ME452

Robotics and Control

AR501
Robot Kinematics Dynamics and Control

Dr. Amit Shukla

Chairperson, Centre of AI and Robotics, CAIR@IIT Mandi
Indian Institute of Technology Mandi, Himachal Pradesh, India

Dr. Amit Shukla, Robotics & Control, IIT Mandi 1

 CHAPTER 4

Forward Kinematics

4
Content:

4.1 Product of Exponentials Formula

4.1.1 First Formulation: Screw Axes in the Base Frame
4.1.2 Examples
4.1.3 Second Formulation: Screw Axes in the End-Effector Frame
4.2 The Universal Robot Description Format
The forward kinematics of a robot refers to the calculation of the position and
orientation of its end-effector frame from its joint coordinates 𝜃.

The link lengths are L1, L2, and L3

a fixed frame {0}, an end-effector
frame {4}

Figure 4.1: Forward kinematics of a 3R planar open chain. For each frame, the 𝑥" and
𝑦" -axis is shown, the 𝑧̂ -axes are parallel and out of the page.
The Cartesian position (x, y) and orientation ∅ of the end-effector frame as functions
of the joint angles (𝜃1, 𝜃2, 𝜃3) are then given by

A more systematic method of deriving the forward kinematics might involve

attaching reference frames to each link, in Figure 4.1 the three link reference
frames are respectively labeled {1}, {2}, and {3}. The forward kinematics can
then be written as a product of four homogeneous transformation matrices:

T04 = T01T12T23T34
where

Observe that T34 is constant and that each remaining T 𝑖 −1,i depends only on the
joint variable 𝜃𝑖 . As an alternative to this approach, let us define M to be the
position and orientation of frame {4} when all joint angles are set to zero.
consider each revolute joint axis to be a zero-pitch screw axis. If 𝜃1 and 𝜃2 are held at
their zero position, then the screw axis corresponding to rotating about joint 3 can be
expressed in the {0} frame as

by simple visual inspection of 𝐹𝑖𝑔𝑢𝑟𝑒 4.1. When the arm is stretched out straight to the
right at its zero configuration, imagine a turntable rotating with an angular velocity of
𝜔! = 1 rad/s about the axis of joint 3. The linear velocity 𝜗! of the point on the
turntable at the origin of {0} is in the − 𝑦 ! 0 -direction at a rate of L1 +L2 units/s.
Algebraically, 𝜗3 = − 𝜔! × 𝑞! , where 𝑞! is any point on the axis of joint 3 expressed in
{0}, e.g., 𝑞! = (L1 + L2, 0, 0).
The screw axis S3 can be expressed in se(3) matrix form as

for any 𝜃3, the matrix exponential representation for screw

motions from the previous chapter allows us to write

for 𝜃1 = 0 and any fixed (but arbitrary) 𝜃3, rotation about joint 2 can be viewed as
applying a screw motion to the rigid (link 2)/(link 3) pair, i.e.,

where [S3] and M are as defined previously, and

The forward kinematics can be expressed

as a product of matrix exponentials, Each
corresponding to a screw motion.
 4.1. Product of Exponentials Formula
Finally, keeping θ2 and θ3 fixed, rotation about joint
1 can be viewed as applying a screw motion to the
nd entire
✓3 fixed, rotation
rigid about
three-link joint 1 can
assembly. We be viewed
can as applying
therefore
he write,
entire for
rigid three-link
arbitrary assembly.
values We
of (θ1, θ2, can therefore write,
θ3),
of (✓1 , ✓2 , ✓3 ),

T04 = e[S1 ]✓1 e[S2 ]✓2 e[S3 ]✓3 M, (4.10)

2 3
0 1 0 0
Thus, the forward kinematics can be
6 1 0 0 0 7
[S1 ] = 6 7. expressed as a product
(4.11) of matrix
4 0 0 0 0 5 exponentials, each corresponding to a screw
0 0 0 0 motion. Note that this latter derivation of
the forward kinematics does not use any
nematics can be expressed as a productlink reference
of matrix frames; only {0} and M must
exponen-
ding to a screw motion. Note that thisbelatter
defined.
derivation of
ics does not use any link reference frames; only {0} and M
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1 PRODUCT OF EXPONENTIALS FORMULA

To use the PoE formula, it is only necessary to assign a stationary frame {s} and a
frame {b} at the end-effector, described by M (zero position matrix). It is common to
define a frame at each link, though, typically at the joint axis, these are needed for
the D–H representation and they are useful for displaying a graphic rendering of a
geometric model of the robot and for defining the mass properties of the link, Thus
when we are defining the kinematics of an n-joint robot, we may either

