n x + n 2 x = 0. is similar to the natural frequency n and is similar to
the product of natural frequency and damping n .
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

47 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

Equations of motion involve second derivative of m constraint

equations (q, t) = 0.
Small numerical errors in acceleration (
q) due to integration results in
increasing errors in q and q.
Baumgarte stabilization (Baumgarte 1983) Replace second

derivative constraint equation [(q)]

q + [(q)]
q + (t)
= 0 with
q + (t))

+ 2 (q, t) = 0
([]
q + []
+ 2 ((t) + [(q)]q)

and are constants.

Similar to a spring-mass-damper system6 , for proper choice of and

, limt {q(t), q(t)}

0
Not clear how to choose and !
6 The

equations of motion of an unforced mass-spring-damper system is written as

x + 2 n x + n 2 x = 0. is similar to the natural frequency n and is similar to
the product of natural frequency and damping n .
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

47 / 74

S IMULATION OF E QUATIONS OF M OTION

PARALLEL MANIPULATORS (C ONTD .)

Equations of motion involve second derivative of m constraint

equations (q, t) = 0.
Small numerical errors in acceleration (
q) due to integration results in
increasing errors in q and q.
Baumgarte stabilization (Baumgarte 1983) Replace second

derivative constraint equation [(q)]

q + [(q)]
q + (t)
= 0 with
q + (t))

+ 2 (q, t) = 0
([]
q + []
+ 2 ((t) + [(q)]q)

and are constants.

Similar to a spring-mass-damper system6 , for proper choice of and

, limt {q(t), q(t)}

0
Not clear how to choose and !
6 The

equations of motion of an unforced mass-spring-damper system is written as

x + 2 n x + n 2 x = 0. is similar to the natural frequency n and is similar to
the product of natural frequency and damping n .
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

47 / 74

E XAMPLES
Initial conditions
1 = 90 , 2 = 45
External torques
1 = 2 = 0

S IMULATIONS OF A P LANAR 2R ROBOT

Lo cation of cg of Link 2

g
Y

Link 2
(m2 , l 2 , r2 , I 2 )

Lo cation of cg of Link 1
O2

{ 0}

2
Link 1
(m1 , l 1 , r 1 , I 1 )

Link
1

X 0

O1

Figure 10: A 2R manipulator

1
2

Length
(m)
1.0
1.0

Mass
(kg)
12.456
12.456
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

C.G.
(m)
0.773
0.583
.

Inertia
(kg m2 )
1.042
1.042
.

NPTEL, 2010

48 / 74

E XAMPLES
S IMULATIONS OF A P LANAR 2R ROBOT (C ONTD .)
60
1.65

theta1 in degrees

70
80

1.7

90
100

1.75

120

10

theta2 in degrees

50

Y Coordinate

110

1.8

1.85

1.9

1.95

50

5
6
Time t in seconds

Figure 11: Plot of 1 and 2 Vs time

10

2
0.8

0.6

0.4

0.2

0
X coordinate

0.2

0.4

0.6

0.8

Figure 12: Path traced by the tip

Motion of link 2 causes motion of link 1 Coupled ODEs.

The path traced by the tip is quite complicated.
Planar 2R forward dynamics simulation movie
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

49 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM
0
Y

Almost folded configuration

initial 1 and 1 small.

l2

l1

1 actuated by a torsional spring.

Right-hand side of equations of
motion is modified as

l3
l0

0
X
Figure a

0
Y

= 0 k 1

l2

l3
l1
1

0
X
l0

Figure b

Initial (pre-loaded) torque

0 = 1.96 N-m and the spring
constant is given as
k = 0.1 N m/rad.

Figure 13: A four-bar mechanism in two

configurations
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

50 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM

The mass, length, location of centre of

inertia
Link Length Mass
(m)
(kg)
0
1.241

1
1.241 20.15
2
1.2
8.25
3
1.2
8.25

mass and Izz component of

C.G.
(m)

1.2
0.6
0.6

Inertia
(kgm2 )

9.6
0.06
0.06

As the spring unwinds, link 1 rotates in counter-clockwise direction.

Link lengths chosen such that 1 cannot rotate beyond a certain angle.
Links 2 and 3 lock when they align & 4-bar becomes a structure!
For 1 = 0.01 radians, 1 = 0.0102, 6.2698 radians (4-bar direct
kinematics) choose initial 1 = 0.0102 radians.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

51 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM

The mass, length, location of centre of

inertia
Link Length Mass
(m)
(kg)
0
1.241

1
1.241 20.15
2
1.2
8.25
3
1.2
8.25

mass and Izz component of

C.G.
(m)

1.2
0.6
0.6

Inertia
(kgm2 )

9.6
0.06
0.06

As the spring unwinds, link 1 rotates in counter-clockwise direction.

Link lengths chosen such that 1 cannot rotate beyond a certain angle.
Links 2 and 3 lock when they align & 4-bar becomes a structure!
For 1 = 0.01 radians, 1 = 0.0102, 6.2698 radians (4-bar direct
kinematics) choose initial 1 = 0.0102 radians.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

51 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM

The mass, length, location of centre of

inertia
Link Length Mass
(m)
(kg)
0
1.241

1
1.241 20.15
2
1.2
8.25
3
1.2
8.25

mass and Izz component of

C.G.
(m)

1.2
0.6
0.6

Inertia
(kgm2 )

9.6
0.06
0.06

As the spring unwinds, link 1 rotates in counter-clockwise direction.

Link lengths chosen such that 1 cannot rotate beyond a certain angle.
Links 2 and 3 lock when they align & 4-bar becomes a structure!
For 1 = 0.01 radians, 1 = 0.0102, 6.2698 radians (4-bar direct
kinematics) choose initial 1 = 0.0102 radians.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

51 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM

The mass, length, location of centre of

inertia
Link Length Mass
(m)
(kg)
0
1.241

1
1.241 20.15
2
1.2
8.25
3
1.2
8.25

mass and Izz component of

C.G.
(m)

1.2
0.6
0.6

Inertia
(kgm2 )

9.6
0.06
0.06

As the spring unwinds, link 1 rotates in counter-clockwise direction.

