This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

4, 477-496

**A 4-WHEEL SKID-STEERING MOBILE ROBOT1
**

Krzysztof KOZLOWSKI*, Dariusz PAZDERSKI*

Systems Engineering Poznan University of Technology ul. Piotrowo 3A, 60-965 Poznan, Poland e-mail: Krzysztof.Kozlowski, Dariusz.Pazd

amcs

**K. Kozlowski and D. Pazderski
**

Yg

1

1 II .

Z

?comJ 6

xi

Xg

e found in (Morin and Samson, 2002).

g the Lyapunov technique. It is important to note that, in contrast to the previous work done by (Caracciolo et al., 1999; Kozlowski and P

transformation of the SSMR kinematics into a new space is presented, too. Next, the control scheme on the dynamic level is obtained t Fig. 2. SSMR in the inertial frame.

g. 1). Additionally, it offers a possibility to develop a controller based on the backstepping technique. f COM in the inertial frame can be writ-ten as COM = (X, Y, Z). Since in this paper the plane motion is considered only, the Z-coordinate

X Y 6 ]T is the state vector describing generalized coordinates of the robot (i.e., the COM position, X and Y , and the orienta-tion 6 of th

Fig. 1. Decomposition of an electrically driven mobile robot.

ic Model

e kinematic model of an SSMR, it is as-sumed that the robot is placed on a plane surface with the

' y //> \ul0 W %^y.= u.denotes the Euclidean norm.c. the following relat ( 3) . In con?trast to most wheeled vehicles.d xc . is generally nonzero. where a. This property comes from the mech 4. After combining Eqns. (4) and (7). ciolo etal. the lateral velocity iy ICR = (xicr. . based on the geometry of iu. Wheel velocities. d i satisfy the following relationshi (7) = = d cx a.Modeiing and control o f a 4-wheei skid-steering mobile robot yl v2 V 2 icr ^ d i ?. H Defining the coordinates of the ICR in the local frame as of the SMRR. b and c are positive kinematic parameters of the robot depicted in Fig. v .. vi|| where the symbol H .sin 9 cos 9 vx vy Fig.d yc ) (5) allows us to rewrite (4) as follow be developed: = y icR d1 y vy . XICR (6) = d2y From Fig.v ' x . 3. 5 it is clear that the coordinates ofcy + c. Consequently. 5.= M d oy or.ii l dil l l l lv l l | | = --. in a more detailed form. d2 x (2) = = d cx + b. (1) Xg are defined with respect to the local frame from the instan-taneous center of rotation (ICR). ii iu. 1999): X Y cos 9 sin 9 . y icR) = (. = d vectors d3 y = = d cy ..

Additionally. (12) tegrable.x ICR cos 6 (16) S (q) = 0 0 +b w . this is respectively. velocities of the left and right wheels. resp ( 4) . v F and v B are the lateral coordinates of the velocities of(q) n. we can write -xicr 0 W-xicr . It is in-teresting to see that t ollowing approximated relations between the angular wheel velocities and the ve-locities of the robot can be developed: nowledge of the x i-axis projection of the ICR.sin 6 cos 6 x ICR X Y 6 = A (q) q = 0. respectively. (1) has been used. 1999).amcs K. In consequence. Kozlowski and D.. The active force Fi and reactive force Ni are related to the wheel torque and gravity. Since the generalized ve?locity vl q is always in the null space of q = S the front es of the left and right wheel velocities. In addition. '1 (14) -c ing transformation describing the relationship between the wheel velocities and the velocity of the robot: vr where and S T (q) A T (q) = 0 (15) 1 c vx vf cos 6 (9) x ICR sin 6 sin 6 1 . a nonholonomic sys?tem because of the constraint described by (12). the wheel forces Itis Wr w 2c (11) mostly depends on the longitudinal slip and can be valid only if this phe-nomenon is not dominant. First. assuming that the effective radius is ri = r for each wheel. and rear wheels. considering the linear velocity v x and the angular velocity as control signa w W l + Wr 2 vx r (11) n = e for such vehicles (see Caracciolo etal. it de-scribes a nonholonomic constraint which can be rewritten in the Pfaffian form: r ir it . Pazderski where Eqn. vb ance with (2) and (8). There-fore.= 1 L a Wr J vl [ v r (10) r is underactuated.Wl + caused by unknown lateral skidding ground interaction forces. the parameters r an vy + xicr 6 = 0.

