direct and inverse kinematics

Direct and inverse kinematics

Robot usually directly measures its inner kinematic parameters - joint coordinates.Those coordinates measure the position of joints. We denote them usually as ~q, jointcoordinate of the revolute joint is denoted as , joint coordinate of the prismatic jointis denoted as d.

User is interested in the position of the end effector or the position of themanipulated rigid body. It has 6 DOF and it could be described in number of ways,e.g. by the transformation matrix describing position of the end effector coordinatesystem in the world coordinate system.

We are interested in the mapping between those two descriptions of the robotposition.

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Students often confuse position of the end effector (6DOF) with the position of the center of the gripper (point- has 3 DOF). This is a crucial error as orientation of the

gripper is important for both manipulation or operation (e.g.welding).

ROBOTICS: Vladimr Smutn Slide 1, Page 1

Direct kinematics

Direct (forward) kinematics is a mapping from joint coordinate space to space ofend-effector positions. That is we know the position of all (or some) individual jointsand we are looking for the position of the end effector. Mathematically:

~q T(~q)

Direct kinematics could be immediately used in coordinate measurement systems.Sensors in the joints will inform us about the relative position of the links, jointcoordinates. The goal is to calculate the position of the reference point of themeasuring system.

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Let us emphasize, that direct kinematics is mapping, nota mathematical function. It could have none, one, several, or

infinitely many solutions for particular ~q.


Inverse kinematics

Inverse kinematics is a mapping from space of end-effector positions to jointcoordinate space . That is we know the position of the end effector and we arelooking for the coordinates of all individual joints. Mathematically:

T ~q(T)

Inverse kinematics is needed in robot control, one knows the required position of thegripper, but for control the joint coordinates are needed.

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Let us emphasize, that inverse kinematics is mapping, nota mathematical function. It could have none, one, several, or

infinitely many solutions for particular ~q.


Open kinematic chain

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Open kinematic chain modelling

Open kinematic chain is formed by the sequence of links con-nected by joints. If we know the description joints using ge-

ometrical tranformations we can easily find transformationof point coordinates from end effector coordinate system tobase coordinate system and vice versa. This transformationis called kinematic equations.


Open kinematic chain modeling in plane

y1

x1

y3x3y2

x2

y5

x5y4

x4

y0

x0

y6

x6

l1

d2

l3

1

2

3

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Homogeneous coordinate system transformation could bedescribed by transformation matrix A

x = Axb,

A =[R xo0 0 1

]=

cos() sin() xosin() cos() yo0 0 1

,where is relative rotation of second coordinate system tofirst coordinate system. Inverse matrix is immediately:

A1 =[RT RTxo0 0 1

]= (1)

=

cos() sin() cos()xo sin()yo sin() cos() sin()xo cos()yo0 0 1

, (2)The simple serial planar manipulator is shown on the figure.The manipulator consist of the revolute joint (joint variable1), link of the length l1, then there is a prismatic joint withthe joint variable d2, which is tilted by angle 2. Consequentjoint is revolute 3. At the end of the link with the lengthl3 is gripper. The angle of the gripper to the base coordinateframe is denoted as , the origin of the gripper in the basecoordinate system is G0 = (x0, y0)

T .We shall assign coordinate systems to each end of the link,

the transformations between coordinate systems will thenhave the simplest form of either pure translation (prismatic

joint) or pure rotation, revolute joint. Starting with the basecoordinate system 0, rotating by 1, translation by l1, rotationby 2, translation by d2, rotation by 3, and translation byl3. The individual transformation matrices will look like:

A01 =

cos(1) sin(1) 0sin(1) cos(1) 00 0 1

,

A12 =

1 0 l10 1 00 0 1

,A23 =

cos(2) sin(2) 0sin(2) cos(2) 00 0 1

,A34 =

1 0 d20 1 00 0 1

,A45 =

cos(2) sin(2) 0sin(2) cos(2) 00 0 1

,A56 =

1 0 l30 1 00 0 1

,x0 = A01A

12A

23A

34A

45A

56x6.


