110 Roborics-Control and Programening Pure manslation along an axi: Pure rotation about an axis. Combined rotation and translation. we 35. ‘Translation of a Point Translation is motion of a point in space with respect to a frame by a finite distance along coor, axes. The directional unit vectors remain in the same direction and thus do not change. When is moved forwards relative to frame, it may be visualized either as a forward movement of vec, backward motion of frame. In these two cases the mathematics involved is identical. Fig, 3.4 sho... translation of a point from A to A’ in space. The reference frame is (0). The directional (rotzy Vectors of the frames at points A and A’ with respect to frame (0) remain the same. a Veg Fig. 3.4 Pure translation of a point A ‘Thus location of the point at A’ can be found by adding the translational vector (d) to the oriciz:! position vector (p} at A. Here both the frames are referred with respect to base coordinate freme.(0 In matrix form, the new point representation may be found by pre-multiplying the frame with « me: representing the transformation. Since the directional vectors do not change in translation, t: transformation [T] will be simply 100 4 o10d = (dd, ¥ 3 (T}=Trans(didsd=| gy o G9) 0001 where d,, dy and d, are the components of d along X, Y and Z axes of the frame at A (which is in the direction of base frame-(0}). Here the first three columns represent zero rotational motion (unit) matrix), while the last column represents the components of translation scaled to 1. The location of ti point A’ is therefore given by multiplying the original frame matrix representing point A with this transformation matrix (T]. Thus new point is represented as: 10 0 d,}[n, o, a, py] fn, oO, a, py td, 0 10 dyJJn, oy a, py|_|ny oy a, py td, G0 001 4,{/n, 0, a, p,| |n, 0, a, p,+d, ooo 1j[0 0 0 1} lo 0 o 1 Here the homogencous representation facilitated the multiplication of matrices conveniently.Coordinate Frames mations: Example 3.2 late Frames & Transformations 111 0.527 -0.574 0.628 |A frame represented by the matrix | 9359 9.819 0.439 -0.766 0 0.643 0 0 o ind representation of new frame. has been moved 6 units along the x- Hana axis and 4 units along the 2-axis, Fi Solution: ‘The transformation matrix is 1006 100 o14 001 ooo New frame is given by the product: 0 0 6)[ 0527 -0574 0.628 6 10 || 0369 0819 0.439 4 0 1 4/|-0766 0 0.643 6 oo1 0 0 ool 0.527 -0.574 0.628 12 0.369 0819 0439 4 -0.766 0 0.643 10 0 0 ool eooe Here the upper 3x3 sub-matrix remains unaltered. 3.5.2 Pure rotation of Vectors About an Axis Rotations of a point about each of the three axes by an angle @ can be accomplished by rotation transformations. Consider a point A within a frame having origin at point O. Initially the direction of frame at A shown by n,o and a vectors coincides with inertial reference frame X-Y-Z. ‘As the frame rotates about X-axis by an angle @, the point A attached to the frame will also be displaced, Then n-o-a frame for representation of point A no longer coincides with original frame X-Y- Zas shown in Fig. 3.5 (a). Let original coordinates of point A be (pa.Po,P,) and new coordinates after rotation as: (Px, Py» Pz)- As seen from the two-dimensional representation in Fig. 3.5(b), the value of p, does not change as the frame has rotated about the X-axis, but py and p, are different from p, and p,. This transformation is described by the following set of three equations: Pa= Pa Py=P. cos @-p, sin @ Pe=Po sin @ +p, cos 8112 Robotics-Control and Programming (a) Coordinates of point A after (b) Equivalent 2-D representation rotating about X-axis Fig. 3.5 Rotation of a point A about X-axis This can be written in matrix form as Px 10 0 Pa py} =[0 cosé sind] )p, p,} [0 sind cos@ | |p, Thus in order to get new coordinates of point A in the reference frame (X-Y-Z) the original coordinates must be pre-multiplied by a rotation matrix. This rotation matrix is valid only for a pure rotation about X-axis and is denoted by (R], or sometimes Rot(X,6). Denoting cos @ as c® and sin 8 as $B, we can write this rotation matrix as: 10 0 Rot(X,8)=[R]x =|0 c8 -s6 GAD 0 30 As a homogeneous matrix it can be written as: 10 0 0 _|o <8 -s9 0 “]0 s8 8 0 oo 01 my G2) Likewise: Rotational transformation matrix about the Y-axis can be shown as: 8 0 0 Rot(¥,6)=[R=| 0 1 0 3.13) -38 0 c@Coordinate Frames & Transformations 113 0 9 0 Its homogeneous transformation matrix form is: (Ty=] ° 1 9 9 G.14) -s8 0 8 0 ; 0001 Rotational transformation matrix about the Z-axis is obiained as: c8 -s8 0 Rot(Z,8)=[RJz=|s0 8 0 3.15) oo 1 8-30 0 0 lis homogeneous transformation matrix formis: (T]=|S° © 9 9 6.16) o 10 0 0 01 Itis possible to develop the rotation transformation about any other veetor in which X,Y,Z are not the coordinate system. Example 33 A point P=[2 3 4J" is attached to a rotating frame. The frame rotates 90° about the X-axis of the reference frame. Find the coordinates of the point relative to the reference frame after the rotation. Solut ‘The original point given can be represented as a homogeneous vector {P}=(2 3 4 1)" ‘The rotational transformation is given by substituting 6=90°- 10 0 oj f1 000 0 cos90 -sin90 0/ |o 0 -1 0 0 sin90 cos90 0/ |0 1 0 0 0 0 0 1fJlooo1 New coordinates after rotation in homogeneous form are given by: 10 0 O}(2) (2 (The| P’ _{o 0 -1 of |3|_|-4 0 I=(TIxtP}=| 9 1 0 offatq}3 ooo ly lt Hence the new coordinates of the point are [2 -4 3]" Altematively, we can use the rotational matrix directly to multiply with the three dimensional vector. Example 3.4 The coordinates of point in frame-(1} are [4 3 2". The position vector P is rotated about the Z-axis by 30°. Find the coordinates of the new point (say Q). Solution: The 30° rotation of P about the Z-axis of frame-(1} is114 Roboties-Control and Programming ‘cos30 —sin30 0} [0.866 -0.5 0 [Riz =Rot(Z,30")=| sin30 cos30 of=| 0.5 0.866 0 0 0 1 0 o 1 New point Q is given by the product [Rlz(P} Substituting [R]z and (P}, we get: 0.866 -0.5 0] (4) [1.964 {Q)=| 05 0.866 0} 37= 4.598 0 0 1} (2. 2 ¢,The new position vector is: Q=[1.964 4.598 2)" 3.5.3 Combined Rotation and Translation of vectors Combined transformations consist of a number of successive translations and rotations about the fixed reference frame or any current frame. Such complex transformation can be resolved into a set of translations and rotations in a particular order. This order is very important and if the order of wo successive transformations is changed, the result may be completely different. For example, one may rotate the frame about X-axis and then rotate about Y-axis in order to reach the final target point. The final frame would be entirely different if the order is changed, Consider a frame subjected to following three successive transformations relative to the reference frame (X-Y-Z): () —_Rotation of an angle 6 degrees about the X-axis (i) Followed by a translation of [a, , ¢]" (iii) Followed by rotation of angle @ degrees about the Z-axis. A point A in the X-Y-Z reference frame is attached with another frame. As the frame rotates or translates relative to the reference frame the point A within the frame moves as well and the coordinates of the point relative to the reference frame change. After first transformation (rotation about X-axis), the coordinates of point A relative to reference frame can be calculated by: 10 0 0 0 ob -s6 0 Ay) = a ; {AD =l9 go cp 0 AITUIA Gin 0001 At the end of the second transformation (translation), the coordinates of the point relative to the original reference frame will be 100. 010b {Arley g yo {Arb =Utal (Ad=(hNTItA) ooo! And after the third transformation (rotation about Z-axis), the coordinates of the point relative f° the reference frame will be {A)=(M{AD 3.18)a, Coordinate Frames & Transformations 1S cg -sp 0 0 so ep 00 (Asl=]q go M2) SUT Ad =U TIT NAL Cs PETA} B19 0 0 O18 Hence the revised veetor (As) at the conclusion of a set of transformations is found by pre- atpying the original vector (A) by product of transformation matrices {Ty[ Ty] Tuk. Ht should be id here that all cansformations are refered with common X-Y-Z, cooainate syst, The onder of totges cannot be changed. It should be noted thatthe order of matrices is writen in the apposite ero transformations to be performed, This is also called composite homogenous transforma gxample3S point P=I7 3 2)" is attached to a frame (n, 0,3) and is subjected tothe ansformations deseribed below: (1) Rotation of 90° about the Z-axis. (2) Followed by a rotation of 90° about the Y-axis. 3) Followed by a translation of (4, -3, 7] Find the coordinates of the point relative to the reference frame at the conclusion of the transformation. Solution: ‘The matrix equation representing the transformations is (Axe) =(TTIT (Acca) 1 0 0 4 ]f-cos90 0 sin90 0][cos90 -sin90 0 0) [7 _Jo 1 0-3] 0 1 0 6} |sin90 cos99 0 0} /3 “Jo 0 1 7 ||-sin90 0 cos90 O}| 0 o 1 O})2 ooo rfl 0 0 o I)f 0 o ott 100 4)][0 01 OJfo -1 0 0} {7) [6 _{o 1 0 -3/]0 1 0 o}f1 0 0 of)a{_ja =loo 1 7] J-1 0 0 ofjo o 1 of jaf iol oo0 1flo oo alo o oly lt ‘Thus coordinates of the point at conclusion of transformations are (6 4 10)" Example 3.6 Frame-(B) is rotated with respect to frame-(A) about the X-axis by an angle of 30°. The position of the origin of frame-{B) as seen from frame-{A) is (po)=15 7 7)". Obtain the transform: i {Tp} which describes frame-(B) relative to frame-(A). Also, find the description of point P in the frame-(A} if position its position referred to frame-(B} is ("p) = (6. 4, 21". Solution: The homogeneous transform matrix describing frame-(B) with respect to frame-(A) is tye] ORal (Ps! (ooo 1 Frame-(B} is rotated to frame-(A} about X-axis by 30°.116 — Roborics-Control and Programming 1 0 0 1 0 0 (Rel =Rot(X,30)=|0 cos30 -sin30] =|0 0.866 -0.5 0 sin30 cos30} [0 05 0.866. As position vector given is (“pa}=[5 7 7)", we can write the homogeneous transformation matrix ay 1 0 0 Ptpja]2 0865-05 7 0 05 0.866 7 oo 0 1 Given the point P referred to frame-[B) as ("p}= (6 4 2)" Point P referred to frame-(A) is given by Ts) {°p} 1 0 0 S](6) f _|0 0866 -05 7) J4|_ | 9.464 (or 0 05 0.866 7} }2{~ )10.732 0 0 o ahi 1 ‘What happens if you are given {“p] and asked to compute (p}? Would you attempt to invert the entire matrix (“T,]? Here the rules of first 3x3 rotational sub-matrix are useful. That is [RY'={RJ. The first 3 elements of fourth column in matrix (*Tp] need to be computed. Example 3.7 Frames-(A) and (B) have coincident origin and differ only in the orientation. Frame-(B) is initially coincident with frame-(A). Certain rotations are carried out about the axis of the fi First rotation about X-axis by 60° then about same X-axis by 45° Obtain the equivalent rotation matrix “Rp. Solution: Rotations are in order X-X-Y about the fixed axes, hence, itis a case of fixed angle representation. Therefore the resultant rotation matrix is: (*’Ro]=Rot(¥,30°) Rot(X,45°) Rot(X,60°) cos30 0 sin30][1 0 o yfi oo 0 © 1 0 JJ0 cosas -sin4s|}0 cos60 —sin60 ~sin30 0 cos30}|0 sin45 cos45 }]0 sin60 cos60 0.866 0.482 -0.129 0 -0258 -0.96 -05 0835 -0.223 xed frame-(A): and finally about Y-axis by 30° Example 3.8 A moving coordinate frame {U,V,W} undergoes a screw transformation, that is it is translated by ¢ units along Z-axis and rotated by an angle of 180° about same axis of stationary reference coordina! frame (X,Y,2Z}. Assuming the frames (U.V,W] and {X,Y,Z} are initially coincident, obtain t® homogeneous transformation matrix for the screw transformation,Coordinate Frames & Transformations 117 lution: ‘The overall transformation matrix forthe given situation is M=(TT) cos180 sin180 00 1000 sin180 cos180 0 0 0 (mi| jo 100 where 0 0 1 of™IM=} Gg la 0 o o1 ooo0 -1 000 0-100 Th lo ota 0001 Itcan be noticed that even if the order of transformations is reversed the overall screw transformation matrix will not change. This is an important property of screw transformations. 35.4. Inverse of Homogeneous Transformation matrix During the kinematic modeling of robots, inverse transformations are often needed. Consider a manipulator linkage holding a tool at the end of its hand. It has to be moved towards a part P as shown in Fig. 3.6. Zp ’ Part frame (P} Ye ig. 3.6 Manipulator with important frames A reference frame (0) is selected at the base of the linkage. The end-effector (hand) is described by the frame-(H}, while the end of the tool is described by frame-(E}. Thus the tool frame (H} can be telated to base reference frame {0} in two independent ways: i) Through the part P and ) Through the robot.118 Robotics-Control and Programming When referred the frame (E) through the robot, we write the ransfo (Tel = CT Te) On the other hand, when the frame (E} is dseribed with reference (0 Part frame (P}, we get: (Te) =CTr] (Te) G21) Here [!'T,] is the transformation of the tool end frame (E} relative to the robot's hand frame (14) This is known in practice. Similarly the transformation (°Tr], representing the relationship between the part frame and base reference frame (0) is also known using a vision system. The transformation of part frame relative to frame {E} described by the matrix [Te] is also commonly available. So the ‘unknown matrix (°T], can be obtained by equating above two matrix multiplications, mation as: G.2n) Thos (Ts) ('Te}= (THI Te) 6.23) Post-multiplying this equation with [Te], we get: (°F) Te) Te = Te) Te Te" (3.23) This gives (Tu) as follows: (3.24) (Td = (Te) Te) Tu Here the inverse transformation r it is convenient to decompose the matrix it orthogonality property of rotation matrix, transpose, while the inverse of upper right product defined below: stead of evaluating inverse of the matrix as a whole, ro rotational and translational sub-matrices. Using the the inverse of upper left 3x3 matrix is written as its 3x1 position vector is computed as a negative of the dot is needed. Ir 1 0 0 Consider a rotation matrix Rot(X,0)=[R}x=/0 8 $0], for which inverse has to be obtained, 0 -s@ c@ To do this, the following steps are followed: (@) calculate the determinant of the matrix (b) Transpose the matrix (c) replace each element of the (d) Divide the converted matrix by the determinant. determinant is unity and finally it can be shown that the inverse of transposed matrix by its own minor ‘This matrix is thus a unitary matrix. It is true for all For this rotation matrix [R}x, the rotation matrix is same as its transpose. rotation matrices. For homogeneous 4x4 transformation matrix, the matrix inverse can be written by dividing matrix into two portions: The rotation portion of the matrix can be simply transposed, while the transtational part (upper right 3x1) portion of the homogeneous matrix is computed as the nest of the dot product of vector {p) with each of the {n}, (o} and (a) vectors. The last row is not affected The lengthy process of computation of inverse can be brought down with this approach ‘Mathematically, let [T] be the homogeneous transformation matrix defined by My 0, ay Px mme|ty 9 8y Ps nm, 0, a Pz 0001 (in vector form)_| RI “[(0 0 0} jts inverse is obtained as: hen sam y ryt =|°* a, a 0 0 _| (Rt “lio 0 0} i 9 Coordinate Frames & Transformations 119 al (in matrix form) a, (xn, +pyny +p,n,) Ox 7x04 +40, +7202) a, ~(P.a, +Pyay +p,a,) ° 1 af 1 (in vector form) (in matrix form) Here (Pax +PyMy *P2M2) =(P).(n), (Pex *PyOy + P22) =(PI-(0} and (pxdx tPyay +P242) =(P}-(a} are the vector dot products defined by the vectors an tae talanciamy Janek, (0)=0, i +0y j+0,k, {a}=aci tay j+a,k and (p)=pxi +p, j+p.k