INVERSE KINEMATICS COMPUTATION IN ROBOTICS USING CONFORMAL GEOMETRIC ALGEBRA

12 WANG Chaoqun , 2 2 WU Hongtao MIAO Qunhua

1 Nanjing Agricultural University, Nanjing, Jiangsu, 210031, China, [email protected] Nanjing Aeronautic and Astronautic University, Nanjing, Jiangsu, 210016, China, [email protected]

Keywords: Inverse Kinematics, Conformal Geometric Algebra.

Serial

Manipulator,

mathematical knowledge about trigonometry, rotation matrices, screw etc. has to be available, in our approach only some basic operations with basic geometric entities like planes, lines and sphere are needed [6].

Conformal Geometric Algebra is a kind of new geometric representation and geometric computation tool. In this paper, an approach to the inverse kinematics of a serial manipulator is presented based on the Conformal Geometric Algebra. First, the basics of the Conformal Geometric Algebra are introduced. Then, an algorithm for the inverse kinematics of a 5-dof serial manipulator is presented, which can be able to comprehend without prior knowledge of geometric algebra.
1
Introduction

Early in the development of robotics it was realized that the projective geometry is very well suitable to represent transformations. Now we can realize that another change of paradigm is lying ahead of us which is based on the so-called Conformal Geometric Algebra [1]. In robotics many mathematical approaches like vector algebra, trigonometry, homogenous coordinates, quaternions or dual quaternions are used for different applications. This is why people, working in this area, have to learn and understand diverse approaches and how to translate between them, for instance between rotation matrices and quaternions. Conformal Geometric Algebra is able to unify all of these different approaches in one mathematical system which is easy to understand [2]. The main contribution of this paper is the description of the inverse kinematics of a 5-dof serial manipulator using conformal geometric algebra. The SCORBOT-ER 4u robot arm (see figure 1) is an arm-like manipulator for educational use. It has 5 degrees of freedom which are: from base to end effector, the base rotation, the shoulder rotation, the elbow rotation, the wrist pitch and the wrist roll. The inverse kinematics is about finding a set of joint variables that give rise to a particular end effector pose. The inverse kinematics plays an important role in control of robot, directly relevant to kinematics analysis, offline programming and trajectory planning and so on. There are some traditional methods for kinematics of robot, such as matrix transformation method [3], screw theory [4], method of geometry [5]. While in these algorithms a lot of

Figure 1: SCORBOT-ER 4u Robot Arm

Algebra

The 5D Conformal Geometric Algebra is an extension of the 4D projective geometric algebra. It provides a great variety of basic geometric entities to compute with. Points, spheres and planes are easily represented as vectors (grade 1 blades) in Conformal Geometric Algebra. Transformations and motions are easily described in Conformal Geometric Algebra based on algebraic objects.
2.1

The products of Geometric Algebra

The three most often used products of Geometric Algebra are the outer, the inner and the geometric product. The notations of these products are listed in table 1. We will use the outer product mainly for the construction and intersection of geometric objects while the inner product will be used for the computation of angles and distances. The geometric product will be used mainly for the description of transformations.

Table 1: Notations of Conformal Geometric Algebra

2.3.1

Rotor

A and B outer product of A and B inner product of A and B dual of A inverse of A reverse of A
geometric product of conformal origin conformal infinity

meaning

alternative

N'B A.B, LC(A,B)

The operator (3) describes a so-called rotor.

IIA

eo eX)
2.2

cO einf

The basic geometric entities

Table 2 presents the basic geometric entities of Conformal Geometric Algebra: points, spheres, planes, circles, lines and point pairs. They have two algebraic representations: 'standard' and 'direct'. These representations are duals of each other (the superscription of '*' denotes the dualization operator). In table 2 x and n are marked bold to indicate that they represent 3D entities by linear combinations of the 3D basis vectors e], e2 and e3. (1) The {Si} represent different spheres and the {nJ represent different planes. In the direct representation the outer product '''' indicates the construction of geometric objects with the help of points {Pi} that lie on it. A sphere is for instance defined by 4 points ( Pj"P2"P3"P4 ) ' In the standard representation the meaning of the outer product is the intersection of geometric entities. For instance a line is defined by the intersection of two planes ( 7ri" 7ri ) ' Table 2: List of the conformal geometric entities
entity Point Sphere Plane Circlc Line Point Pair

bivector and rjJ is the rotation angle around this axis. Please note that L can be an arbitrary line and not only going through the origin. The rotation of a geometric object
o is performed with the help of the operation o rotated = RoR
2.3.2

(4)

Translator

In Conformal Geometric Algebra, a translation can be expressed in a multiplicative way with the help of a translator Tdefined by
-te 2 1

T= e

00

=1

- - teX)
2

(5)

where t is a vector representing the direction and length of translation. The translation of a geometric object 0 is performed with the help of the operation
o translated = To T
2.3.3

(6)

Rigid body motion

j 2 P=x+"2x eoo+eo

Standard representation

Dircct representation

2 S=p-l.r e 2
7r=n +deoo Z=Sj /\S2

00

S* = P" * = P" J[
*

/\ /\

P2

/\ /\

P3 P3

/\ /\

P4 eoo

A motion in 3D includes both a rotation and a translation. In Conformal Geometric Algebra a rigid body motion is described by one operator M, a so-called motor
M= RT

P2

Z =/\P2 /\ * L = /\ P2 /\ eX) * Pp =/\?z

(7)

L = 7rj /\ 7r2 Pp =S\ /\S2 /\S3

with R being a rotor and Tbeing a translator. A rigid body motion of an object
0

is described by ( 8)

2.3

Transformations and motions

M 0rigid _body _motion = Mo

All kind of transformations of an object 0 are done in Conformal Geometric Algebra with the help of the following geometric product
o transformed =VoV

3. The inverse kinematics of the SCORBOT-ER

(2)

with V being a so-called versor and with V as its reverse.