Based on the previous deliberations. Next. 7. which allows us to neglect the term /z v a to sim-plify the model. namely.Modeling and control o f a 4-wheel skid-steering mobile robot page. we obtain b L (q. the relation yU. Since there is symmetry along the longitudi?nal midline. where ks ^ 1 is a constant which determines the approx-imation accuracy according to the 2 that Fi is linearly dependent on the wheel control input Ti. Fi = 1 . the following equ 2 (a + b) (26) However.arctan (ksa) = sgn (x). 4 i=1 (18) nal forces. properties. we obtain the followin Fii = fJ-ici 'mg s gn (23) (24) Ni a = N2b. (27) (20) . (21) Fig. respectively. (19) 2 (a + b) a Considering mg. It is very important to note that th sgn (a) 2 = . q ) = T (q. F f (a) = N sgn (a) + /Zv a. clude the nonholonomic constraint (12).arctan ( k s a ) . Th he robot and I is the mo?ment of inertia of the robot about the COM. N2 = N3 = the kinetic energy of the vehicle and neglect-ing the energy of rotating wheels. it is important to note that friction modeling is quite complicated since it is highly nonlinear and depends on many variables. (17) lim . it is assumed that the mass distribution is homogene 1 T = 2 m (X2 + Y2) + 2 162. 7) and neglecting addi-tional dynamic(vyi). Active and resistive forces of the vehicle. q) (25) N1 = N4 = mg. For simplicity. it is assumed that the potential energ mass and g is the gravity acceleration. Considering the four wheels of the vehicle (Fig.cN ^ |/z v a| is valid. the friction forces for one wheel can be written as ce to the wheel. the dynamic equation of the robot can be obtained. N4a = N3b.

a J2 F ii (v yi) + b J2 F ii (v yi) i=1. (28) cos 6 Ti (38) F= 1 r i=1 4 sin 6 i=1 Ti M = 0 m 0 0 (29) c (.3 B(q)= 1 cos 6 cos 6 sin 6 r c c sin 6 (41) Next. which results in F x =cos ^ F i. (30) where li=1 and r denote the torques produced by the wheels on the left and right sides of the vehicle. respec-tivel i=1 t t F r y (q) = sin 6 ^ F si (v xi )+cos 6 ^ F ii (v yi) . (33) and (40). Pazderski g the partial derivative of kinetic energy and its time-derivative. 7.T 1 . the following dynamic model is obtained: F si (v i=3. (36) mic equations become In consequence. a new torque control input t is defined as (39) TR T3 inertial frame can be calculated: ssipation of energy are considered. According to Fig.2 t. the vector F of active forces has the fol-lowing form: Fx Fy M (37) F= . Therefore a constraint has to be imposed on (4 ces generated by the actua-tors which make the robot move can be expressed in the inertial frame as follows: (q) A. using (28). (42) To define generalized resistive forces. (43) suitable to express (43) in terms of the internal velocity vector n.4 i=2. the follow-ing resultant forces expressed in the + T4 F r x (q) = cos 6 4 4 t= TL = T1 + T2 X F si (v xi) .T 2 + T 3 + T 4) To simplify the notation.sin 6 ^ F ii (v yi) . (34) (35) The active torque around the COM is calculated as M = c (-F 1 . (34)-(36) and assuming that ) mX 16 m 0 Mq. (31) where B is the inpu i=1 i=1 The resistant moment around the center of mass Mr can be obtained as Mr (q) = .4 +c F si (v xi )^X (32) M (q) q + R (q) = B (q) i=1. the vector R ( q ) = [ Fr x (q) Fr y ( q ) Mr (q (33) the dynamics of a free body only and does not include the nonholonomic constraint (13).amcs as dt V dq where K. Kozlowski and D. Therefore. the inertial forces can be obtained the radius of each wheel is the same Using (17). i=1 4 F y = sin 6 F i.F 2 + F 3 + F 4 ). (43) is multiplied from the left by S T (q).