Link lengths chosen such that 1 cannot rotate beyond a certain angle.
Links 2 and 3 lock when they align & 4-bar becomes a structure!
For 1 = 0.01 radians, 1 = 0.0102, 6.2698 radians (4-bar direct
kinematics) choose initial 1 = 0.0102 radians.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

51 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM

The mass, length, location of centre of

inertia
Link Length Mass
(m)
(kg)
0
1.241

1
1.241 20.15
2
1.2
8.25
3
1.2
8.25

mass and Izz component of

C.G.
(m)

1.2
0.6
0.6

Inertia
(kgm2 )

9.6
0.06
0.06

As the spring unwinds, link 1 rotates in counter-clockwise direction.

Link lengths chosen such that 1 cannot rotate beyond a certain angle.
Links 2 and 3 lock when they align & 4-bar becomes a structure!
For 1 = 0.01 radians, 1 = 0.0102, 6.2698 radians (4-bar direct
kinematics) choose initial 1 = 0.0102 radians.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

51 / 74

E XAMPLES
S IMULATIONS OF A 4- BAR M ECHANISM
400

500
lambda1
lambda2

350

400

300
300
250
200
200
100
150
0
100
theta1
phi2
phi1

50

10

12

100

14

200

time in sec

10

12

14

time in sec

Figure 14: Plot of 1 , 2 and 1 Vs time

Figure 15: Plot of 1 and 2 Vs time

At 1 = 150.4 links 2 and 3 align singular configuration.

At t = 12.25 seconds & 1 = 150 simulation is stopped.
As t 12.25 seconds, Lagrange multipliers 1 , 2
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

52 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

53 / 74

I NTRODUCTION
Multi-body system with large
number of links redundant
robots, proteins, automobile etc.
Classical model of protein 20
types of amino acid residues in a
serial chain
50-500 residues assumed to be
rigid bodies
Two DOF between two residues
( , ) 100 to 1000 joint
variables!
Figure 16: Amino acid chain in a protein

Direct and inverse dynamics of large multi-body systems

Efficient O(N) (or better) formulations desired
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

54 / 74

I NTRODUCTION
Multi-body system with large
number of links redundant
robots, proteins, automobile etc.
Classical model of protein 20
types of amino acid residues in a
serial chain
50-500 residues assumed to be
rigid bodies
Two DOF between two residues
( , ) 100 to 1000 joint
variables!
Figure 16: Amino acid chain in a protein

Direct and inverse dynamics of large multi-body systems

Efficient O(N) (or better) formulations desired
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

54 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
I NTRODUCTION

(t) find (t).

Inverse dynamics Given q(t), q(t)
&q
Newton-Euler formulation Newtons Law and Euler Equation for
each link {i}
F = mi 0 V Ci
N = Ci [Ii ]0 i + 0 i Ci [Ii ]0 i

(37)

mi , Ci and [Ii ] are the mass, centre of mass and inertia of link {i}.
Requires computation of position/orientation, velocity and
acceleration.
Position & orientation computed using i1
i [T ] (see Module 2, Lecture
1)
Linear and angular velocities can be computed using propagation
formulas (see Module 5, Lecture 1)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

55 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
I NTRODUCTION

(t) find (t).

Inverse dynamics Given q(t), q(t)
&q
Newton-Euler formulation Newtons Law and Euler Equation for
each link {i}
F = mi 0 V Ci
N = Ci [Ii ]0 i + 0 i Ci [Ii ]0 i

(37)

mi , Ci and [Ii ] are the mass, centre of mass and inertia of link {i}.
Requires computation of position/orientation, velocity and
acceleration.
Position & orientation computed using i1
i [T ] (see Module 2, Lecture
1)
Linear and angular velocities can be computed using propagation
formulas (see Module 5, Lecture 1)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

55 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
I NTRODUCTION

(t) find (t).

Inverse dynamics Given q(t), q(t)
&q
Newton-Euler formulation Newtons Law and Euler Equation for
each link {i}
F = mi 0 V Ci
N = Ci [Ii ]0 i + 0 i Ci [Ii ]0 i

(37)

mi , Ci and [Ii ] are the mass, centre of mass and inertia of link {i}.
Requires computation of position/orientation, velocity and
acceleration.
Position & orientation computed using i1
i [T ] (see Module 2, Lecture
1)
Linear and angular velocities can be computed using propagation
formulas (see Module 5, Lecture 1)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

55 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
I NTRODUCTION

(t) find (t).

Inverse dynamics Given q(t), q(t)
&q
Newton-Euler formulation Newtons Law and Euler Equation for
each link {i}
F = mi 0 V Ci
N = Ci [Ii ]0 i + 0 i Ci [Ii ]0 i

(37)

mi , Ci and [Ii ] are the mass, centre of mass and inertia of link {i}.
Requires computation of position/orientation, velocity and
acceleration.
Position & orientation computed using i1
i [T ] (see Module 2, Lecture
1)
Linear and angular velocities can be computed using propagation
formulas (see Module 5, Lecture 1)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

55 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
I NTRODUCTION

(t) find (t).

Inverse dynamics Given q(t), q(t)
&q
Newton-Euler formulation Newtons Law and Euler Equation for
each link {i}
F = mi 0 V Ci
N = Ci [Ii ]0 i + 0 i Ci [Ii ]0 i

(37)

mi , Ci and [Ii ] are the mass, centre of mass and inertia of link {i}.
Requires computation of position/orientation, velocity and
acceleration.
Position & orientation computed using i1
i [T ] (see Module 2, Lecture
1)
Linear and angular velocities can be computed using propagation
formulas (see Module 5, Lecture 1)
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

55 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION

For rotary (R) joint

i
i
Vi

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi )
i1 [R](

=
=

(38)

For prismatic (P) joint

i

Vi

i
i1
i1
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi ) + d i (0
i1 [R](

0 1)T

(39)

Acceleration of an arbitrary point p on rigid body {i} differentiate

velocity with time
0

V p =

V Oi + 0i [R]i V p + 2 0 i 0i [R]i Vp + 0 i 0i [R]i p

+0 i (0 i 0i [R]i p)
When i p is constant, then i Vp = i V p = 0.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

56 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION

For rotary (R) joint

i
i
Vi

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi )
i1 [R](

=
=

(38)

For prismatic (P) joint

i

Vi

i
i1
i1
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi ) + d i (0
i1 [R](

0 1)T

(39)