• minimally use the frames {s} and {b} if we are only interested in the kinematics,
• Refer to {s} as frame {0}, use frames {𝑖} for 𝑖 = 1, . . . ,n (the frames for links 𝑖
at joints 𝑖), and use one more frame {n+1} (corresponding to {b}) at the end-
effector. The frame {n + 1} (i.e., {b}) is fixed relative to {n}, but it is at a more
convenient location to represent the configuration of the end-effector. In some
cases we dispense with frame {n+1} and simply refer to {n} as the end-effector
frame {b}.
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.1 FIRST FORMULATION: SCREW AXES IN THE BASE FRAME

Figure 4.2: Illustration of the PoE formula for an n-link spatial open chain
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.1 FIRST FORMULATION: SCREW AXES IN THE BASE FRAME

Figure 4.2, consisting of n 1-DOF joints that are connected serially. To apply the PoE
formula,
• choose a fixed base frame {s} and an end-effector frame {b} attached to the last
link.
• Place the robot in its zero position by setting all joint values to zero, with the direction
of positive displacement (rotation for revolute joints, translation for prismatic joints)
for each joint specified.
• M ∈ SE(3) denote the configuration of the end-effector frame relative to the fixed
base frame when the robot is in its zero position.
Now suppose that joint n is displaced to some joint value 𝜃n. The end-effector frame M
then undergoes a displacement of the form

T ∈ SE(3) is the new configuration of the end-

effector frame and Sn = (𝜔n, 𝜗n) is the screw axis of
joint n as expressed in the fixed base frame.
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.1 FIRST FORMULATION: SCREW AXES IN THE BASE FRAME

If we assume that joint (n−1) is also allowed to vary then this has the effect of
applying a screw motion to link (n−1). The end-effector frame thus undergoes a
displacement of the form

Continuing with this reasoning and now allowing all the joints (𝜃1, . . . , 𝜃n) to vary, it
follows that
(a)
(space form of the product of exponentials formula)

the product of exponentials formula describing the forward kinematics of an n-DOF

open chain.
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.1 FIRST FORMULATION: SCREW AXES IN THE BASE FRAME

to calculate the forward kinematics of an open chain using the space form of the PoE
formula (a), we need the following elements:

• The end-effector configuration M∈SE(3) when the robot is at its home position,

• The screw axes S1, . . . , Sn expressed in the fixed base frame, corresponding to the
joint motions when the robot is at its home position.

• The joint variables 𝜃1, . . . ,𝜃n.

4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES
Example 4.1 (3R spatial open chain). Consider the 3R open chain of Figure 4.3,
shown in its home position (all joint variables set equal to zero). Choose the fixed
frame {0} and end-effector frame {3} as indicated in the figure and express all
vectors and homogeneous transformations in terms of the fixed frame.

Figure 4.3: A 3R spatial open chain.

4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

The forward kinematics has the form

where M∈SE(3) is the end-effector frame configuration when the robot is in its zero
position. By inspection M can be obtained as

The screw axis S1=(𝜔1,𝜗1) for joint axis 1 is then given by 𝜔1 =(0,0,1) and 𝜗1 =(0,0,0)
To determine the screw axis S2 for joint axis 2, observe that joint axis 2 points in the
-𝑦" 0-direction, so that 𝜔2=(0,-1,0). Choose q2=(L1,0,0), in which case 𝜗2 =- 𝜔2, q2=(0,
0,-L1). Finally, to determine the screw axis S3 for joint axis 3, note that 𝜔3=(1,0,0).
Choosing q3=(0,0,-L2), it follows that 𝜗3=- 𝜔3 q3=(0,-L2,0).
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

we have the following 4×4 matrix representations for the three joint screw axes S1,
S2,and S3:

It will be more convenient to list the screw axes in the following tabular form:
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

Example 4.2 (3R planar open chain). For the robot in Figure 4.1, we expressed the
end-effector home configuration M and the screw axes Si as follows:

Since the motion is in the 𝑥" -𝑦-plane,

" we could equivalently write each screw axis Si as
a 3-vector (𝜔z,𝜗x, 𝜗y):
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

and M as an element of SE(2):

In this case, the forward kinematics would use the simplified matrix exponential for
planar motions.