8. Conside and uvaR u uvaL u va1 + u va2 va3 u UL = 1 U mi = 1 Um2 (57) Assuming that all motors and gears have the same param-eters and using (51)-(57). Therefore. ?va = 2L a . 8 a simplified scheme of the drive on the right side of the robot is de-picted.Modeling and control o f a 4-wheel skid-steering mobile robot where C = S T M S = m x ICR m0 M = S T MS = 0 the 0 of 9 rotor. while u m denotes the angular velocity i . we can write i k i i ai . Second. (5 ) 2 he SSMR is not fixed. respectively. k e is the electromotive force coefficient. According to Fig. it can move L e and resis-tance of the rotors.ia + 2R a i a + 2 kenLOw . Since each pair of motors c c (56) + va4 r DC brushed motors with mechanical gears. theoretically. respectively. a new input (47) (48) (49) (50) F rx (<?) ? = ST ? = X lCR F ry (q) + M r J' uvaL u B = ST B = 1 1 vaR (55) where u vaL and u vaR denote the motor voltage signals on the left and right sides of the vehicle. Since the i-th actuator is equipped with gears characterized by a ratio n > 1. 1971) is introduced to limit the magnitude oflateral slippage in open loop con?trol. the ide he voltage equation of the armature can be approximated by the fol-lowing linear relationship: uiva a ~jT i ai + R a i ai + k e u mi . (51) Tmi -holonomic constraint (see Gutowski. (53) (54) m x2 CR + 1 J' where some additional dynamic effects concerning inertia and backlash of gears have been neglected. -9 Xicr Ti n k i i ai . In Fig. the following voltage-curre uvoR T - 2k j n i a . (58) (59) dt orque Tm produced by the i-th motor is linearly dependent on the rotor current i ai .

qr (t). . It is interesting to see that sele 9 0 (q r ) = . (68) is transformed into a form which is similar to the nonholonomic integrator considered in = const. 5 and Eqn. for the trajec-tory tracking problem. nr (t). This constraint can be written a The control objective is formulated as a practical stabiliza-tion of the position and the orientation of the robot consid-ered in lim ||q (t)||< e.. of the robot wheelbase.qr (67) Using (14) and (64). z = u. Calculating a time derivative o con-straint on the kinematic and dynamic model of the vehi-cle. (69) where vx and w are control inputs. 2001). xICR in (16). qr (t) is a vector function which depends explicitly on time. In the second case.3.qr = X Y iT (63) 92 x0 sin 6 . However.amcs K. Pazderski to be bounded for all times: (65) nr (t).S (q r ) n r . . g 0 is a drift of the system.e. (47)-(50) is replaced by x0 and it is assumed th q= q . 3. Kozlowski and D. 91 ( q ) = cos 6 sin 6 0 (70) (71) q = q . To this end. we reformulate (68) as follows: Q= 9 0 + 9 1 (q) vx + 9 2 (q) w. t. Kinematic Transformation (61) xo G (-a. (9) that if the xICR coordinate goes out. b) . we get S(q)= cos 6 x0 sin 6 sin 6 -x0 cos 6 0 1 (62) Q= S (q) n . g 1 (q) and g 2 (q) are vector which allows us to approximate the relation between the angular and linear lateral velocities of the robot.S (q r ) n r.x (q) = 0 cos 6 (72) 1 aused by the refer-ence kinematics signals and for n r = 0 (i. when the regulation problem is considered) the system has no drift. (60) where x0 denotes a fixed coordinate of the ICR which should be selected as ere e denotes an arbitrary small positive constant. q = u T J T z + f. qr (t) G L?. his section we introduce a kinematic transformation which will be used later on in designing a control law.>oo (66) vy + xo 6 = 0. then the robot skids into a lateral direction and loses its stab sed to include an operational nonholonomic constraint (mathematical constraint) which is based on (12). (73) (74) . Fur troller based on (Dixon et al. (68) Taking into account (62).

Then (73) can be rewritten as + 29 (Xcos9 + Ysin9). i. z2 = Xcos 9 + Ysin 9. (77) + ^9?(t) X(t)cos 9 (t)+Y(t)sin 9 (t) dt.Ycos 9 - (88) . (80) and (81) in (77).Ycos9) To calculate the integral in (84). the variable w has the fol-lowing form: w = -9 Xcos9 + Ysin9 ) x 9) .Y cos 9 = Then (85) can be rewritten as d dt X sin9 . (86) x q 9J= 9 X cos 9 + Y sin 9 = 9 X cos 9 + Y sin 9 (87) Finally. we see (note that it is just Eqn. tion matrix P (9. according to (84). 1995). the following change of coordinates is proposed: Z1 = 9. q + 2 Xsin 9 . (13) X sin 9 . 1993). (84) ed on (De Luca and Oriolo. (83) Therefore the variable w can be expressed as w = -9 (Xcos9 + Ysin9^ W = Z2 M1 . it is convenient to con?sider the expression z1 is the orientation error and |z2| denotes the length of projection of the vector [ X Y ]T onto the direction determined by the actual orien = 9 (Xcos 9 + using the error kinematics given by (68) and assuming that vrx = 0 and ur = 0. using (74).9). it follows that and (74) are first-order systems (see Murray and Sas-try. u is an input signal and f denotes the drift. It is possible to find the following global dif-feomorfism that preserves t W = 9 Xcos9 + ysin9 X cos 9 + Y sin 9 (82) Z = W Zl Z2 = P (76) or d dt \ e x p n d t w 0 . q r (t) = const.Modeling and control o f a 4-wheel skid-steering mobile robot Taking the time derivative of (78) and (79). (78) (79) dt (Xsin9 .Ycos9 - x q 9. we can auxiliary vari-ables. the regulation problem is considered first.e.J (Xcos 9 + Ysin 9) interpreted as a new aux-iliary state vector of the system (68) described in a three-dimensional space. while J is the following skew-symmetric matrix: write Z1 = 9.. J = 0 -1 1 0 (75) d (X cos 9 + Y sin 9^ Z2 = dt (80) (81) Next.Zi?2.