Acceleration of an arbitrary point p on rigid body {i} differentiate

velocity with time
0

V p =

V Oi + 0i [R]i V p + 2 0 i 0i [R]i Vp + 0 i 0i [R]i p

+0 i (0 i 0i [R]i p)
When i p is constant, then i Vp = i V p = 0.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

56 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION

For rotary (R) joint

i
i
Vi

i
i1
i1 + i (0 0 1)T
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi )
i1 [R](

=
=

(38)

For prismatic (P) joint

i

Vi

i
i1
i1
i1 [R]
i
i1
Vi1 +i1 i1 i1 Oi ) + d i (0
i1 [R](

0 1)T

(39)

Acceleration of an arbitrary point p on rigid body {i} differentiate

velocity with time
0

V p =

V Oi + 0i [R]i V p + 2 0 i 0i [R]i Vp + 0 i 0i [R]i p

+0 i (0 i 0i [R]i p)
When i p is constant, then i Vp = i V p = 0.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

56 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION (C ONTD .)

When joint i + 1 is rotary (R)

i+1
i+1

V i+1 =

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
(40)
i
i+1
i+1
i
i
i+1
i+1

[R] i + i [R] i i+1

Zi+1 + i+1
Zi+1
i

When joint i + 1 is prismatic (P)

i+1

V i+1 =

i+1

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
i
i+1 + di+1 i+1 Z
i+1
+2i+1 i+1 d i+1 i+1 Z
i+1
[R]i i
i

(41)

The acceleration of the centre of mass of link i is

i

+ i i i pC + i i (i i i pC )
V Ci = i Vi
i
i

(42)

ip
Ci

is the position vector of the centre of mass of link i with respect

to origin Oi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

57 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION (C ONTD .)

When joint i + 1 is rotary (R)

i+1
i+1

V i+1 =

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
(40)
i
i+1
i+1
i
i
i+1
i+1
i+1 + i+1
i+1
[R] i + i [R] i i+1
Z
Z
i

When joint i + 1 is prismatic (P)

i+1

V i+1 =

i+1

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
i
i+1 + di+1 i+1 Z
i+1
+2i+1 i+1 d i+1 i+1 Z
i+1
[R]i i
i

(41)

The acceleration of the centre of mass of link i is

i

+ i i i pC + i i (i i i pC )
V Ci = i Vi
i
i

(42)

ip
Ci

is the position vector of the centre of mass of link i with respect

to origin Oi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

57 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
V ELOCITY PROPAGATION & ACCELERATION (C ONTD .)

When joint i + 1 is rotary (R)

i+1
i+1

V i+1 =

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
(40)
i
i+1
i+1
i
i
i+1
i+1
i+1 + i+1
i+1
[R] i + i [R] i i+1
Z
Z
i

When joint i + 1 is prismatic (P)

i+1

V i+1 =

i+1

i+1 =

i+1
[R][i V i + i i i pi+1 + i i (i i i pi+1 )]
i
i+1 + di+1 i+1 Z
i+1
+2i+1 i+1 d i+1 i+1 Z
i+1
[R]i i
i

(41)

The acceleration of the centre of mass of link i is

i

+ i i i pC + i i (i i i pC )
V Ci = i Vi
i
i

(42)

ip
Ci

is the position vector of the centre of mass of link i with respect

to origin Oi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

57 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION

Use propagation formulas for position/orientation of links.

Outward iterations for velocities and accelerations
i : 0 N1
i+1
i+1 = i+1 [R]i i + i+1 (0 0 1)T

i+1 =
i+1
Vi+1 =
i+1
VC
=
i+1

i +1

i
i+1
[R]i i + i+1
[R]i i i+1 (0 0 1)T + i+1 (0
i
i
i+1
[R](i V i + i i i pi+1 + i i (i i i pi+1 ))
i
i+1
Vi+1 + i+1 i+1 i+1 pCi +1
+i+1 i+1 (i+1 i+1 i+1 pCi +1 )

0 1)T
(43)

Use of Newtons Law and Euler equations for each link i

i+1
Fi+1 = mi+1 i+1 V C

(44)

i +1

i+1

Ni+1 =

Ci +1

[I ]i+1

i+1

i+1 +

i+1

i+1
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

Ci +1

[I ]i+1
.

i+1

i+1
.

NPTEL, 2010

58 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION

Use propagation formulas for position/orientation of links.

Outward iterations for velocities and accelerations
i : 0 N1
i+1
i+1 = i+1 [R]i i + i+1 (0 0 1)T

i+1 =
i+1
Vi+1 =
i+1
VC
=
i+1

i +1

i
i+1
[R]i i + i+1
[R]i i i+1 (0 0 1)T + i+1 (0
i
i
i+1
[R](i V i + i i i pi+1 + i i (i i i pi+1 ))
i
i+1
Vi+1 + i+1 i+1 i+1 pCi +1
+i+1 i+1 (i+1 i+1 i+1 pCi +1 )

0 1)T
(43)

Use of Newtons Law and Euler equations for each link i

i+1
Fi+1 = mi+1 i+1 V C

(44)

i +1

i+1

Ni+1 =

Ci +1

[I ]i+1

i+1

i+1 +

i+1

i+1
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

Ci +1

[I ]i+1
.

i+1

i+1
.

NPTEL, 2010

58 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION

Use propagation formulas for position/orientation of links.

Outward iterations for velocities and accelerations
i : 0 N1
i+1
i+1 = i+1 [R]i i + i+1 (0 0 1)T

i+1 =
i+1
Vi+1 =
i+1
VC
=
i+1

i +1

i
i+1
[R]i i + i+1
[R]i i i+1 (0 0 1)T + i+1 (0
i
i
i+1
[R](i V i + i i i pi+1 + i i (i i i pi+1 ))
i
i+1
Vi+1 + i+1 i+1 i+1 pCi +1
+i+1 i+1 (i+1 i+1 i+1 pCi +1 )

0 1)T
(43)

Use of Newtons Law and Euler equations for each link i

i+1
Fi+1 = mi+1 i+1 V C

(44)

i +1

i+1

Ni+1 =

Ci +1

[I ]i+1

i+1

i+1 +

i+1

i+1
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

Ci +1

[I ]i+1
.

i+1

i+1
.