Figure: 4.1
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES
Example 4.3 (6R spatial open chain). We now derive the forward kinematics of the
6R open chain of Figure 4.4. 6-DOF arms play an important role in robotics because
they have the minimum number of joints that allows the end-effector to move a rigid
body in all its degrees of freedom, subject only to limits on the robot's workspace. For
this reason, 6-DOF robot arms are sometimes called general purpose manipulators.
The zero position and the direction of positive rotation for each joint axis are as
shown in the figure. A fixed frame {s} and end-effector frame {b} are also assigned
as shown. The end-effector frame M in the zero position is then
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

Figure 4.4: PoE forward kinematics for the 6R open chain.

4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.2 EXAMPLES

Example 4.4 (An RRPRRR spatial open chain). In this example we consider the 6-
degree-of-freedom RRPRRR spatial open chain of Figure 4.5. The end-effector frame
in the zero position is given by
The screw axes 𝑆" = (𝜔" , 𝜗" ) are listed in the
following table

Note that the third joint is prismatic, so that 𝜔! = 0 and 𝜗! is a unit vector in the
direction of positive translation.
4.1 PRODUCT OF EXPONENTIALS FORMULA
4.1.3 Second Formulation: Screw Axes in the End-Effector Frame
# "# $# "# $#
The matrix identity 𝑒 = 𝑀%& 𝑒 $ M
can also be expressed as M𝑒 # = 𝑒$M .
M
Beginning with the rightmost term of the previously derived product of exponentials
formula, if we repeatedly apply this identity then after n iterations, we obtain

------------(b)

where each [𝛽" ] is given by 𝑀%& [𝑆" ]M, i.e., 𝛽 i =[Ad 𝑀%& ] 𝑆" , i = 1,…..,n. Equation (b) is
an alternative form of the product of exponentials formula, representing the joint axes
as screw axes 𝛽" in the end-effector (body) frame when the robot is at its zero position.
We call Equation (b) the body form of the product of exponentials formula.
4.2 the universal robot description format
4.2 The universal robot description format
The Universal Robot Description Format (URDF) is an XML (eXtensible Markup Language)
file format used by the Robot Operating System (ROS) to describe the kinematics, inertial
properties, and link geometry of robots. A URDF le describes the joints and links of a
robot:
Joints. Joints connect two links: a parent link and a child link. A few of the possible joint
types include prismatic, revolute, continuous, and fixed. Each joint has an origin frame
that defines the position and orientation of the child link frame relative to the parent link
frame when the joint variable is zero. The origin is on the joint's axis. Each joint has an
axis 3-vector, a unit vector expressed in the child link's frame, in the direction of positive
rotation for a revolute joint or positive translation for a prismatic joint.
Links. While the joints fully describe the kinematics of a robot, the links define its mass
properties. These start to be needed in Chapter 8, when we begin to study the dynamics
of robots. The elements of a link include its mass, an origin frame that defines the
position and orientation of a frame at the link's center of mass relative to the link's joint
frame described above, and an inertia matrix, relative to the link's center of mass frame,
specified by the six elements on or above the diagonal.
4.2 the universal robot description format

A URDF le can represent any robot with a tree structure. This includes serial-chain
robot arms and robot hands, but not a Stewart platform or other mechanisms with
closed loops. An example of a robot with a tree structure is shown in Figure 4.10.

The orientation of a frame {b} relative to a frame fag is represented using Roll, Pitch,
yaw coordinates: first, a roll about the fixed 𝑥" a-axis, then a pitch about the fixed 𝑦" a-
axis, then a yaw about the fixed 𝑧̂ a-axis.
4.2 the universal robot description format

Figure 4.10: A five-link robot

represented as a tree, where
the nodes of the tree are the
links and the edges of the tree
are the joints
THE DENAVIT-HARTENBERG CONVENTION
 erive the forward kinematics from the knowledge of the relative displacements
The Denavit-Hartenberg convention
etween adjacent link frames. Assume that a fixed reference frame has been es-
The basic idea underlying the Denavit–Hartenberg approach to forward kinematics is to
ablished and thatframes
attach reference a reference
to eachframe
link of(the
the end-e↵ector
open chain and frame)
then tohas beenthe
derive attached
forward
kinematics
o some pointfromonthe
theknowledge of of
last link the the
relative
open displacements
chain. For between
a chain adjacent link frames.
consisting of n
ne-degree-of-freedom joints, the links are numbered sequentially
For a chain consisting of n one-degree-of-freedom joints, the links are numbered
from 0 to n:
hesequentially
ground link fromis 0labeled 0, and
to n: the groundthelink
end-e↵ector
is labelled 0,frame is attached
and the to frame
end-effector link n. is
attached frames
Reference to link n.attached to the links are also correspondingly labeled from
0}Reference
(the fixed frame)
frames to {n}
attached to (the end-e↵ector
the links frame). The joint
are also correspondingly variable
labelled from {0} corre-
(the
ponding to the
fixed frame) ith joint
to {n} is denotedframe).
(the end-effector ✓i . TheThe forward kinematics
joint variable of the
corresponding to n-link
the ith
joint
pen is denoted
chain can thenθi. The
beforward kinematics
expressed as of the n-link open chain can then be expressed
as