Q) (101) 0 0 1 cos 6 sin 6 1 0 0 ( d = k2 + ^ + w ( -2 .x0 wr cos Z1 ) u = u 0 . 2001) which solves a unified tracking and In this Z2 . Kozlowski and D.Y cos 6^ . n = Tu + n. (99) Z1 M2 + 2 (Z2 Wr + x0 wr . can be= ^Z d + ( k1 W+f + W??1) Jz d. u = T -1 n - while k1.x Wr trajectory tracking problem when the reference ve rx 3.Z1 Z2 + 2 (97) 6z2 is an auxiliary velocity vector which^x0 6 .. Based on (94).vrx sin Z1 . (90) results in k 1 w+f ua = M1 Z2 - JZ d + ^1 Z d .6z2 . (101) One can easily check that this matrix is nonsingular. Kinematic Control Law - 2x0 6? + 2x0 wr = 6Z2 + 6r section raZ2 + 26r Z2control structure based on (Dixon et al. k2. which introduces a time-varying feedback to overcome Brockett's obstruction (see Brocke + 2 ( x0wr .2 cos 6 -2x0 (d ( k1W + f 0 7 l .Z1 M2 From (99) it is clear that the signal u a is modulated by an auxiliary signal z d generated by the tunable oscillator de-scribed by the + 2 [-vrx sin Z1 + Wr (x0 +Z2 .x0 sin Z1 + l) wr Using (68).v sin Z1 for0the cos + 2 is nonzero only . Pazderski where ars in (73). which transforms (68) into (73)-(74).an auxiliary variable that has the meaning of a distance.vrx sin Z1 . (92) (100) Summarizing. (96) (89) n= vrx cos Z1 + Wr (. a0. we have w = . (78) and (79) in (89).amcs K.x0 wr sin Z1 d. the trajectory tracking case is considered.x0 Wr cos Z1) = M1 Z2 .Y cos 6 is X cos6 -Fsin6 sin6 . (98) = Z1 Z2 . O1 and (d are auxiliary terms defined as 6 cos 6+2 sin 6 -6 sin 6. the velocity + 2(9 Xcos 6 + Ysin 6 where + 2 (X sin 6 .x0wr cos Z1 ) . the matrix P in (76). P .Y cos6 . (94) It is easy to show that the envelope of ||z d (t)|| is determined by the scalar function ( vrx cos Z1 .Z1 Z2 The control law has the following form: + 2 ( x0 wr - vrx sin z1 . d2 (93) (5d = ?0 The velocity transformation can be obtained from (74) and (80)-(81) as follows: exp(-?1 t ) + e1.26 kinematic . we conclude that 0 1 1 l 1 or T = (95) T -1 = 1 -l 10 Xcos 6 + Ysin 6 w= + and l = XXsin 6 .x0 cos Z1)]. Wr . a1 and e1 are positive constant design parameters. .4.vrx sin Z1 .x0 wr cos z1 ) + 2 ( x0 wr . (90) Next.2x0 Q.Z1 ZZ2 + 2Z2 Wr where u a . (91) and therefore f = 2 [ . written as where z d = [ Zd1 Zd2 ]T .vrx sinZ1 + wr (x0 + Z2 - x0 cosZ1) ] . To this end. substituting (74) into Eqn..k2 Z. Taking again the time derivative of (88).