Geometry Intermezzo: Relative Position of Two Straight Lines andTransversal

Transversal of two lines is the shortest line connecting the point on one line with thepoint on the other line.

A=B AB

A=BA B

Relative position of two straight lines in space could be:

coincident lines, both end points of degenerate transversal could be placed to anypoint on the line,

parallel lines, transversal could be placed anywhere along the parallel lines,

crossing lines, degenerate transversal is located in the intersection of lines,

nonparallel and nonintersecting lines, the transversal is perpendicular to bothlines.

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Open kinematic chain modeling in space the Denavit-Hartenbergnotation

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Unique and efficient description of transformations can befound by the method Denavit-Hartenberg. The description isthen called Denavit-Hartenberg notation.

Eulers theorem about the existence of motion axis saysroughly that each motion in 3D space could be represented ascomposition of rotation around certain axis and translationaround the same axis. This theorem allows to formulate thealgorithm for forward kinematics of open kinematic chains.DH notation is just one of those fomalisms. DH notation isbased on the idea of mathematical induction. Therefore DHnotation could be used only for open kinematic chains(think about why).

We describe the joint i.

1. Find the axes of rotation of joints i 1, i, i+ 1.2. Find the common normal of joint axes i 1 and i and

axes of joints i a i+ 1.

3. Find points Oi1, Hi, Oi.

4. Axis zi shall be placed into the axis of the joint i+ 1.

5. Axis xi shall be placed into the common normal HiOi.

6. Axis yi forms with the other axes right-hand coordinatesystem.

7. Name the distance of points |Oi1, Hi| = di.8. Name the distance of points |Hi, Oi| = ai.9. Name the angle between common normals i.

10. Name the angle between axes i, i+ 1 i.

11. The origin of a base coordinate system Oo can be placedanywhere on the joint axis and axis x0 can be orientedarbitrarily. For example to get d1 = 0.

12. The origin On of the end effector coordinate system andorientation of the axis zn can be placed arbitrarily whenother rules hold.

13. When the axis of two consecutive joints are parallel, thecommon normal position can be placed arbitrarily, e.g.to get di = 0.

14. The position of joint axis can be arbitrarily chosen forprismatic joints.


Adjacent coordinate frames in DH

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Position of end effector in base coordinate system

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Base and end effector coordinate frames in DH

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


5-R-1-P manipulator

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

The tranformation in the joint is uniquelly described byfour parameters i, di, ai, i. Parameters ai, i are constant,one of the parameters di, i is changing when the joint moves.

The joints are usually:

Revolute, then di is constant and i is changing, Prismatic, then i is constant and di is changing.The matrix of transformation A can be calculated

Ai1i = Ai1int A

inti ,

where

Ai1int =

cos i sin i 0 0sin i cos i 0 0

0 0 1 di0 0 0 1

,

Ainti =

1 0 0 ai0 cosi sini 00 sini cosi 00 0 0 1

.

It can be shown, that:

Ai1i =

cos i sin i cosi sin i sini ai cos isin i cos i cosi cos i sini ai sin i0 sini cosi di0 0 0 1

.

Denote qi the parameter i, di, which is changing. Expres-sion ?? can be redrawn to

x0 = A01(q1)A12(q2)A

23(q3)A

34(q4) . . .A

n1n (qn)x

n.

For each value of the vector q = (q1, q2, q3, q4, . . . qn) Q =Rn we can calculate coordinates of point P in base coordinatesystem from given P coordinates in end effector coordinatesystem and vice versa.

Kinematic equation are always solvable analytically foropen kinematic chain.


DenavitHartenberg

Matrix representing transformation from one link to the succesive link

Ai1i =

cos i sin i cosi sin i sini ai cos isin i cos i cosi cos i sini ai sin i0 sini cosi di0 0 0 1

.1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Inverse kinematics of 5-R-1-P manipulator

...