NPTEL, 2010

58 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)
iF

and i Ni are known from outward iteration

Use of free-body diagram (see Module 5 Lecture 5 on Statics)

Compute joint torques from i Fi and i Ni by inward iteration
i:N 1
i
i

fi

ni

i
i+1
fi+1 +i Fi
i+1 [R]
i
i+1
ni+1 +i pi+1 ii+1 [R]i+1 fi+1 +i
i+1 [R]
i
i
ni i Z

pCi i Fi +i Ni
(45)

To include gravity set 0 V 0 = g the fixed link (or base) is

accelerating upward with 1.0g acceleration.
Algorithm given is for rotary (R) jointed serial manipulator
Substitute equations for Prismatic (P) joint if present.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

59 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)
iF

and i Ni are known from outward iteration

Use of free-body diagram (see Module 5 Lecture 5 on Statics)

Compute joint torques from i Fi and i Ni by inward iteration
i:N 1
i
i

fi

ni

i
i+1
fi+1 +i Fi
i+1 [R]
i
i+1
ni+1 +i pi+1 ii+1 [R]i+1 fi+1 +i
i+1 [R]
i
i
ni i Z

pCi i Fi +i Ni
(45)

To include gravity set 0 V 0 = g the fixed link (or base) is

accelerating upward with 1.0g acceleration.
Algorithm given is for rotary (R) jointed serial manipulator
Substitute equations for Prismatic (P) joint if present.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

59 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)
iF

and i Ni are known from outward iteration

Use of free-body diagram (see Module 5 Lecture 5 on Statics)

Compute joint torques from i Fi and i Ni by inward iteration
i:N 1
i
i

fi

ni

i
i+1
fi+1 +i Fi
i+1 [R]
i
i+1
ni+1 +i pi+1 ii+1 [R]i+1 fi+1 +i
i+1 [R]
i
i
ni i Z

pCi i Fi +i Ni
(45)

To include gravity set 0 V 0 = g the fixed link (or base) is

accelerating upward with 1.0g acceleration.
Algorithm given is for rotary (R) jointed serial manipulator
Substitute equations for Prismatic (P) joint if present.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

59 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)
iF

and i Ni are known from outward iteration

Use of free-body diagram (see Module 5 Lecture 5 on Statics)

Compute joint torques from i Fi and i Ni by inward iteration
i:N 1
i
i

fi

ni

i
i+1
fi+1 +i Fi
i+1 [R]
i
i+1
ni+1 +i pi+1 ii+1 [R]i+1 fi+1 +i
i+1 [R]
i
i
ni i Z

pCi i Fi +i Ni
(45)

To include gravity set 0 V 0 = g the fixed link (or base) is

accelerating upward with 1.0g acceleration.
Algorithm given is for rotary (R) jointed serial manipulator
Substitute equations for Prismatic (P) joint if present.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

59 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)
iF

and i Ni are known from outward iteration

Use of free-body diagram (see Module 5 Lecture 5 on Statics)

Compute joint torques from i Fi and i Ni by inward iteration
i:N 1
i
i

fi

ni

i
i+1
fi+1 +i Fi
i+1 [R]
i
i+1
ni+1 +i pi+1 ii+1 [R]i+1 fi+1 +i
i+1 [R]
i
i
ni i Z

pCi i Fi +i Ni
(45)

To include gravity set 0 V 0 = g the fixed link (or base) is

accelerating upward with 1.0g acceleration.
Algorithm given is for rotary (R) jointed serial manipulator
Substitute equations for Prismatic (P) joint if present.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

59 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)

Newton-Euler algorithm has O(N) computational complexity

The computation in sets of equations (43 -45) is performed onlyonce.
There are no iteration or loops.
Number of multiplications and additions is proportional to N (number
of links).

Very easily adapted for any serial manipulators with rotary (R),
prismatic (P) or any other joint.
if
i

and i ni can be used to obtain all components of reactions at joints

Useful for design.
Can be used in symbolic computation of equations of motion.
Extensively used in robotics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

60 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)

Newton-Euler algorithm has O(N) computational complexity

The computation in sets of equations (43 -45) is performed onlyonce.
There are no iteration or loops.
Number of multiplications and additions is proportional to N (number
of links).

Very easily adapted for any serial manipulators with rotary (R),
prismatic (P) or any other joint.
if
i

and i ni can be used to obtain all components of reactions at joints

Useful for design.
Can be used in symbolic computation of equations of motion.
Extensively used in robotics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

60 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)

Newton-Euler algorithm has O(N) computational complexity

The computation in sets of equations (43 -45) is performed onlyonce.
There are no iteration or loops.
Number of multiplications and additions is proportional to N (number
of links).

Very easily adapted for any serial manipulators with rotary (R),
prismatic (P) or any other joint.
if
i

and i ni can be used to obtain all components of reactions at joints

Useful for design.
Can be used in symbolic computation of equations of motion.
Extensively used in robotics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

60 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)

Newton-Euler algorithm has O(N) computational complexity

The computation in sets of equations (43 -45) is performed onlyonce.
There are no iteration or loops.
Number of multiplications and additions is proportional to N (number
of links).

Very easily adapted for any serial manipulators with rotary (R),
prismatic (P) or any other joint.
if
i

and i ni can be used to obtain all components of reactions at joints

Useful for design.
Can be used in symbolic computation of equations of motion.
Extensively used in robotics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

60 / 74

R ECURSIVE INVERSE DYNAMICS OF SERIAL

MANIPULATORS
N EWTON -E ULER F ORMULATION (C ONTD .)

Newton-Euler algorithm has O(N) computational complexity

The computation in sets of equations (43 -45) is performed onlyonce.
There are no iteration or loops.
Number of multiplications and additions is proportional to N (number
of links).

Very easily adapted for any serial manipulators with rotary (R),
prismatic (P) or any other joint.
if
i

and i ni can be used to obtain all components of reactions at joints

Useful for design.
Can be used in symbolic computation of equations of motion.
Extensively used in robotics.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

60 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

I NTRODUCTION

Forward dynamics: Given (t) obtain q(t).

Involves two steps
(t)
Obtain q
(t) with initial conditions to obtain q(t)

Integrate q
and q(t)

(t) does not include

O(N) algorithms for forward dynamics give q
integration step.
Brute force approach
Obtain equations of motion using Newton-Euler formulation O(N)
steps
q F(q, q)

[M(q)]
q = [C(q, q)]
by inverting [M(q)] O(N 3 ) steps using Gauss Elimination.
Obtain q
Although O(N 3 ), coefficient of N 3 is small (Walker and Orin, 1982)
and hence efficient for serial manipulators with N 6.