. z) 1 m2 (w. z) = . denoted by u d ( t) . is design ud = u a k 2 z. (109) a z d (t)T z d (t) = z d (0)T z d (0) = J2 (t). R = T T C n+M n+R . The proposed Lyapunov function candidate has the following form: M U + C u + R = B 1 2 1 ~T -w2 + -z T z. 2 2 - t. l (105) n. the envelope of l z d ( t) l is described by v t > 0 y z d (t) l = Jd (t). (113) . (110) where (107) V (w. on errors govern-ing w and z for the closed-loop kinematic controller can be found in (Dixon et al. 2001).. C = T T (CT + M T ) . m1 (w. Based on the Rayleigh-Ritz theorem (Dixon et al. n]. B = T T B . ( d ( t) can be seen as a function used for motion planning. it is convenient to introthe vector zd (t) and next use (100) to obtain duce the following parametrization: (103) The solution of (103) can be written as cos z d (0) = Jd (0) <p sin <p where 92 G (104) (n.f. (112) al z d whose norm approaches the constant e1 (c.k1w2 .k2 z T z. z) and m2 (w. (108) Remark 2.. (111) signal. In this way. where | G R2. a desired velocity signal. using a method which relies on linearizing the dynamic equa-tions is not possible because of unknown lateral skidding for matic modelT T . Using it. z) ?T?. In this case. we get V (w. According to (105). z) are positive scalar functions dependent on the actual tracking errors which denote the actu ent states. Assuming that z d (0)T z d (0) = (0).Modeling and control o f a 4-wheel skid-steering mobile robot we calculate the time derivative of the quadratic norm of condition. z) M T T Af T. 2003) and using the fact that M is a symmetric and positive definite 1 m1 (w. the kinematic control law given by (98)-(102) ensures that the position and orientation he right by satisfies the ed on the Lyapunov stability the-ory. (102)). It is essential if the initial error is quite high and the v Since the transformed velocity signal u (t) is no longer a control input. we get constraint given by (60).

U) #11 < p. z.k 1 w 2 . i.c) u . n r) tf.k1w2 . K.e..(w Jz + z Ta is defined as P2 According to (Spong. Based on ( 1 = . Dixon et al. . an additional control input Ta is introduced.6. equations governing the transformed errors w and + UT fw Jz + z + Ydtf . 1992. zg) + u T (w Jz + Q + and y dn y d21 y d31 y d41 yd12 y Y d (u d.57 U = 0.Bt z for the closed-loop system have the fol-lowing form: To achieve robustness to the parametric uncertainty w tf. (114) where i T Calculating the term tf m V 2 (w. Defining the difference between this parameter vector and the parameter vector tf that denotes the best-guess es + uT fw Jz + z+Mu +(5u d +R . it is assumed that the friction coef-ficients are the same for each wheel. we J ^sm (115) + 5 ud + R r = Y d (u d. we see that T T T m0 J0 Ms0"0 Mt 0 m 0 (118) uT (5T .tf. U T (w Jz+z+M"u dKozlowski and. Kozlowski and D. +55u d +R Bt) . 2001). z = .Bt) 0 tf = where #0 = #0 . z. Kozlowski and (114) in (1 V2 (w. a new ve-locity signal error is defined (Dixon et al. (127) that tf is bounded it is possible to find a coef-ficient p which fulfils the following inequality: V ?2 (w. u) = 2 w 2 + 1 z T z + 1 2u T Mu.k2zT z d = ^i and or tf is uncertain.k 2 z T zg (120) u = u d . Using (127). 2J (125) It is easy to prove that M" = (5 + (5 . u) To sim-plify the control law. 2001. we see the dynamic model (110) can be expressed in termstheu d (t)derivative of V2 yields z. (128) Now. 2 Since 57 . U d.k 2 z + (121) a + u.= u) is = aVvector of dynamical parametersM u d (w.amcs V2 (w. z. (123) where e is a positive constant scalar which can be set up as arbitrarily small. w Ju (122) T = B . Ta = Yd | Yd U T (131) z.. u) = ww+ z T z+u T Mu+2u T Mu. the following Lyapunov func-tion candidate is proposed: V 2 (w. (122) and then (108) in (124).u [ c . Using thisre-lationship and adding and subtracting the term -1 uTCu to the right-han T V 2 (w.B t + 1M u . iT (117) (126) + . For stability analysis.57 is a skew-symmetric matrix. the term etailed derivation of the closed-loop error system can be found in (Dixon et al. u d.k2zT z (119) + 2001. U) = . z.k 1 w + u T Jz + u T Jz.Majchrzak. = . making use of level. 2001). Pazderski Taking of time and then linearly parametrized.u.. 6. n r) = (116) d22 y d32 y d42 + Cu + R . Mazur. (124) M"u from (110) and using (121). q .k1w2 . 2002.. the time derivative of the Lyapunov function can be rewritten as = . q.