T0n (✓1 , . . . , ✓n ) = T01 (✓1 )T12 (✓2 ) · · · Tn 1,n (✓n ), (C.1)

where Ti,i𝑇"%&,"
where 1 ∈ 2SE(3)
SE(3) denotes
denotes the relative
the relative displacement
displacement between linkbetween
frames {𝑖 link frames
− 1} and
i {𝑖}. and {i}. on
1}Depending Depending
how the linkon how the
reference link have
frames reference frameseach
been chosen, have been
𝑇"%&," can chosen,
be
achobtained
Ti 1,i incan a straightforward
be obtained fashion.
in a straightforward fashion.
 The Denavit-Hartenberg convention
C.1. Assigning Link Frames
The first rule is that the 𝑧̂! -axis coincides with joint axis 𝑖 and the
𝑧̂!"# -axis coincides with joint axis 𝑖 − 1.
axis i 1
axis i
Determines the origin of the link reference frame: First, find the line
segment that orthogonally intersects both the joint axes
𝑧̂!"# and 𝑧̂! .

ẑi Connecting joint axes 𝑖 − 1 and 𝑖 by a mutually perpendicular line,

↵i 1 ŷi x̂i di the origin of frame {𝑖 − 1} is then located at the point where
ẑi
this line intersects joint axis 𝑖 − 1.
1 i
ŷi 1 ai 1
x̂i Determining the remaining 𝑥*! - and 𝑦*! -axes of each link reference
1
frame is now straightforward: the 𝑥*! -axis is chosen to be in the
direction of the mutually perpendicular line pointing from the
(𝑖 − 1) −axis to the 𝑖-axis. The 𝑦*! -axis is then uniquely
Figure C.1: Illustration of the Denavit–Hartenberg parameters. determined from the cross product 𝑥*! ×𝑦*! = 𝑧̂! .

erved. Figure C.1 illustrates the frame-assignment convention for two

ent revolute joints i 1 and i that are connected by link i 1.
 The Denavit-Hartenberg convention
C.1. Assigning Link Frames
Having assigned reference frames in this fashion for links 𝑖 and 𝑖 −
axis i 1 1, we now define four parameters that exactly specify 𝑇!"#,! :
axis i
• The length of the mutually perpendicular line, denoted by the
scalar 𝑎!"# , is called the link length of link 𝑖 − 1. Despite its
name, this link length does not necessarily correspond to the
actual length of the physical link.
ẑi • The link twist 𝛼!"# is the angle from 𝑧̂!"# to 𝑧̂! , measured about
↵i ŷi x̂i di
1
𝑥*!"# .
ẑi 1 i
• The link offset 𝑑! is the distance from the intersection of 𝑥*!"#
ŷi 1 ai 1
and 𝑧̂! to the origin of the link-𝑖 frame (the positive direction is
x̂i 1
defined to be along the 𝑧̂! - axis).
• The joint angle ∅! is the angle from 𝑥*!"# to 𝑥*! , measured about
the 𝑧̂! - axis.

Figure C.1: Illustration of the Denavit–Hartenberg parameters.

erved. Figure C.1 illustrates the frame-assignment convention for two

ent revolute joints i 1 and i that are connected by link i 1.
 The Denavit-Hartenberg convention
C.1. Assigning Link Frames
Having assigned reference frames in this fashion for links 𝑖 and 𝑖 −
axis i 1 1, we now define four parameters that exactly specify 𝑇!"#,! :
axis i
• The length of the mutually perpendicular line, denoted by the
scalar 𝑎!"# , is called the link length of link 𝑖 − 1. Despite its
name, this link length does not necessarily correspond to the
actual length of the physical link.
ẑi • The link twist 𝛼!"# is the angle from 𝑧̂!"# to 𝑧̂! , measured about
↵i ŷi x̂i di
1
𝑥*!"# .
ẑi 1 i
• The link offset 𝑑! is the distance from the intersection of 𝑥*!"#
ŷi 1 ai 1
and 𝑧̂! to the origin of the link-𝑖 frame (the positive direction is
x̂i 1
defined to be along the 𝑧̂! - axis).
• The joint angle ∅! is the angle from 𝑥*!"# to 𝑥*! , measured about
the 𝑧̂! - axis.

Figure C.1: Illustration of the Denavit–Hartenberg parameters.

erved. Figure C.1 illustrates the frame-assignment convention for two

ent revolute joints i 1 and i that are connected by link i 1.
 The Denavit-Hartenberg convention
C.1. Assigning Link Frames

axis i 1 These parameters constitute the Denavit

axis i Hartenberg (D–H) parameters. For an open
chain with 𝑛 one-degree-of-freedom joints,
the 4𝑛 D–H parameters are sufficient to
completely describe the forward
ẑi kinematics.
↵i 1 ŷi x̂i di

ẑi 1 i In the case of an open chain with all joints

ŷi 1 ai 1
revolute, the link lengths 𝑎"%& , twists 𝛼"%& ,
x̂i 1
and offset parameters 𝑑" are all constant,
while the joint angle parameters ∅" act as
the joint variables.
Figure C.1: Illustration of the Denavit–Hartenberg parameters.

erved. Figure C.1 illustrates the frame-assignment convention for two

 The Denavit-Hartenberg convention

We now consider the cases where the mutually perpendicular line is undefined or fails to
be unique, or where some of the joints are prismatic; finally, we consider how to choose
the ground and end-effector frames.
1. When Adjacent Revolute Joint Axes Intersect: If two adjacent revolute
joint axes intersect each other then a mutually perpendicular line between 586 C.1. Assigning Link Fr
the joint axes fails to exist. In this case the link length is set to zero, and we
axis i 1
choose 𝑥!!"# to be perpendicular to the plane spanned by 𝑧!"# ̂ and 𝑧!̂ . There axis i

are two possibilities, both of which are acceptable: one leads to a positive
value of the twist angle 𝛼!"# while the other leads to a negative value.
2. When Adjacent Revolute Joint Axes Are Parallel: The second special case ↵i
ẑi
1 ŷi x̂i di
occurs when two adjacent revolute joint axes are parallel. In this case there
ẑi 1 i
ŷi
exist many possibilities for a mutually perpendicular line, all of which are 1

x̂i
ai 1

valid A useful guide is to try to choose the mutually perpendicular line that
is the most physically intuitive and that results in as many zero parameters
as possible. Figure C.1: Illustration of the Denavit–Hartenberg parameters.

is observed. Figure C.1 illustrates the frame-assignment convention for

adjacent revolute joints i 1 and i that are connected by link i 1.
 588 C.1. Assigning Link Fra
The Denavit-Hartenberg
convention
axis i 1
Prismatic Joints:
axis i
For prismatic joints, the 𝑧!̂ -
ẑi
direction of the link reference ↵i 1 ŷi
frame is chosen to be along the x̂i
ẑi 1 di
positive direction of translation.
This convention is consistent with ŷi 1 ai 1
x̂i 1
that for revolute joints, in which i

the 𝑧!̂ -axis indicates the positive

axis of rotation. With this choice
the link offset 𝑑! is the joint
variable and the joint angle ∅Figure
! is
C.2: Link frame assignment convention for prismatic joints. Joint i 1
revolute joint, while joint i is a prismatic joint.
constant.