The following inverse kinematics algorithm has been developed using Conformal Geometric Algebra to solve the 5-dof kinematic chain for the SCORBOT-ER 4u robot arm (see figure 2). The chain consists of joints with one degree of

freedom each. The base (PI) defines the first joint and lies in the origin, rotating about the z axis. The axes of the following three joints are parallel to each other and all perpendicular to z axis. The final joint rotates about the a vector of the end effector, leading to the end-effector point (P,). Note that the joint point P4 and Ps are at the same position. The coordinates of the end-effector (Px, Pv, Pz), the orientation of the end effector ( n,o,a) and the iength of the links (1],12'/3,14) are needed to solve the inverse kinematics chain. Please refer to table 3 for a list of the input and output parameters of the algorithm.

position, so the translator TI is needed to translate P, about 14 in negative direction of a (see figure 3) .

(9)

(10)
z

p x

Figure 3: Step 1 - translating P, by 14 in negative direction of a to get P4 and P5.

Similarly, since there is only one rotation about the z-axis, joint 2 is on the z-axis direction of the base. Another translator T2 is needed to translate PI about 11 in direction of z axis (e2) (see figure 4). (11)

(12 ) Figure 2: The chain of the SCORBOT-ER 4u robot arm with the indication of the points PI to P, Table 3: Input/output of the inverse kinematics algorithm
Input var description end-effector x-value end-effector y-value end-effector z-value normal vector of end-effector orientation vector of end-effector approach vector of end-effector output var description angle of joint 1 angle of joint angle of joint angle of joint
x z

Px P" Pz
n 0 a

(}I (}2 (}3 (}4 (}5

2 3 4

Figure 4: Step 2 - translating PI by 11 in direction of z-axis to get P2.

angle of joint 5

II L2 Lj 14

I # link length of 2# link length of 3# link length of 4# link

3.1 Computation of the position of joint 4 and joint 5 in the kinematics chain (P4, P5)

By taking a closer look at the kinematic chain, one will notice, that P], P2, P3, P4, P5 define a plane 7[1, which includes the z axis. The intersection of the sphere S2 (around P2 with radius 12) and S4 (around P4 with radius 13) results in a circle Z3 P3 must be located on circle Z3 and on plane 7[1 as well, thus the intersection of Z3 andni results in a point pair again, from which P3 can be selected (see figure 5). (13 )

Since there is only one rotation about the z axis, joints point P4 and P5 are on the negative direction of the a vector of the end-effector. Note that point P4 and P5 are at the same

(14)

The plane 7[2 spanned by the x axis and the z axis. Since the y axis is perpendicular to this plane, we get (20)

(15)

The line L23 through P2 and P3 (21)

(16) (17)

The line L34 through P3 and P4 (22)

The z-axis Lz (23)

.---- a

Pt

Now, we are able to compute all the joint angles

(24)

arccos

L3 .r: * IL23 I Lz* I

(25)

(26) Figure 5: Step 3 - The intersection of the spheres around P2 with radius 12 and around P4 with radius 13 results in the gray circle. Intersecting the circle with the red plane 7[] returns P3
3.4

Compute the angles of the joints

(27)

Angles between two objects 0], 02 like two lines or two planes can be computed using the inner product of the normalized dual object:

COS

(8)

* 01'O2* 10; 1 0; I

(18)

85 is the angle between the 0 vector of end effector and the axis of the joint 4 which is parallel to that of joint 2, thus 85 is the angle between the 0 vector of end effector and the axis of the joint 2. Since the axis L2 of the joint 2 is rotate about z axis with 85 fromy axis, so (28)

or (19)

8 s
-e3

arccos

L* '0 * IL;1 0 *1

(29)

Now, all the auxiliary planes and lines that are needed for the computation of the angles of the joints are calculated (see figure 5) . We need

OJ

where R

is the rotor.

4. Conclusions

Inverse kinematics is an important topic in robotics. It does not have analytical solution for all joint configurations, and efficient computation often is difficult. Owing its properties of geometric intuitiveness, compactness, high efficiency, unification and elegance, Conformal Geometric Algebra gives a geometrically intuitive approach for solving inverse kinematics. In this paper, we demonstrated the power of Conformal Geometric Algebra to solve the inverse kinematics of a 5-dof manipulator. In this framework the computations of intersection of the entities like sphere, planes, lines and circles is easier, this help us to solve in a series of steps the problem of the inverse kinematics.
References

[1] D. Hildenbrand. "Geometric computing in computer graphics using conformal geometric algebra", Computers & Graphics, 29, pp. 795-803, (2005) . [2] D. Hildenbrand. "Geometric Computing in Computer Graphics and Robotics using Conformal Geometric Algebra ", PhD dissertation, TU Darmstadt, (2006) . [3] Craig, John 1. "Introduction to Robotics: Mechanics and Control", Addison Wesley, (1989) . [4] Huang, Z., Zhao, Y. S., Zhao, T. S. "Advanced Spatial Mechanism", Beijing: China Higher Education Press, (2006) . [5] Li, Y. H., Ye, B. S., Zhu, Z. H. "Geometric Method of Analyzing Inverse Robotic kinematics and Application", Chinese Robot, 3, ppAI-44, (1989) . [6] D. Hildenbrand, 1. Zamora, E. Bay-Corrochano. "Inverse Kinematics Computation in Computer Graphics and Robotics Using Conformal Geometric Algebra", Advanced in Applied Clifford Algebras, 18, pp.699-713, (2008) .