O(N 3 ) not very useful when N is large (as in protein chains with N
between 50 and 500).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

61 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

I NTRODUCTION

Forward dynamics: Given (t) obtain q(t).

Involves two steps
(t)
Obtain q
(t) with initial conditions to obtain q(t)

Integrate q
and q(t)

(t) does not include

O(N) algorithms for forward dynamics give q
integration step.
Brute force approach
Obtain equations of motion using Newton-Euler formulation O(N)
steps
q F(q, q)

[M(q)]
q = [C(q, q)]
by inverting [M(q)] O(N 3 ) steps using Gauss Elimination.
Obtain q
Although O(N 3 ), coefficient of N 3 is small (Walker and Orin, 1982)
and hence efficient for serial manipulators with N 6.

O(N 3 ) not very useful when N is large (as in protein chains with N
between 50 and 500).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

61 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

I NTRODUCTION

Forward dynamics: Given (t) obtain q(t).

Involves two steps
(t)
Obtain q
(t) with initial conditions to obtain q(t)

Integrate q
and q(t)

(t) does not include

O(N) algorithms for forward dynamics give q
integration step.
Brute force approach
Obtain equations of motion using Newton-Euler formulation O(N)
steps
q F(q, q)

[M(q)]
q = [C(q, q)]
by inverting [M(q)] O(N 3 ) steps using Gauss Elimination.
Obtain q
Although O(N 3 ), coefficient of N 3 is small (Walker and Orin, 1982)
and hence efficient for serial manipulators with N 6.

O(N 3 ) not very useful when N is large (as in protein chains with N
between 50 and 500).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

61 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

I NTRODUCTION

Forward dynamics: Given (t) obtain q(t).

Involves two steps
(t)
Obtain q
(t) with initial conditions to obtain q(t)

Integrate q
and q(t)

(t) does not include

O(N) algorithms for forward dynamics give q
integration step.
Brute force approach
Obtain equations of motion using Newton-Euler formulation O(N)
steps
q F(q, q)

[M(q)]
q = [C(q, q)]
by inverting [M(q)] O(N 3 ) steps using Gauss Elimination.
Obtain q
Although O(N 3 ), coefficient of N 3 is small (Walker and Orin, 1982)
and hence efficient for serial manipulators with N 6.

O(N 3 ) not very useful when N is large (as in protein chains with N
between 50 and 500).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

61 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

I NTRODUCTION

Forward dynamics: Given (t) obtain q(t).

Involves two steps
(t)
Obtain q
(t) with initial conditions to obtain q(t)

Integrate q
and q(t)

(t) does not include

O(N) algorithms for forward dynamics give q
integration step.
Brute force approach
Obtain equations of motion using Newton-Euler formulation O(N)
steps
q F(q, q)

[M(q)]
q = [C(q, q)]
by inverting [M(q)] O(N 3 ) steps using Gauss Elimination.
Obtain q
Although O(N 3 ), coefficient of N 3 is small (Walker and Orin, 1982)
and hence efficient for serial manipulators with N 6.

O(N 3 ) not very useful when N is large (as in protein chains with N
between 50 and 500).
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

61 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

Simplest possible
example: Body 1 slides
on horizontal rail fixed to
ground and Body 2 slides
on vertical rail fixed to
Body 1.

A RTICULATED - BODY ALGORITHM - KEY IDEA

Rigid-body

1
Motion of prismatic
joint 2

Y2

No rotations of bodies
X , Y coordinates enough
to describe two bodies!

X2

Rigid-body

2
X1

Motion of prismatic joint

Absolute coordinates for

body 1 x1 , y1 .

Absolute coordinates for

body 2 x2 , y2 .
Figure 17: Planar 2P example (Featherstone, 1983)

Constraints y1 = 0 &
x2 x1 = 0
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

62 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Two Lagrange multipliers 1 and 2 for two constraints

Equations of motion and algebraic constraints for body 1
[
](
) ( )
(
) (
)
m1 0
x1
fx1
1
0
+
1 =

2
0 m1
y1
fy1
0
1
[
]
x1
[0 1]
= 0
y1
Equations of motion and algebraic constraints for body 2
[
](
)
( )
(
)
1
m2 0
x2
fx2
+
2 =
0 m2
y2
0
fy2
[
]
[
]
x2
x1
[1 0]
= 0 [1 0]
y2
y1
fxi , fyi external forces for yi , yi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

63 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Two Lagrange multipliers 1 and 2 for two constraints

Equations of motion and algebraic constraints for body 1
[
](
) ( )
(
) (
)
m1 0
x1
fx1
1
0
1 =
2
+

0 m1
y1
fy1
0
1
[
]
x1
[0 1]
= 0
y1
Equations of motion and algebraic constraints for body 2
](
)
( )
[
(
)
m2 0
x2
1
fx2
+
2 =
0
0 m2
y2
fy2
[
]
[
]
x2
x1
[1 0]
= 0 [1 0]
y2
y1
fxi , fyi external forces for yi , yi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

63 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Two Lagrange multipliers 1 and 2 for two constraints

Equations of motion and algebraic constraints for body 1
[
](
) ( )
(
) (
)
m1 0
x1
fx1
1
0
1 =
2
+

0 m1
y1
fy1
0
1
[
]
x1
[0 1]
= 0
y1
Equations of motion and algebraic constraints for body 2
](
)
( )
[
(
)
m2 0
x2
1
fx2
+
2 =
0
0 m2
y2
fy2
[
]
[
]
x2
x1
[1 0]
= 0 [1 0]
y2
y1
fxi , fyi external forces for yi , yi .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

63 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Effect of Body 2 seen in equations of motion on Body 1

Equations of motion for Bodies 1 and 2 are coupled both Lagrange
multipliers 1 and 2 appear in equation of motion for body 1
Not possible to get O(N) algorithm to obtain accelerations xi , yi 1
and 2 at best can be solved by a O(N 3 ) algorithm.
For O(N) recursive forward dynamics algorithm obtain equation of
motion for all connected bodies similar to Body 2 (terminal body)!!
Achieved by Featherstone (1983)(see Additional Material) For the
2P example, equations of motion for Body 1 are
) ( )
[
](
(
)
m1 + m2 0
x1
0
fx1 + fx2
+
1 =
y1
0
m1
1
fy1