L-1 ki n (u va .k3 U T ii - where Ta is defined by (131).T . (132) = V2 ^ -gT-g. we get dt d i a = (2ki n) 1 T . (140) and (141). V3 (w. we obtain (137) Rearranging the terms in (137). z g.BQ + -QTQ < 0. based on (130). Next. + -Q [T is developed. 0 < we finally get u YdTgl pe fwJz + z +~Miid + fiud + R .ii T k 3 ii + e 2 . the following in-equality can be obtained: T d = B -1 (wJz + z + Y dtf0 + T a + k3 U) . (138) ^w . ii. 1 z T ~ . To determine the voltage control law on the motor level. the following desired torque signal Td is designed: After the substitution of (135) in (146).2 . Based on (144).2Ra i a . (135) U T .3 "d . z. zg. we can prove the follow-ingCalculating its time derivative and using (125). it is imposed that the following inequality is sa V2 (w. 20 e typical robust control term used in (Spong. z. ii.^ 489^J Modeling and control o f a 4-wheel skid-steering mobile robot amcs The next step of the backstepping procedure consists in defining an error torque signal (see Kozlowski and Maj-chrzak. ii) < . V3 (w.2ke n^w) g ia from (58) and then taking its time derivative.k1w2 - (139) k2 z gTz g . where k4 is an adjustable positive coefficient. ii) < -k 1 w 2 .. (134) (142) In order to design the control voltage u v a . z. the following control law on the volt-age level is proposed: (136) Uva = La (ki n)-1 ^7"d + k4 -g + BTiUj +Rai a + ke?W w . k4QT-Q + e2.2Ra i a .k 2 z T zg . . zg. we have 1^3 (w. the control law considering the drive model which was presented in Section T2. ii) < -k 1 w 2 . we get inequality: V2 (w. the following Lyapu = .k1w2 .2ke n^w ).k 2 z T zg . It is an important property (140) -g = td . l. where t denotes the actual torque signal. (146) T = L-1 ki n (u va . w Law on the Motor Level < ii 0. using (138) in (143). -g) = 2w2 + 2z 1 z + 2U Mii + 2QTT . ii) V ii T ii + |YdTii II |YdT U || pe p 2 + e2 = V1 (w.BTd 2 |YdT U || p + e2 < e2 . z) (133) + UT Ti Since the last term in (133) is bounded. the control signal of the drive system.k2 z T z . 1992) and allows us to limit the chattering of the control signal.k 3/ 3 (w.k4-gT-g. 2 (141) Tu) T (tf P2 Y dT U YdTUll p + e2 After some simple calculations. z. ii) .U T k 3 U + Yd 1 1 . Therefore.t . -g) = (w. Using (145) in (141).

z) which appears in (112). (149) in 147) gives V3 (w. I = 0. Y) and /z t (X.4 0. f) + ?2.55 [s?Als]. signedTby Jedwabny et al. R = 3. z. Table 1.9 [fi].1 0. 10.2 2 2 > km ( w2 +z T z+U T M U+ fTf ) . z.e. / 1 . 0. Y ). Friction coefficients: (a) W 2 0 X [m] ( X . a = a e k n 0 (152) s196 va max| = 12 [V] where 71 .2 0.22 [mH]. (119) are fulfilled.min Tu. 72 . Pazderski ere m2 is an upper bound of the function m2 ( w. + ~ T. T k . =12. It shows that the phase shift between zd1 and zd2 plays a very impo . ti.8 1 K. f) 0 Y [m] -2 -2 (b) Fig. 2002). (2004). (15 er and the vehicle model are bounded (it is proved by Dixon et al.m2] Motors l fl l < exp(-71 t) + ?2/?2 + /?3 exp(-72t ) +L ^46 0. we have V3 G L?.8 i mined for a kine-matic and a dynamic controller (without a motor model)..6 k1 22 kmin h -w ? min +- k 2 T h min ? z T z- k4 T T. U.0036 [kg ."0. / 2 . i.15 =*.05 0 2 V3 (w. z. Surface m e (0. We have dTO. k4 1 m2 (148) 0. c = 34 [mm]. Consequent 0.2 l VA 1. k = 8. |i max | =1. Y) describing the fric-tion coe illustrated with respect to the initial value of the oscillator signal z d .2. Y ). Model and controller pa nditions (60).. r = 26. k2. The friction surface is characterized by the functions (X. U. Kozlowski and D. (b) /ks ( X . k u. f) < 7V3 (w. T min 0. the proposed con?trol law given by (98)-(102). = 8.5 [mm] Dynamic m = 1 [kg]. / 3 and / 4 are positive constants. (139) and (145) ensures the position and orientation trac Kinematic a = b = 39 [mm].amcs Now choose the parameter k3 high enough such that V t > 0 k3 > max k1. Kozlowski and Majchrzak. The torque signal is limited and cannot achieve a value greate /z s 1 0 ] .52 [2^^^]. 2001.