...

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Inverse kinematics

We call inverse kinematics the task when there is give a ma-trix

T(q) = A01(q1)A12(q2)A

23(q3)A

34(q4) . . .A

n1n (qn). (3)

and we are looking for values of vector q. The system ofnonlinear equations (usually trigonometric) is typically notpossible to solve analytically.

Solving inverse kinematics:

Analytically, if possible. There is not a unique descrip-tion, how to solve the system.

Numerically. Look up table, precalculated for the working spaceW Q.

There are a manipulator structures, which can be solved ana-lytically. We call them solvable.

The sufficient condition of solvability is e.g. when the 6DOF robot has three consecutive revolute joint with axes in-tersecting in one point.

The other property of inverse kinematics is ambiquity ofsolution in singular points. There is often the subspace Qs ofthe space Q, which gives the same T.

q Qs : T(q) = TTo decide which q solving 3 to select has to be taken into

account mainly:

1. Is the selected value q applicable, i.e. can the robotbe sent to q?

2. How to reach the singular point. The function q shallallways be a continuous function of time. The precee-ding values of q should make with the selected valuecontinous function of time.

3. How to continue from the singular point? The future va-lues of q should form with the selected value continousfunction of time.

4. Will not the selected value of q guide us to the situationwhere we will not be able to satisfy above conditions?

5. Will the required operational space limit us duringmanipulation? The example is the insertion of the seatinto car.

We design sometimes redundant robots (with more de-grees of freedom, e.g. 8), to increase the space Qs, from whichwe select q to allow more freedom to fulfill above require-ments.

Think about following:

Is it possible to design the robot with prismatic jointsonly which can position arbitrarily the rigid body in 3Dspace? Why?

Choose some manipulating task and design thestructure of redundant robot for it.


Inverse Kinematics - example

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25



Robot arm

Outside dimensions Operating range diagram 2-13

2.4 Outside dimensions Operating range diagram(1) RV-6S/6SC

Fig.2-4 Outside dimensions : RV-6S/6SC

115

96

122

204

160

102.5

205

115 140

4-9 installation hole 2-6 holes (prepared holes for 8 positioning pins)

View D bottom view drawing : Detail of installation dimension

6.3a (Installation)

6.3a

(In

stal

latio

n)

31.5

45

20H7+0.0210 depth 7.5

408 -0.0390 depth 7.5

5H7 +0.0120 depth 5

View A: Detail of mechanical interface

4-M5 screw, depth 9

110

R211

51.5

58.5

70

77 84

85

5461

7873

200

80

37

32

34020165

50(for customer use)

View C: Detail of screw holes for fixing wiring hookup

Screw holes for fixing wiring hookup (M4)

162 165

204

A

B

200115 140

158

158

R98

80.5

100

280

350

120

90

53

20

85 85315

140

90

63

200

* Dimensions when installing a solenoid valve (optional)

*C

(Maintenance space)

Machine cable

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25



2-14 Outside dimensions Operating range diagram

Robot arm

Fig.2-5 Operating range diagram : RV-6S/6SC

Restriction on wide angle in the rear sectionNote1) J232 -200 degree when -45 degree J2 15 degree. Note2) J23 8 degree when J1 75 degree, 2 -45 degree. Note3) J23 -40 degree when J1 75 degree, 2 -45 degree. Restriction on wide angle in the front sectionNote4) J3 -40 degree when -105 degree J1 95 degree, J2 123 degree. Note5) J2 110 degree when J1 -105 degree, J1 -95 degree. However, J2 - J3 150 degree when 85 degree J2 110 degree.

170

170

170

R258

R526

R202

R696

170

P-point path: Reverse range (alternate long and short dash line)

P-point path: Entire range (solid line)

P

R611

100

961

280

350

179

85 315308 238

85

R280

R280

R173

135

R33192

R331

76

17

R287

258437

444

421

294

474

Flange downward limit line(dotted line)

Restriction on wide angle in the rear section Note1)

Areas as restricted by Note1) and Note3) within the operating range



Restriction on wide angle in the front section Note5)


594

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25



2-14 Outside dimensions Operating range diagram

Robot arm

Fig.2-5 Operating range diagram : RV-6S/6SC

Restriction on wide angle in the rear sectionNote1) J232 -200 degree when -45 degree J2 15 degree. Note2) J23 8 degree when J1 75 degree, 2 -45 degree. Note3) J23 -40 degree when J1 75 degree, 2 -45 degree. Restriction on wide angle in the front sectionNote4) J3 -40 degree when -105 degree J1 95 degree, J2 123 degree. Note5) J2 110 degree when J1 -105 degree, J1 -95 degree. However, J2 - J3 150 degree when 85 degree J2 110 degree.

170

170

170

R258

R526

R202

R696

170

P-point path: Reverse range (alternate long and short dash line)

P-point path: Entire range (solid line)

P

R611

100

961

280

350

179

85 315308 238

85

R280

R280

R173

135

R33192

R331

76

17

R287

258437

444

421

294

474

Flange downward limit line(dotted line)


Areas as restricted by Note1) and Note3) within the operating range





594

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Multiple configurations in inverse kinematics

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Inverse Kinematics - Multiple Configurations

6Appendix

Configuration flag Appendix-63

6 AppendixAppendix 1 Configuration flag

The configuration flag indicates the robot posture.For the 6-axis type robot, the robot hand end is saved with the position data configured of X, Y, Z, A, B and C. However, even with the same position data, there are several postures that the robot can change to. The posture is expressed by this configuration flag, and the posture is saved with FL1 in the position constant (X, Y, Z, A, B, C) (FL1, FL2).The types of configuration flags are shown below.

(1) RIGHT/LEFTQ is center of J5 axis rotation in comparison with the plane through the J2 axis vertical to the ground. .

Fig.6-1 Configuration flag (RIGHT/LEFT)

(2) ABOVE/BELOWQ is center of J5 axis rotation in comparison with the plane through both the J3 and the J2 axis. .

Fig.6-2 Configuration flag (ABOVE/BELOW)

RIGHT LEFT

J2 axisRotation center

(Flag )

Note) "&B" is shows the binary



ABOVE

BELOW (Flag )


6Appendix

Configuration flag Appendix-63

6 AppendixAppendix 1 Configuration flag

The configuration flag indicates the robot posture.For the 6-axis type robot, the robot hand end is saved with the position data configured of X, Y, Z, A, B and C. However, even with the same position data, there are several postures that the robot can change to. The posture is expressed by this configuration flag, and the posture is saved with FL1 in the position constant (X, Y, Z, A, B, C) (FL1, FL2).The types of configuration flags are shown below.

(1) RIGHT/LEFTQ is center of J5 axis rotation in comparison with the plane through the J2 axis vertical to the ground. .

Fig.6-1 Configuration flag (RIGHT/LEFT)

(2) ABOVE/BELOWQ is center of J5 axis rotation in comparison with the plane through both the J3 and the J2 axis. .

Fig.6-2 Configuration flag (ABOVE/BELOW)

RIGHT LEFT


(Flag )




ABOVE

BELOW (Flag )


Appendix-64 Configuration flag

6Appendix

(3) NONFLIP/FLIP (6-axis robot only.)This means in which side the J6 axis is in comparison with the plane through both the J4 and the J5 axis..

Fig.6-3 Configuration flag (NONFLIP/FLIP)

J4 axis

J6 axis flange surface

FLIP

NON FILIP

(Flag )


1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Example: Multiple configurations for simple planar manipulator

l2

x

y

l1

2 solutions

solutions = singular point8

= singular surface1 (double) solution

2 solutions

Working envelope

0 solutions

solutions = singular point8

= singular surface1 (double) solution

10.5 0 0.5 1 1

0.50

0.51

200

0

200

400

600

800

1000

y

Working space of the 2DOF planar RR manipulator l1=0.5,l2=0.5

x

[de

grees

]

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

2DOF planar manipulator with 2 revolute joints have twosolutions within the circle, where it can reach. On the borderof the circle there is a single solution, where two solutions ba-sically coincide (compare 1 solution of quadratic equation),this border is singular surface, where one configuration canswitch to the other configuration. Outside of the circle thereis no solution. There is infinitely many solutions in the centerof the circle, this is another singular point of the robot.

This planar manipulator has only 2 DOF but it operatesin the 3D working space, that is e.g. x, y, and orientationof the gripper . DOF deficiency thus causes that only somepoints in the working space are reachable, that is only somecombinations of (x, y, ). The picture of the working space

shows the reachable points, green color represent first confi-guration, red color representing second configuration. The axis is a singular point, where any orientation is reachable,the boundary between green and red surface is also singular,where both solutions meet. Working space is shown here inthe interval < 0, 720o >, the spiral is actually from to.

It shall be stressed that ideally the working space shalloccupy some compact but dense region, where all orientationsof the end effector could be reached in all locations. Visuali-sation of six dimensional working space of spatial manipulatoris of course difficult.


Example: Multiple configurations for simple planar manipulator

10.5 0 0.5 1 1

0.50

0.51

200

0

200

400

600

800

1000

y


x

[de

grees

]

10.5 0 0.5 1 1

0.50

0.51

200

0

200

400

600

800

1000

y


x

[de

grees

]

l1

l2

x

y

l1

l2

x

y

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Manipulator with links of different length cannot reach near first joint.


PUMA

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25


Mapping between Joint and Working space

Joints space Working (Cartezian) space

0 1 2 3 4 5 6 70

1

2

3

4

5

6

7joint space

1

2

6 4 2 0 2 4 664

2

0

2

4

6working space

xy

3 2 1 0 1 2 3

1

0.5

0

0.5

1

1.5

2

2.5

3

3.5

4

joint space

1

2

6 4 2 0 2 4 664

2

0

2

4

6working space

x

y

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Mapping between joint space and working space is for robot with revolute joints quite nonlinear.


Direct and inverse kinematics - summary

Kinematics Structure Solutions DifficultySerial 1 Easy

Direct Hybrid 0, 1, N, DifficultParallel 0, 1, N, DifficultSerial 0, 1, N, Difficult

Inverse Hybrid 0, 1, N, DifficultParallel 1 Easy

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

Number of solutions and difficulty to solve the particularkinematics for particular robot is given by the mathematicalnature of the problem, the tranformation is described by setof nonlinear equations, which has to be solved. The equati-ons are basically polynomial in variables or their sines andcosines, goniometric functions causing the nonlinearity. The

equtions have in some cases unique solution, e.g. direct kine-matics of the open kinematic chain (serial manipulator) andare relatively easily solvable. In other cases the task is notsolvable analytically or its solution is not known. Numericalmethods are used in such cases or such structures are avoidedaltogether.


Motion in other coordinate systems

Joint coordinates Cartezian world coordinates

Cylindrical world coordinates Cartesian tool coordinates

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

21 22

23 24

25

The robot controller usually allows using a pendant inter-active control of end-effector position in various coordinatesystems:

joint coordinates (standard), cartezian coordinates in world coordinate system (al-

most standard),

cylindrical coordinate system in world coordinate sys-tem,

cartesian coordinate system in end-effector coordinatessystem,. . .


direct and inverse kinematics

Documents

position of joints

relative position

joint coordinate space

space of end

end effector coordinatesystem

world coordinate system

coordinate system tofi

vladimr smutn slide