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

64 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Effect of Body 2 seen in equations of motion on Body 1

Equations of motion for Bodies 1 and 2 are coupled both Lagrange
multipliers 1 and 2 appear in equation of motion for body 1
Not possible to get O(N) algorithm to obtain accelerations xi , yi 1
and 2 at best can be solved by a O(N 3 ) algorithm.
For O(N) recursive forward dynamics algorithm obtain equation of
motion for all connected bodies similar to Body 2 (terminal body)!!
Achieved by Featherstone (1983)(see Additional Material) For the
2P example, equations of motion for Body 1 are
) ( )
[
](
(
)
m1 + m2 0
x1
0
fx1 + fx2
+
1 =
y1
0
m1
1
fy1

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

64 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Effect of Body 2 seen in equations of motion on Body 1

Equations of motion for Bodies 1 and 2 are coupled both Lagrange
multipliers 1 and 2 appear in equation of motion for body 1
Not possible to get O(N) algorithm to obtain accelerations xi , yi 1
and 2 at best can be solved by a O(N 3 ) algorithm.
For O(N) recursive forward dynamics algorithm obtain equation of
motion for all connected bodies similar to Body 2 (terminal body)!!
Achieved by Featherstone (1983)(see Additional Material) For the
2P example, equations of motion for Body 1 are
) ( )
[
](
(
)
m1 + m2 0
x1
0
fx1 + fx2
+
1 =
y1
0
m1
1
fy1

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

64 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Effect of Body 2 seen in equations of motion on Body 1

Equations of motion for Bodies 1 and 2 are coupled both Lagrange
multipliers 1 and 2 appear in equation of motion for body 1
Not possible to get O(N) algorithm to obtain accelerations xi , yi 1
and 2 at best can be solved by a O(N 3 ) algorithm.
For O(N) recursive forward dynamics algorithm obtain equation of
motion for all connected bodies similar to Body 2 (terminal body)!!
Achieved by Featherstone (1983)(see Additional Material) For the
2P example, equations of motion for Body 1 are
) ( )
[
](
(
)
m1 + m2 0
x1
0
fx1 + fx2
+
1 =
y1
0
m1
1
fy1

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

64 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM - KEY IDEA (C ONTD .)

Effect of Body 2 seen in equations of motion on Body 1

Equations of motion for Bodies 1 and 2 are coupled both Lagrange
multipliers 1 and 2 appear in equation of motion for body 1
Not possible to get O(N) algorithm to obtain accelerations xi , yi 1
and 2 at best can be solved by a O(N 3 ) algorithm.
For O(N) recursive forward dynamics algorithm obtain equation of
motion for all connected bodies similar to Body 2 (terminal body)!!
Achieved by Featherstone (1983)(see Additional Material) For the
2P example, equations of motion for Body 1 are
) ( )
[
](
(
)
m1 + m2 0
x1
0
fx1 + fx2
+
1 =
y1
0
m1
1
fy1

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

64 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM G ENERAL APPROACH

Equations of motion of a single rigid-body under the action of force F

and moment NC acting at the centre of mass
F = mV C
NC

[I ]

where m, C [I ] are the mass and inertia, respectively.

No C [I ] term since all quantities are with respect to coordinate
frame {C } at centre of mass.
Rewrite above equations as

[
]
F
m [U] [0]

FC =
=
AC
C [I ]
[0]
NC
[U]: 3 3 identity matrix,
FC : 6 1 entity of external force and moment (see Module 5, Lecture
5).
.
.
.
.
.
.
AC : 6 1 entity of linear and angular acceleration.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

65 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM G ENERAL APPROACH

Equations of motion of a single rigid-body under the action of force F

and moment NC acting at the centre of mass
F = mV C
NC

[I ]

where m, C [I ] are the mass and inertia, respectively.

No C [I ] term since all quantities are with respect to coordinate
frame {C } at centre of mass.
Rewrite above equations as

[
]
F
m [U] [0]

FC =
=
AC
C [I ]
[0]
NC
[U]: 3 3 identity matrix,
FC : 6 1 entity of external force and moment (see Module 5, Lecture
5).
.
.
.
.
.
.
AC : 6 1 entity of linear and angular acceleration.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

65 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM G ENERAL APPROACH

Equations of motion of a single rigid-body under the action of force F

and moment NC acting at the centre of mass
F = mV C
NC

[I ]

where m, C [I ] are the mass and inertia, respectively.

No C [I ] term since all quantities are with respect to coordinate
frame {C } at centre of mass.
Rewrite above equations as

[
]
F
m [U] [0]

FC =
=
AC
C [I ]
[0]
NC
[U]: 3 3 identity matrix,
FC : 6 1 entity of external force and moment (see Module 5, Lecture
5).
.
.
.
.
.
.
AC : 6 1 entity of linear and angular acceleration.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

65 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM (C ONTD .)

Newton-Euler equations for an arbitrary point O

F = m[U](V O rC )
NO = C [I ] + rC F
where the centre of mass is located by rC from O.
In a compact form FO = [I ]AO where
[I ] is a 6 6 equivalent inertia matrix
[I ] consists of C [I ], m, [U], and
3 3 skew-symmetric matrix [rC ]

0
rCz
0
[rC ] = m rCz
rCy
rCx

rCy
rCx [U]
0

Above equations must be modified for rigid bodies connected by joints

Need to account for the
effects of link i + 1 through N on link i.
effects of generalized forces Qi+1 through QN .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

66 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM (C ONTD .)

Newton-Euler equations for an arbitrary point O

F = m[U](V O rC )
NO = C [I ] + rC F
where the centre of mass is located by rC from O.
In a compact form FO = [I ]AO where
[I ] is a 6 6 equivalent inertia matrix
[I ] consists of C [I ], m, [U], and
3 3 skew-symmetric matrix [rC ]

0
rCz
0
[rC ] = m rCz
rCy
rCx

rCy
rCx [U]
0

Above equations must be modified for rigid bodies connected by joints

Need to account for the
effects of link i + 1 through N on link i.
effects of generalized forces Qi+1 through QN .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

66 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED - BODY ALGORITHM (C ONTD .)

Newton-Euler equations for an arbitrary point O

F = m[U](V O rC )
NO = C [I ] + rC F
where the centre of mass is located by rC from O.
In a compact form FO = [I ]AO where
[I ] is a 6 6 equivalent inertia matrix
[I ] consists of C [I ], m, [U], and
3 3 skew-symmetric matrix [rC ]

0
rCz
0
[rC ] = m rCz
rCy
rCx

rCy
rCx [U]
0

Above equations must be modified for rigid bodies connected by joints

Need to account for the
effects of link i + 1 through N on link i.
effects of generalized forces Qi+1 through QN .
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

66 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

For an arbitrary link i, seek an equation of the form

A
Fi = [I ]A
i Ai + Pi

(46)

6 6 matrix [I ]i A : articulated-body inertia (ABI)

6 1 PiA : bias term containing effects of links after {i}.
Also seek to obtain [I ]i A and Pi A in O(N) steps!
Following formulas (Featherstone 1983, 1987) achieve the
requirements:
[I ]A
= [I ]i + [I ]A
i
i+1
A
PiA = Pi+1
+

T
A
[I ]A
i+1 Si+1 Si+1 [I ]i+1
T [I ]A S
Si+1
i+1 i+1

T
A
[I ]A
i+1 Si+1 (Qi+1 Si+1 Pi+1 )
T [I ]A
Si+1
i+1 Si+1

, [I ]A
N = [I ]N

, PNA = 0

(47)

Si+1 is a 6 1 entity representing the i + 1th joint axis7

7 Joint axis can be represented by a pair of 3 1 vectors (Q ; Q ) with Q Q = 0
0
0
(see Module 2, Additional Material).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

67 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

For an arbitrary link i, seek an equation of the form

A
Fi = [I ]A
i Ai + Pi

(46)

6 6 matrix [I ]i A : articulated-body inertia (ABI)

6 1 PiA : bias term containing effects of links after {i}.
Also seek to obtain [I ]i A and Pi A in O(N) steps!
Following formulas (Featherstone 1983, 1987) achieve the
requirements:
[I ]A
= [I ]i + [I ]A
i
i+1
A
PiA = Pi+1
+

T
A
[I ]A
i+1 Si+1 Si+1 [I ]i+1
T [I ]A S
Si+1
i+1 i+1

T
A
[I ]A
i+1 Si+1 (Qi+1 Si+1 Pi+1 )
T [I ]A
Si+1
i+1 Si+1

, [I ]A
N = [I ]N

, PNA = 0

(47)

Si+1 is a 6 1 entity representing the i + 1th joint axis7

7 Joint axis can be represented by a pair of 3 1 vectors (Q ; Q ) with Q Q = 0
0
0
(see Module 2, Additional Material).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

67 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

For an arbitrary link i, seek an equation of the form

A
Fi = [I ]A
i Ai + Pi

(46)

6 6 matrix [I ]i A : articulated-body inertia (ABI)

6 1 PiA : bias term containing effects of links after {i}.
Also seek to obtain [I ]i A and Pi A in O(N) steps!
Following formulas (Featherstone 1983, 1987) achieve the
requirements:
[I ]A
= [I ]i + [I ]A
i
i+1
A
PiA = Pi+1
+

T
A
[I ]A
i+1 Si+1 Si+1 [I ]i+1
T [I ]A S
Si+1
i+1 i+1

T
A
[I ]A
i+1 Si+1 (Qi+1 Si+1 Pi+1 )
T [I ]A
Si+1
i+1 Si+1

, [I ]A
N = [I ]N

, PNA = 0

(47)

Si+1 is a 6 1 entity representing the i + 1th joint axis7

7 Joint axis can be represented by a pair of 3 1 vectors (Q ; Q ) with Q Q = 0
0
0
(see Module 2, Additional Material).
.
.
.
.
.
.
A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

67 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

Articulated-body inertia and the bias term for the end-effector (link
N) is known (see the planar 2P example).
[I ]A
N = [I ]N , and
PNA = 0.
A
Start with i = N 1 and compute [I ]A
i and Pi for each i O(N)
algorithm.
A
i for each i as
From [I ]A
i and Pi , obtain q

The acceleration Ai is related to Ai 1 by

Ai = Ai 1 + Si qi , A0 = 0

(48)

The generalised force Qi is the component of Fi along Si

SiT Fi = Qi

(49)

Finally after simplification,

qi =

A
T
Qi SiT [I ]A
i Ai 1 Si Pi
A
T
Si [I ]i Si
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(50)
.

NPTEL, 2010

68 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

Articulated-body inertia and the bias term for the end-effector (link
N) is known (see the planar 2P example).
[I ]A
N = [I ]N , and
PNA = 0.
A
Start with i = N 1 and compute [I ]A
i and Pi for each i O(N)
algorithm.
A
i for each i as
From [I ]A
i and Pi , obtain q

The acceleration Ai is related to Ai 1 by

Ai = Ai 1 + Si qi , A0 = 0

(48)

The generalised force Qi is the component of Fi along Si

SiT Fi = Qi

(49)

Finally after simplification,

qi =

A
T
Qi SiT [I ]A
i Ai 1 Si Pi
A
T
Si [I ]i Si
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(50)
.

NPTEL, 2010

68 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)

Articulated-body inertia and the bias term for the end-effector (link
N) is known (see the planar 2P example).
[I ]A
N = [I ]N , and
PNA = 0.
A
Start with i = N 1 and compute [I ]A
i and Pi for each i O(N)
algorithm.
A
i for each i as
From [I ]A
i and Pi , obtain q

The acceleration Ai is related to Ai 1 by

Ai = Ai 1 + Si qi , A0 = 0

(48)

The generalised force Qi is the component of Fi along Si

SiT Fi = Qi

(49)

Finally after simplification,

qi =

A
T
Qi SiT [I ]A
i Ai 1 Si Pi
A
T
Si [I ]i Si
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

(50)
.

NPTEL, 2010

68 / 74

F ORWARD DYNAMICS OF SERIAL MANIPULATORS

A RTICULATED -B ODY A LGORITHM (C ONTD .)
A
Fixed base A0 = 0, compute q1 from [I ]A
1 , P1 and for a given Q1 .
Iterate for i = 1 to N to obtain all qi s.
Overall algorithm is O(N) since all sub-parts are O(N) (Featherstone
1983, 1987).
Root Node (Fixed)

1
2

Can be used for multi-body

system in a tree structure.

3
4
5

0 is the root node (fixed base).

10, 11 etc. are leaf nodes

(end-effectors)

7
8

9
11

Leaves (End-effector)

10

12

Figure 18: A typical tree structure

.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

69 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Recursive inverse and forward dynamics algorithms cannot be directly
applied to parallel manipulators and closed-loop mechanisms.
Presence of passive joints and loop-closure constraints relating passive
and active joints impractical to eliminate passive joints.
Equations of motion with m loop-closure constraints and m Lagrange
multipliers
)
[
](
) (
[C]q G F

q
[M] []T
=
q

[] [0]
[]
n + m equations in n qi s and m i s.
using Gaussian elimination O((n + m)3 )
Solve for and q
complexity.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

70 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Recursive inverse and forward dynamics algorithms cannot be directly
applied to parallel manipulators and closed-loop mechanisms.
Presence of passive joints and loop-closure constraints relating passive
and active joints impractical to eliminate passive joints.
Equations of motion with m loop-closure constraints and m Lagrange
multipliers
)
[
](
) (
[C]q G F

q
[M] []T
=
q

[] [0]
[]
n + m equations in n qi s and m i s.
using Gaussian elimination O((n + m)3 )
Solve for and q
complexity.
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

70 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Recursive inverse and forward dynamics algorithms cannot be directly
applied to parallel manipulators and closed-loop mechanisms.
Presence of passive joints and loop-closure constraints relating passive
and active joints impractical to eliminate passive joints.
Equations of motion with m loop-closure constraints and m Lagrange
multipliers
[
) (
)
](
[C]q G F

q
[M] []T
=
q

[] [0]
[]
n + m equations in n qi s and m i s.
using Gaussian elimination O((n + m)3 )
Solve for and q
complexity.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

70 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Recursive inverse and forward dynamics algorithms cannot be directly
applied to parallel manipulators and closed-loop mechanisms.
Presence of passive joints and loop-closure constraints relating passive
and active joints impractical to eliminate passive joints.
Equations of motion with m loop-closure constraints and m Lagrange
multipliers
[
) (
)
](
[C]q G F

q
[M] []T
=
q

[] [0]
[]
n + m equations in n qi s and m i s.
using Gaussian elimination O((n + m)3 )
Solve for and q
complexity.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

70 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Recursive inverse and forward dynamics algorithms cannot be directly
applied to parallel manipulators and closed-loop mechanisms.
Presence of passive joints and loop-closure constraints relating passive
and active joints impractical to eliminate passive joints.
Equations of motion with m loop-closure constraints and m Lagrange
multipliers
[
) (
)
](
[C]q G F

q
[M] []T
=
q

[] [0]
[]
n + m equations in n qi s and m i s.
using Gaussian elimination O((n + m)3 )
Solve for and q
complexity.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

70 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

I MPROVED ALGORITHM

Form [M] etc. terms using an O(n) inverse dynamics algorithm.

Form [] can be formed using a O(m2 ) algorithm.
Using O(m3 ) Gaussian elimination algorithm, solve from
q [][M]1 ( [C]q G F)
([][M]1 []T ) = []
Complexity of obtaining O(nm2 + m3 )
For known , parallel manipulator is equivalent to a serial
manipulator with extra right-hand side forcing terms.
using an O(n) algorithm.
Solve for q

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

71 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Efficient forward dynamics algorithm for large multi-body systems

(such as proteins) with closed-loops still a subject of research!
O(n + m3 ) algorithm called MEXX (Lubich, et al., 1992).
Sequential regularisation method iterative O(n) (Ascher & Lin 1999)
Requires k iterations for numerical convergence.
k claimed to be independent of the number of loops m!
k is small!

Parallel O(log n) algorithms are useful for very large multi-body

systems.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

72 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Efficient forward dynamics algorithm for large multi-body systems

(such as proteins) with closed-loops still a subject of research!
O(n + m3 ) algorithm called MEXX (Lubich, et al., 1992).
Sequential regularisation method iterative O(n) (Ascher & Lin 1999)
Requires k iterations for numerical convergence.
k claimed to be independent of the number of loops m!
k is small!

Parallel O(log n) algorithms are useful for very large multi-body

systems.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

72 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Efficient forward dynamics algorithm for large multi-body systems

(such as proteins) with closed-loops still a subject of research!
O(n + m3 ) algorithm called MEXX (Lubich, et al., 1992).
Sequential regularisation method iterative O(n) (Ascher & Lin 1999)
Requires k iterations for numerical convergence.
k claimed to be independent of the number of loops m!
k is small!

Parallel O(log n) algorithms are useful for very large multi-body

systems.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

72 / 74

F ORWARD DYNAMICS OF PARALLEL MANIPULATORS

Efficient forward dynamics algorithm for large multi-body systems

(such as proteins) with closed-loops still a subject of research!
O(n + m3 ) algorithm called MEXX (Lubich, et al., 1992).
Sequential regularisation method iterative O(n) (Ascher & Lin 1999)
Requires k iterations for numerical convergence.
k claimed to be independent of the number of loops m!
k is small!

Parallel O(log n) algorithms are useful for very large multi-body

systems.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

72 / 74

O UTLINE
.
. .1 C ONTENTS
.
. .2 L ECTURE 1
Introduction
Lagrangian formulation
.
. .3 L ECTURE 2
Examples of Equations of Motion
.
. .4 L ECTURE 3
Inverse Dynamics & Simulation of Equations of Motion

.
. .5 L ECTURE 4
Recursive Formulations of Dynamics of Manipulators

.
. .6 M ODULE 6 A DDITIONAL M ATERIAL
Maple Tutorial, ADAMS Tutorial, Problems, References and
Suggested Reading
.

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

73 / 74

M ODULE 6 A DDITIONAL M ATERIAL

Introduction to Maple and symbolic equations of motion using

R
Maple
ADAMS software for modeling and simulation of multi-body systems
Exercise Problems
References & Suggested Reading

A SHITAVA G HOSAL (IIS C )

ROBOTICS : A DVANCED C ONCEPTS & A NALYSIS

NPTEL, 2010

74 / 74