2 0 -Ti \par \cell { N ' 0. zd (t) indirectly determines the actual position and the orientation error. x r 1 Modeling and control o f a 4-wheel skid-steering mobile A \par / v robot s.1 53 0. . The steady-state position and orientation errors are bounded as follows: \n s p rs b ou e u < 1[m ]. and (d) en that the initial regulation error is highly re-duced.0 /6% -6/ 6% \par -y / t 1/6% .5 /' y 0 X [ml v 0.3 0.2 0 -0-1 aj /X -0.5 1 1/3% / ? TL 0. i.8 0.1 -0.4 0.e. Th . 1.6 ^ 0.5 l -0-0.6 1. as a consequence. the regula \c 2 1 # \n s p rs b s3 ou e u < 0 1[ra ].8 0.3 0.5 m for t > 25 [ s ] : \nosupersub X (cf..2 0. The limitation of the control signal results in signal chattering with a high frequency and.2 0.2 -5/6% III 1 8(0)=1 ? 8d(0)=1.4 0.1 / ' .5.5 0 y X [ml (b) 0.2 y v' \cell 1/2% > ^2/3?t 5/6% -1/6% "1/2 % / / 1 /' 1 -0.1 -0. (c) Torque signal for Sd ( 0 ) = 2.2 0 1 4 5 0 2 3 time [sl (d) ect to z d (0): (a) Performed path with respect to p. (b) Performed path with respect to Sd (0). m \nosupersub Y \n s p rs b ou e u < 0 [m ] . (98)-(100)). 1 -2/3% 1 1 0 .25 y ^ 0.2 l 0.0 d and ) approaches the trajectory gener-ated by the oscillator. 1 < i- 0 a 0 tr?ai "-0.

amcs 1.5 I 0 X -0.5 0q(t) " 0.2t [m].5. In Figs. a1 = 0.5 I K.0 d r allows us to obtain(t) is calculated states and to rapidly limit motion steady-state tracking errors are bounded as be considered for a pra while 9r good transient by the reference model (64). (e) the following parameters were assumed: k 1 = 2. Therefore. choosing zd ( 0 ) should follows: \n s p rs bX o u e u <5[m ].8 0.5 5 10 15 20 time [sl 25 30 20 10 r 15 10 5 .2i 1 0.4 0.0 d ented in Figs. 14(b) and 14(c) it is clear that the error and velocit . o u e u \n s p rs b m \n s p rs bY o u e u < o u e u \n s p rs b \nosupersub 9\nosupersub 4[m ] m < 0 4[ra ]. The regul for t > 25 [ s ] : \n s p rs bX ou e u \nosupersub < 1 [m ]. 0 =s -5 -10 -15 0 (c) (d) 30 ulation case for q(0) = [ 0 1 0 ]T: (a) Performed path. 1.6 ^ 0.2 0 "?:0. (b) Regulation errors.1 [m].2t . 1 h rw 1 + 0.5 qr(t) X [ml (a) o q(0) " a qr(0) .2. but according to Figs.2t cos0. Pazderski X-X r Y-Y r 9-9 r ii I cr> 1 0. 1 1. regulation errors. the path.5 1. \n s p rs bY m ou e u 0 \n s p rs b ou e u < 1 [m ] m Yr (t) = sin0. auxiliary signals and voltage signals are shown. The changes. for t > 30 [s] : 14(a)- and In Fig.7. 14(e) the friction coefficients are illustrated. a 1 = 0. . (5d (0) = 1. . Sd ( 0 ) = 3. and (d) Voltage signal. The reference Xr (t) = ' 1 1 +0. 9 = n. 13(a)-(d). (c) Auxiliary error and oscillator signal. Kozlowski and D.2t and \n s p rs b ou e u < 0 5[ra ].3. 92 = n/12.