Prismatic Joints
For prismatic joints, the ẑ-direction of the link reference frame is chosen to
 Once
rm all the transformations
transformations T between T i
adjacent 1,i T=iframes
between
link 6 sin
1,i =arei cos
Rot(x̂,
adjacent known
↵i link
1↵i cos
in
i cos ↵are
)Trans(x̂,
1frames i 1ai known sin ↵i in
1 )Trans(ẑ,1 did)Rot(ẑ,
i sin ↵i 1 i )
7,
The Denavit-Hartenberg convention sin i 2
i 1,i 4 sin ↵i 1 cos i sin ↵i 1 cos ↵i 1 di cos ↵i 1 5
terms
ir =D–H
of their
parameters,
Rot(x̂,
D–H theparameters,
↵i 1 )Trans(x̂, forward kinematics
ai 1 )Trans(ẑ,
the forward
di )Rot(ẑ,is obtained kinematics
by0 sequen- cos is obtained
i bysin sequen-
0i 01
i) 0
Once
tially
lying all link
2these the transformations
multiplying these
transformations. link !"#,! between
𝑇transformations.
Each link frame adjacent Each 6
transformationlinksin frames
link iscos are
3 iframe known in terms of
↵i transformation
1 cos i cos ↵i is
theirsinD–H ↵i 1 d
cos sin 0 = a 6 1
of parameters, the forward kinematics is obtained by isequentially multiplying these link
i i 1
the form
6 sin i cos ↵i 1Each where 4 sin 7 i sin ↵ i 1 cos i sin ↵ i 1 cos ↵i 1 di
transformations. cos link cos ↵
frame sin ↵
transformation d issin
of ↵ the form 2 3
= 6 i i 1 i 1 i i 1 7,
4 sin i sin ↵i 1 cos i sin ↵i 1
Rot(x̂, ↵i 1 )Trans(x̂, ai 1 )Trans(ẑ, di )Rot(ẑ, cos ↵i 1 ) di cos ↵i 1 5 0 1 0 0 0 0 0
i
2Ti 1,i =0 Rot(x̂, ↵i 10)Trans(x̂, ai0 1 )Trans(ẑ, 1 di )Rot(ẑ,
3 6 6 0i ) cos ↵i 1 sin ↵i 1 0 7 7,
cos i 2 sin i 0where Rot(x̂, ai ↵1i 1 ) = 4 0 sin ↵i 1 cos ↵ 3 0 5
(C.2
6 sin i cos ↵i 1 cos cos sin 0 a2i 1 i 1
cos ii 1
↵ sin ↵ i di sin ↵i 1 7
0 0 0 1
3
6 i i 1 7 1↵ 3 7 0 0 0
4 sin i sin ↵i 6 sin
cos 2 cos
sin ↵ ↵ cos
cos ↵ cos ↵
d 3
cos ↵ sin 5 ↵, 2 d sin
= 6 4
1 ii1 i
i0 1 1 i
i0 1 0i
i 1 i 1 i 1
1 0
i
0 6
6 a0i
i
1
1
cos 7
↵
5 i,1 sin ↵ i 1 0 7
7
0 sin6 i0sin 0 cos↵i↵i 11 cos 0 ↵i sin ↵0i 711 Rot(x̂,
sin cos ↵ ↵
i 1
6 ) d
= i cos ↵ i 17
6 0 1 04 00 sin 0 5
i 1
Rot(x̂, ↵i 1 ) = 6 i 1
Trans(x̂, 7 , a ) (C.2) = 7 ,↵i 1 cos ↵i 1 (C.3
4 0 0 sin ↵i 1 cos ↵i 01 0 5 i 1 0 4 0 0 11 0 5
0 0 0 1
2 0 0 0 13 0 0 02 1 3
where 2
1 0 3 0 0 2 1 30 0 ai 1
1 0 0 a i 1 1 0 306 0
6 6
6 0 cos ↵ i 1 2sin
7 ↵ i 1 0 7
7 6 0 7 1 0 0 7
Rot(x̂, ↵i a1 ) )= =4 6 0 1 0 0 7 1 0 5 , Trans(x̂, (C.2)
0 a 6 0 0=
) 1 0 0 76 7,
Trans(x̂, i 1 ,
0 0 sin0↵i1 1 0cos5↵i 1 0Trans(ẑ, di ) = 4
4 (C.3) i 1 4 0i 50, 1 0 5 (C.4
0 0 00 0 6 00 cos 1↵i 1
0
sin ↵i 1 0 7 0 0 0 0 71 d
Rot(x̂, ↵
2 2 i 1 ) = 1 63 0 0 0 , 1 (C.2) 1
4 0 52 33
1 10 0 0 0 ai 01 3 0 sin ↵i 1 cos ↵i 2 1
cos i 1 1 sin0 i 01 00 0
6 6
6 0 01 1 0 0 00 77 0 7 0 0 6 sin1 6 cos
0 1 0 00 7
0 7
Trans(x̂, a ) =
Trans(ẑ, di ) =4 4
i 1 6 2
7 ,
0 00 0 1 1 0di 55, 1 0 0 Rot(ẑ, 3 (C.3)
i ) (C.4)
Trans(ẑ, = 4 6
di ) =
i 1 6 i 1 7 ,7 . (C.5
ai 1 0 4 0 00 1 d 1i 5 0 5
0 00 0 0 0 11 6
36 0 1 0 3 0 7
0 0 00 0 01 1
2 2 7,
Trans(x̂,1 acos i0 1 )0 =0 4 0i 1 0 1
sin 0 0 5 2 (C.3) 3
6 6
i 1
7 2017 preprint of Modern
May
0
7 Robotics, Lynch and Park, Cambridge cosU. Press,
i 1 2017. sin i 1 0 0
http://modernrobotics.or
 The Denavit-Hartenberg convention
C.1. Assigning LinkAFrames
useful way to visualize 𝑇!"#,! is that it transports
frame {𝑖 − 1} to frame {𝑖} via the following
axis i 1
axis i
sequence of four transformations:
(a) A rotation of frame {𝑖 − 1} about its 𝑥* -axis by an
angle 𝛼!"# .
(b) A translation of this new frame along its 𝑥* -axis by
a distance 𝑎!"# .
ẑi
(c) A translation of the new frame formed by (b)
↵i 1 ŷi x̂i di
along its 𝑧̂ -axis by a distance 𝑑! .
ŷi
ẑi 1
ai
i (d) A rotation of the new frame formed by (c) about
1 1
x̂i 1
its ˆz-axis by an angle ∅! .

Note that switching the order of the first and second

steps will not change the final form of 𝑇!"#,! .
Figure C.1: Illustration of the Denavit–Hartenberg parameters. Similarly, the order of the third and fourth steps
can also be switched without affecting 𝑇!"#,! .
erved. Figure C.1 illustrates the frame-assignment convention for two
ent revolute joints i 1 and i that are connected by link i 1.
 Example C.1 (A 3R spatial open chain). Consider the 3R spatial ope
of Figure 4.3, shown in its zero position (i.e., with all its joint varia
The Denavit-Hartenberg
to zero).convention
The assigned link reference frames are shown in the figure,
corresponding D–H parameters are listed in the following table:

i ↵i 1 ai 1 di i
1 0 0 0 ✓1
2 90 L1 0 ✓2 90
3 90 L2 0 ✓3

Note that framesNote{1} and

that{2} are uniquely
frames {1} andspecified
{2} arefrom our frame
uniquely
ment convention, but specified
that we havefrom some ourlatitude
framein choosing
assignment frames {
{3}. Here we choose convention, but that
the ground frame {0}weto have some
coincide latitude
with framein{1} (r
in ↵0 = a0 = d1 = 0) choosing
and frameframes{3} {0} andsuch
to be Herex̂we=choose
{3}. that x̂2 (resultin
3
the ground frame {0} to coincide with frame
o↵set to the joint angle ✓3 ).
{1} (resulting in 𝛼 = 𝑎 = 𝑑 = 0) and frame {3}
% % #
to be such
Example C.2 (A spatial RRRP that
open 𝑥*' (resulting
𝑥*& =chain). in no
The next offset we c
example
to the joint angle 𝜃& ).
is the four-dof RRRP spatial open chain of Figure C.4, here shown in
position. The link frame assignments are as shown, and the correspondi
parameters are listed in the figure.
The four joint variables are (✓1 , ✓2 , ✓3 , ✓4 ), where ✓4 is the displace
the prismatic joint. As in the previous example, the ground frame {0} a
 The Denavit-Hartenberg convention

Lets consider a four-dof RRRP spatial open chain of

Figure below, here shown in its zero position. The
link frame assignments are as shown, and the i ↵i 1 ai 1 di i
corresponding D–H parameters are listed in the
table.
592 C.4. Examples 1 0 0 0 ✓1
✓4 2 90 0 0 ✓2
ẑ1 ŷ1
ŷ2 x̂3
3 0 L2 0 ✓3 + 90
x̂1 L2 ŷ3
x̂2
ẑ2
4 90 0 ✓4 0
x̂4
ẑ3
✓2 ✓1
ẑ4 The four joint variables are (θ1, θ2, θ3,
ẑ0 ŷ0
✓3
ŷ4 Figure
θ4), C.4:
whereAn
θ4RRRP
is the spatial open chain.
displacement of
the prismatic joint. As in the previous
x̂0
example, the ground frame {0} and
link frame {4} have finalchosen
been link frame {4} have
to make been of
as many chosen
the D–H p
to make as many of the D–H
as possible. parameters zero as possible.

Example C.3 (A spatial 6R open chain). The final example

i ↵i 1 ai 1 di i
1 0 0 0 the
✓ widely used 6R robot arm (Figure C.5). This open chain ha
1
The Denavit-Hartenberg convention
Existence and uniqueness issues:-
Clearly it is not possible to represent any arbitrary homogeneous transformation using
only four parameters. Therefore, we begin by determining just which homogeneous
transformations can be expressed in the form (1). Suppose we are given two frames,
denoted by frames 0 and 1, respectively. Then there exists a unique homogeneous
transformation matrix A that takes the coordinates from frame 1 into those of frame 0.
Now suppose the two frames have the following two additional features.

Coordinate frames satisfying assumptions DH1 and DH2

The Denavit-Hartenberg convention
DH Coordinate Frame Assumptions
(DH1) The axis 𝑋" is perpendicular to the axis 𝑍# .
(DH2) The axis 𝑋" intersects the axis 𝑍# .
Under these conditions, we claim that there exist unique numbers a,d,𝛼 ,𝜃 such that

If (DH1) is satisfied, then 𝑋" is perpendicular to 𝑍# and we have 𝑋" ·𝑍# = 0.

The Denavit-Hartenberg convention
Since 𝑟$" =0, we now need only show that there exist unique angles and such that

Since each row and column of 𝑅#" must have unit length, 𝑟$" = 0 implies that

There exist unique and such that

assumption (DH2) means that the displacement between 𝑂# and 𝑂" can be expressed
as a linear combination of the vectors 𝑍# and 𝑋" . This can be written as
The Denavit-Hartenberg convention
we see that four parameters are enough to specify any homogeneous transformation that
satisfies the constraints (DH1) and (DH2). physical interpretation to each of the four
quantities in eq(1).

• The parameter ‘a’ is the distance between the axes 𝑍# and 𝑍" and is measured along the
axis 𝑋" .
• The angle 𝛼 is the angle between the axes 𝑍# and 𝑍" , measured in a plane normal to
𝑋" .The positive sense for is determined from 𝑍# to 𝑍" by the right-handed rule.
• The Parameter d is the perpendicular distance from the origin 𝑂# to the intersection of
the 𝑋" axis with 𝑍# measured along the 𝑍# axis.
• 𝜃 is the angle between 𝑋# and 𝑋" measured in a plane normal to 𝑍# .

These physical interpretations will prove useful in developing a procedure for assigning
coordinate frames that satisfy the constraints (DH1) and (DH2), and we now turn our
attention to developing such a procedure.
SUMMARY
• An open chain with a fixed reference frame {s} and a reference frame {b} attached to
some point on its last link – this frame is denoted the end-effector frame – the forward
kinematics is the mapping T(𝜃) from the joint values 𝜃 to the position and orientation of
{b} in {s}.

• In the D–H representation the forward kinematics of an open chain is described in terms
of the relative displacements between reference frames attached to each link. If the link
frames are sequentially labeled {0}, . . . , {n + 1}, where {0} is the fixed frame {s},
{i} is a frame attached to link i at joint i (with i = 1, . . . ,n), and {n + 1} is the end-
effector frame {b} then the forward kinematics is expressed as
𝑇#,&'" 𝜃 = 𝑇#" 𝜃" … … … … … . 𝑇&(",& 𝜃& 𝑇&,&'"
𝜃! = joint i variable
𝑇&,&'" = indicates the (fixed) configuration of the end-effector frame in {n}.
If the end-effector frame {b} is chosen to be coincident with {n} then we can dispense
with the frame {n + 1}.
• The D–H convention requires that reference frames assigned to each link obey a strict
convention. Following this convention, the link frame transformation 𝑇!(" , 𝑖 between link
frames { 𝑖 −1} and { 𝑖 } can be parametrized using only four parameters, the D–H
parameters. Three of these parameters describe the kinematic structure, while the
fourth is the joint value. Four numbers is the minimum needed to represent the
displacement between two link frames.
• The forward kinematics can also be expressed as the following product of exponentials
(the space form),
T 𝜃 = 𝑒 [*#],# … … … … … … 𝑒 [*$ ],$ 𝑀
where 𝑆! =(𝜔! ,𝜗! ) denotes the screw axis associated with positive motion along joint 𝑖
expressed in fixed-frame {s} coordinates, 𝜃! is the joint- 𝑖 variable, and M ∈ 𝑆𝐸 3
denotes the position and orientation of the end-effector frame {b} when the robot is in
its zero position. It is not necessary to define individual link frames; it is only necessary
to define M and the screw axes 𝑆" … … … … . . 𝑆& .

• The product of exponentials formula can also be written in the equivalent body form,
T 𝜃 = 𝑀𝑒 [-#],# … … … … … … 𝑒 [-$ ],$
where 𝐵! = 𝐴𝑑.(" 𝑆! , 𝑖 = 1, … … . , 𝑛; 𝐵! = (𝜔! ,𝜗! ) is the screw axis corresponding to joint
axis 𝑖, expressed in {b}, with the robot in its zero position.

• The Universal Robot Description Format (URDF) is a file format used by the Robot
Operating System and other software for representing the Kinematics, inertial
properties, visual properties, and other information for general tree-like robot
mechanisms, including serial chains. A URDF file includes descriptions of joints, which
connect a parent link and a child link and fully specify the kinematics of the robot, as
well as descriptions of links, which specify its inertial properties.