roblem the following reference trajectory was assumed: s necessary to increase the coefficient k 3 to improve the performance of the con?troller.5.4 X 10 -4.15t 1] [m].n ]T: (a) Performed path.2n/3. It is interesting to see that although a trajector (1 + 0. 20 30 40 time [sl A q r (?J 15 10 5 a .2 0 < X [ml 0. 0 =s -5 -10 -15 60 (c) (d) on case for q(0) = [ 0 1 . dynamic and drive model of a 4WD skid-steering mobile robot has been presented.2sin0.y^ Modeling and control o f a 4-wheel skid-steering mobile robot T3 1. a = 0.2 0 -0-0. Sd (0) = 1.2 1 0. s are k 3 = 10. f = . and (d) Voltage signal. (b) Regulation errors.9t)cos0. k 4 = 10.4 o q(0) -0.2 0. since the dynamic properties algorithm considering the kinematic. F for t > 20 [s] : |X | < 12 [mm].6 ^ -2 X-3 i e-e X-X r Y-Y r r 50 60 \cell -0. (c) Auxiliary error and oscillator signal. evious simulations. To solve the ki .02 [rad]. |Y| < 11 [mm] and |9| < 0.q(t) q(t) <X> I t \par \par \cell er> I ^ 0.3. In this simulation experiment the following steady-state errors are recorded (cf.4 0.6 .8 0.

5 >< 0 -0.vs -? i 50 60 time [sl 0.2 -0-1 l (a) 2 dx/dt de/dt dx /dt 1 de /dt r Pi ^ -5 15 10 5 ><-0. Pazderski l 0.2 0 10 20 30 40 """"""" (e) case for q(0) = [ 0.6 T3 2 r e-e r ?i 1.5 0.5 -n/2 ]T: (a) Performed path.6 . (b) Tracking errors.5 u er> I 1 >H 0.X-X r Y-Y WtfS\ 494 K.8 0. (c) Linear and angular velocities.5 -1 0 10 20 30 40 time [sl 50 60 I 0 ni T3 ?-1 -10 -15. Kozlowski and D.2 0 er> I0. and (e) Frict II nu . (d) Voltage signal. 30 (c) 0.8 40 50 60 time [sl ? -? 10 -2 20 (d) 0.

(c) Linear and angular velocities.5 -2 0 time [sl (c) 0.5 0 -0.2 5 10 15 20 25 30 time [sl (e) ase q(0) = [ .4 0. A 0. 0.6 r - 0.. 1 \cell /?.1 0 -n/4 ]T: (a) Performed path.5 0 T3 .. (d) Voltage signal.5 1 0. and (e) Friction c .5 1 0..5 X X -1.dx/dt .Modeling and control o f a 4-wheel skid-steering mobile robot .8 .5 (b) i j 1 /1 ? i ?V -7 \cell (a) w ii t/ i 1.5 -l -1.. (b) Tracking errors. A" / 0 / \par \cell A i ? -1 /A 0.. de/dt 15 10 5 *u 0 3 -5 -10 -15 dx/dt de/dt 5 10 15 20 25 30 0 (d) v ?S -1 v -1.5 i'Y> 1.

No. (2004): Nonholonomic mobile robot MiniTracker re?search a References : A new control algo-rithm for a nonholonomic mobile robot. 9. dbackThe obtained convergence is exponential.Arch. tical stabilization ofdrift-less systems on Lie groups: The transverse function ap-proach. Automat. 11.E. done by Caracciolo etal. 37-69. 37. 12. . the presented approach rors. Contr. .Proc. 48. . InA survey. Vol.liiUfcl. 6-th Int. W. 223-244.. and Kozlowski K. Sci. Topical Meeting Robotics and Re ontrol of robot manipu-lators. Automat. pp. control ofnonholo-nomic wheeled vehicles: contrast .IEEE Trans. : Control o f a four-wheel vehicle using kinematic oscillator.the previous work onholonomic motion plan-ning: Steering using sinusoids. Vol. pp.Amer. 3-4. 16.. toArch.. Climbing and Walking Robots and their Supporting Tec e Dynamics. Vol. Pazderski for the kinematics and the dynamics ofmobile manipulators: A comparative study. Sci. No. ..Oxford: Butterworth-Heinemann. Nuclear Soci. . ki M. Conf. pp. Automat. 10-th Int.E. Contr. (2004): Tracking and regulation control o f a skid steering vehicle. Michalek M. Sci. Contr. Vol.. 277-342. Lawniczak M.. Contr..Arch.IEEE Trans. Contr. 496 K. No. No. 11. er. .IEEE Trans. .. Kozlowski and D. Pazderski D. 48.. . . (1999). Michalski M.

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd