Вы находитесь на странице: 1из 13

1

Sliding mode control of a 4WD4WS vehicle for path tracking


Amit Kumar Anand, Kamal Kalwani, Ramprasad Potluri
Abstract

I. I NTRODUCTION In this work, we consider a vehicle with four wheels. Each wheel has a driving and a steering motors. We call such a vehicle a four wheel drive four wheel steer (4WD4WS) vehicle. 4WD4WS vehicles have several important advantages over conventional vehicles greater stability (due to lowered centery of gravity) and almost 20% greater efciency [1] due to absence of transmission, non-polluting as opposed to an internal combustion engine vehicle, greater manouverability (zero turning radius), and can work well in split- conditions. However, these vehicles are challenging to control. Exist several works that have considered vehicles that are only 4WD (see, for example, [2], [3], [4], [5], [6]). Only a few works exist that have studied the control of 4WD4WS vehicles: [7], [8], [9], [10], [11], [12], [13]. Amit, Praveen, need to do a Google search with keywords such as four wheel drive four wheel steer, four wheel steer four wheel drive, four wheel drive, four wheel steer, four wheel drive four wheel steer + control, etc 4WD4WS vehicles belong to the class of multi wheel drive multi wheel steer (MWDMWS) vehicles. All wheel drive all wheel steer vehicles also belong to this class. Example of MWDMWS vehicles are the various mars and lunar rovers (see, for example, [14], [15]). While the work done with rovers can be useful in understanding the kinematics, dynamics, and control of 4WD4WS vehicle, the kinematics and dynamics used here have been borrowed from the literature on road vehicles. Exist previous works that have studied the control of 4WD4WS vehicles with the goal of path-tracking, improving vehicle stability, and slip reduction. Amit, please cite as many works as you can. We want this vehicle to automatically track a desired path with as little slip as possible. This problem has been studied in a few other works. Which ones? This paper is organized as follows. Section II, presents the mathematical model of the vehicle that we borrowed from [13]. This model is complicated. One main contribution of our work is to bring this model into a form that is amenable to Slotine and Lis sliding mode control method for square systems. Section III reviews Slotine and Lis sliding mode control algorithm for MIMO systems [16, pages 303 306]. There are a few typographical errors in that algorithm, that were corrected in [17]. We present the corrected algorithm. Section IV shows how we apply Slotine and Lis algorithm to our problem. Section V presents the results of simulation, and analyzes the results. Section X concludes with a discussion of future work. Praveen and Amit: Please verify the claims that I am attributing to the various sources that we have cited. For example, was it 10%, 15% or 20% efciency? II. M ATHEMATICAL MODEL
OF A

4WD4WS

VEHICLE

The mathematical model of the vehicle comprises four parts: the vehicle body dynamics, the dynamics of individual wheels, the path-tracking kinematics, and the tyre-road friction that relates the rst two parts. In the remainder of this section, we reproduce these four parts from [13] and [18], that in turn reproduced them from ?????. Note that the resistances (such as air drag) that Peng et al [18] considered are ignored in [13]. We too ignore these resistances. We attach an x y z coordinate system to the center of gravity of the vehicle, and use the following notation: m mass of the vehicle. V velocity vector of the center of gravity (CG) of the vehicle. v is the magnitude V of V. sideslip angle. This is the angle between vehicles actual direction of travel and the heading of the vehicle. yaw rate. The yaw rate of a car or a rigid body is its angular velocity about its yaw axis. This is also equal to the rate of change of heading. Jz moment of inertia of the vehicle about z-axis. Mzj torque around the z-axis due to the friction forces on wheel j.

(a)
Fig. 1. 4WS4WD vehicle body

(b)

A. Vehicle Body Dynamics mv cos mv( + ) = sin 0 Jz


4

sin cos 0

4 0 fxj fyj 0 1 j=1 Mzj lf ld fx2 fy2 lr fx4 fy4

Mzj =
j=1

ld +

lf ld

fx1 fy1 lr

+ fx3 fy3

ld +

Amit, please correct this equation or the gure on which it is based. Here, fxj andfyj are external forces acting on wheel, lf and lr are the distances measured respectively from the CG to the front and the rear axles, and ld is half of the wheel base measured to the center of the tread. B. Wheel Dynamics

(a) Fig2

(b) Fig3

The torque Tj developed by the j-th wheels driving motor overcomes the torque due to the friction force fxj cos j +fyj sin j to accelerate the motor through j according to the fundamental torque equation d j = Tj rej (fxj cos j + fyj sin j ) (1) dt Here, Iwj is the moment of inertia of the j-th driving motor - wheel system referred to the motor shaft, and rej and j are respectively the effective radius and steering angle of wheel j. Iwj C. Path-Tracking Kinematics Is this correct? yc = v sin v cos = 1 ++ ref + yc where the path curvature ref , assumed to be constant, is given. D. Tire/Road Friction SLj SSj fzj cos j Res ksj fzj sin j Sj Sj SSj SLj fzj sin j + Res ksj fzj cos j = Res Sj Sj 1 rej wj cos j ||Vj || = rej wj sin j max(rej wj cos j , ||Vj ||)

fxj = Res fyj SLj SSj

Here, Vj is the velocity of the center, j is the side slip angle, ksj is the attentuation factor of the tyre-tread prole, fzj is the dynamical normal load, and j = j j , all for the j-th wheel. Vj is related to V, , as follows: V1 = V2 = V3 = V4 =

vx1 ||V|| cos ld = ||V|| sin + lf vy1 ||V|| cos + ld vx2 = ||V|| sin + lf vy2 ||V|| cos ld vx3 = ||V|| sin lr vy3 ||V|| cos + ld vx4 = ||V|| sin lr vy4

is the combined wheelslip, SLj is the longitudinal wheel slip, SSj is the lateral wheel slip, and Sj = SLj SSj Res > 0, is a scalar saturation function depending on the magnitude of combined wheel slip ||Sj || and the road condition. III. S LOTINE AND L I S
SLIDING MODE CONTROL ALGORITHM FOR SQUARE

MIMO

SYSTEMS

This algorithm is borrowed from [16, pages 303 306]. We have adapted it for the following nonlinear multi-input system: x1 u1 f1 (x) b11 (x) b12 (x) b1m (x) x2 f2 (x) b21 (x) b22 (x) b2m (x) u2 = + xm fm (x) bm1 (x) bm2 (x) bmm (x) um This system is also written compactly as follows: x = f (x) + B(x)u

(2)

Here, x = x1 x2 xm . This is a square system in the sense that the number of control inputs ui equals the number of outputs xi to be controlled. We want the state x to track a desired time-varying state xd in the presence of parametric uncertainties. 3

Two assumptions are made. First, that the parametric uncertainties are within the range space of the input matrix B (of components bij ). Since B is a square m m matrix, this means that B is invertible over the whole state space, a controllability like assumption. Second, we assume that the estimated input matrix B is invertible, continuously dependent on parametric = B in the absence of parametric uncertainty. uncertainty, and such that B We write uncertainties on f in additive form, and uncertainties on the input matrix B in multiplicative form. |fi fi | Fi , B = (I + )B, i = 1, ..., m |ij | Dij ,

where I is the m m identity matrix. The inequality represents additive uncertainty. It says that fi fi Fi ; fi + Fi , which is the same as saying that fi = fi Fi , where fi is the estimated mean value of fi and Fi is a constant that represents the maximum uncertainty in this estimate. The equality represents multiplicative uncertainty. For example, a resistance whose value is R = 50 10% can be expressed also as R = (1 + )50 , where || D = 0.1. On similar lines, in the case of m = 1, B = (1 + )B with || D. This can alternatively be written as Bmin B Bmax with Bmin = (1 D)B and Bmax = (1 + D)B. If we know Bmin and D = (Bmax Bmin )/(Bmax + Bmin ), and B = (Bmax + Bmin )/2. As an aside we note that, in the general case of m > 1, (1 + 11 ) 11 b (1 + 12 ) 12 b (1 + 21 ) 21 b (1 + 22 ) 22 b (I + )B = (1 + m1 ) m1 (1 + m2 ) m2 b b Bmax , then we can solve for D and B to obtain

(1 + 1m ) 1m b (1 + 2m ) 2m b (1 + mm ) mm b

In our work, we choose, (Amit, please explain how you obtained this; are there other choices for D?) From above , (I D)B B (I + D)B From this we got the inspiration to choose 1+D= Bmax Bmin Bmax D= Bmin Bmax (3)

where Bmax and Bmin are such that the ij-th element of B in Equation (2) is greater than or equal to the ij-th element of Bmin and less than or equal to the ij-th element of Bmax . D is the m m square matrix whose components are Dij . Let xid be the desired value that the variable xi is needed to track, and let si be dened by si = xi xid , i = 1, , m.

The controller design problem is translated into one of nding a control law for the vector u that veries individual sliding conditions of the form 1 d 2 s i |si |, i > 0 2 dt i in the presence of parametric uncertainty. Letting be the vector of components ki sgn(si ) (here, sgn() is the signum function) and choosing the control law to be of the form u = B1 (xd ), where, = f f
m

f1

f2

fm

Sliding conditions are veried if ki and i are chosen so as to satisfy the following inequality. (1 Dii )ki Fi +
j=1

Dij |xjd fj |
j=i

Dij kj + i , i = 1, , m.

In our work, we use i = 0. Amit, we cannot use = 0. It needs to be strictly positive. Using a theorem due to Frobenius-Perron, [16] claims that the last inequality above uniquely denes a set of non-negative ki for the system of Equation (2) satisfying the two assumptions stated above. Amit will verify the above 2 expressions. I would like to see how you are solving for ki and i using MATLABs linprog. this is done

IV. I MPLEMENTATION OF

THE CONTROL ALGORITHM

We present the main contributions of our work in this section: Step 1:Expressing the mathematical model of Section II in the form x = f (x) + b(x)u, where b(x) is not yet in the form of a square matrix. Step 2:Converting the Equation (4) into Equation (2) through suitable approximations. Step 3:Applying the algorithm of Section III on Equation (2). A. Step 1 We have three sets of equations from Section II. The rst set of equations is as follows: Is this correct? sin yc cos = 1 ref + yc The second set of equations is as follows: v 0 = + 0 where, fxj fyj = j=1 Mzj ld
4

(4)

0 1

v +

(5)

cos m sin mv 0

sin m cos mv 0

Jz

4 fxj fyj 0 j=1 Mzj 1 lr


4 j=3

(6)

4 j j=1 (1) Aj

cos j +

4 j=1 Aj cos j 4 j=1 Aj sin j 2 lf j=1 Aj sin j

Aj sin j

(7)

with, Aj =

fzj kj rej j ||Vj ||fzj kj max(rej j ; ||Vj ||)

and we have assumed that j 0. How justied is this assumption? Does our control law work well even if j = 0? The third set of equations is as follows: Tj rej fxj rej fyj 1 cos j , j = 1, , 4. (8) j = Iwj Iwj Iwj sin j In these three sets of equations we can identify the nine states yc , , v, , , 1 , 2 , 3 , 4 , and the twelve control variables Tj (j = 1, , 4), sin j (j = 1, , 4), and cos j (j = 1, , 4). However, as sin and cos are related as sin2 +cos2 = 1, we may use either of the sin j or cos j as our control input. Hence control inputs are reduced to eight. Thus, at this stage, we have a nine-state eight-input system. This is not a square system. To convert it into a square system, we proceed as follows (see Figure 5). Control inputs for Eq.5 are v, + and for Eq.6 and Eq.8 are Tj and either of sin j or cos j . Eq.5 controls the kinematics of the vehicle. Eq.6 and Eq.8 controls the dynamics of the vehicle. Consequently, we have divided the problem of path tracking in two parts. Part one which controls the states yc and with control inputs v and + . Part two controls the states v, , , w1 , w2 , w3 and w4 with control inputs Tj (j = 1, , 4)s and either of sin j or cos j (j = 1, , 4). Henceforth Part one and Part two referred as Sys2 and Sys7 respectively, referring to two and seven states controlled in Part one and Part two respectively. . Variables with 2 and 7 subscript will denote that these variables are used in Sys2 and Sys7 respectively. Sys2 is the side slip angle. Ideally we desire to have no side slip and no variation in it. Therefore desired value for is chosen to be zero. Therefore state space model for controlling the kinematics of the vehicle is Is this correct? sin 0 yc v = cos (9) 1 1 + yc ref 5

Sys7 From Eq. 6, Eq. 8 and Eq. 7 following state space model can be built v 1 2 3 4 0 0 E1 E2 E3 E4 0 0 0 0 1 I2 0 0 Ej = rej fzj kj rej j ||Vj ||fzj kj Ij max(rej j ; ||Vj ||)

= 0 0 0 1 I1 0 0 0

+ 0 0 0 0 0 1 I3 0 0 0 0 0 0 0 1 I4

A1 cos m A1 sin mv ld A1 Jz 0 0 0 0

A2 cos m A2 sin mv ld A2 Jz 0 0 0 0

A3 cos m A3 sin mv ld A3 Jz 0 0 0 0

A4 cos m A4 sin mv ld A4 Jz 0 0 0 0

A1 sin m A1 cos mv lf A1 Jz 0 0 0 0

A2 sin m A2 cos mv lf A2 Jz 0 0 0 0

A3 sin m A3 cos mv lr A3 Jz 0 0 0 0

A4 sin m A4 cos mv lr A4 Jz 0 0 0 0

T1 T2 T3 T4 cos 1 cos 2 cos 3 cos 4 sin 1 sin 2 sin 3 sin 4

where ,

As mentioned before we can use either of sin or cos as our control input. So to reduce the number of control inputs we took

2 the average value of sin 1 , sin 2 , cos 3 and cos 4 to be . So the above state space equation 2 ((A1 + A2 ) sin + (A3 + A4 ) cos ) m v + 2 ((A1 + A2 ) cos (A3 + A4 ) sin ) mv 2 (lf (A2 + A1 ) + ld (A4 A3 )) 1 = + Jz 2 E1 3 E2 4 E3 E4 A2 cos A3 sin A4 sin A1 cos 0 0 0 0 m m m m A1 sin A2 sin A3 cos A4 cos 0 0 0 0 mv mv mv mv ld A1 ld A2 lr A3 lr A4 0 0 0 0 Jz Jz Jz Jz 1 0 0 0 0 0 0 0 I1 1 0 0 0 0 0 0 0 I2 1 0 0 0 0 0 0 0 I3 1 0 0 0 0 0 0 0 I4 B. Step 2

is reduced to the following:

T1 T2 T3 T4 cos 1 cos 2 sin 3 sin 4

Sys7 Rank of B7 matrix is 7 , so this is not full ranked. To make it full rank we suggest to combine the controls of rear wheel i.e. sin 3 and sin 4 . By doing this number of columns in B matrix will be reduced to 7 making it full ranked. So for rear wheels we propose control input P = A3 sin 3 + A4 sin 4 . This form is also amenable to the algorithm mentioned in section III. The new state-space equation is. 2 ((A1 + A2 ) sin + (A3 + A4 ) cos ) m + 2 ((A1 + A2 ) cos (A3 + A4 ) sin ) mv 2 (lf (A2 + A1 ) + ld (A4 A3 )) = + Jz E1 E2 E3 E4 A1 cos A2 cos sin 0 0 0 0 m m m cos A2 sin A1 sin 0 0 0 0 T1 mv mv mv ld A1 ld A2 lr T2 0 0 0 0 Jz Jz Jz T3 1 0 0 0 0 0 0 T4 I1 cos 1 1 0 0 0 0 0 0 cos 2 I2 1 A3 sin 3 + A4 sin 4 0 0 0 0 0 0 I3 1 0 0 0 0 0 0 I4

v 1 2 3 4

C. Step 3 For the purpose of simulation following values have been used 7

TABLE I VALUES OF PARAMETERS USED FOR SIMULATION . T HE INITIAL VALUES ARE FROM [12]. W HERE THE MINIMUM AND MAXIMUM VALUES ARE LISTED FOR A PARAMETER , THE VALUES OF THIS PARAMETER CHANGE FROM Need to explain how you are using this table Variable rej j V m Jz Iwj ld lf lr fzj kj yc ref Initial value 0.31 90 20 /3 0.3 /3 1480 1950 0.7 0.751 1.421 1.029 3000 15 1 1/400 Minimum Value 3 1 0.1 0.1 /3 1 1 0 Maximum value 322 100 /2 /2 /2 6000 30 20 Unit m rad/s m/s rad rad rad kg kgm2 kgm2 m m m N m m

Suppose we start with values very close to zero (say, 0.001). How will our algorithm work? Variables for which maximum and minimum values have been mentioned states that these variables could vary between these two limits over a period of time. Variables for which no maximum or minimum value is mentioned implies that these variables are constant. Minimum value for each variable was required to be chosen non zero otherwise some rows of B7 would become zero and lose rank. Therefore the the algorithm mentioned in Section III cannot be applied. Because of the same reason value of was chosen to asymmetric otherwise the rst row of B2 becomes zero and loses rank. Sys2 Since f2 = [0 0] . Therefore f2 = [0 0] and F2 = [0 0] . Also there is no parametric uncertainty in Sys2.So, 2 = B2 . xd = xd = [0 0] for Sys2. s2i = x2i x2d . D2 is calculated as mentioned in Eq 3. k2i is found through linear B programming satisfying Eq.??. 2 is calculated as mentioned in section III . Sys7 The only parametric uncertainty in Sys7 is in the product of fzj and ksj which varies between 0 to 18000. So for the purpose of calculating f7 and B7 we took the average value 9000. So, 2 re1 1 ||v1 || re2 2 ||v2 || re3 3 ||v3 || re4 4 ||v4 || 9000 (( + ) sin + ( + ) cos ) m max(re1 1 ; ||v1 ||) max(re2 2 ; ||v2 ||) max(re3 3 ; ||v3 ||) max(re4 4 ; ||v4 ||) re1 1 ||v1 || re2 2 ||v2 || re3 3 ||v3 || re4 4 ||v4 || 2 + 9000 (( + ) cos ( + ) sin ) mv max(re1 1 ; ||v1 ||) max(re2 2 ; ||v2 ||) max(re3 3 ; ||v3 ||) max(re4 4 ; ||v4 ||) 2 re2 2 ||v2 || re1 1 ||v1 || re4 4 ||v4 || re3 3 ||v3 || 9000 (lf ( ) + ld ( )) Jz max(re2 2 ; ||v2 ||) max(re1 1 ; ||v1 ||) max(re4 4 ; ||v4 ||) max(re3 3 ; ||v3 ||) re1 f7 = 9000 (re1 1 ||v1 ||) I1 max(re1 ; ||v1 ||) re2 9000 (re2 2 ||v2 ||) I2 max(re2 ; ||v2 ||) re3 (re3 3 ||v3 ||) 9000 I3 max(re3 ; ||v3 ||) re4 9000 (re4 4 ||v4 ||) I4 max(re4 ; ||v4 ||)

Similarly, re1 1 ||v1 || cos max(re1 1 ; ||v1 ||) 9000 m re1 1 ||v1 || sin max(re1 1 ; ||v1 ||) 9000 mv re1 1 ||v1 || ld max(re1 1 ; ||v1 ||) 9000 Jz 0 0 0 0

B7 =

0 0 0 1 I1 0 0 0

0 0 0 0 1 I2 0 0

0 0 0 0 0 1 I3 0

0 0 0 0 0 0 1 I4

re2 2 ||v2 || cos max(re2 2 ; ||v2 ||) 9000 m re2 2 ||v2 || sin max(re2 2 ; ||v2 ||) 9000 m re2 2 ||v2 || ld max(re2 2 ; ||v2 ||) 9000 Jz 0 0 0 0

sin m

cos mv lr Jz 0 0 0 0

and x7d = 0 0 0 0 0 0 0 where v and are the x7d = v 0 v /re1 v /re2 v /re3 v /re4 control inputs in Sys2. So control inputs in Sys2 acts as desired value of v and in Sys7. s7i = x7i x7d . D7 , k7i and 7 is calculated as in C. Step 3 Sys2 section. F7 is calculated as |f7 maxf7 |, where maxf7 is the maximum value of f7 . V. S IMULATION
RESULTS

The results obtained from the simulations are presented and discussed here. We have tried to give a physical realization of our results. A physical insight to the result will, thus, help the readers to understand the solution to the problems that the system is able to tackle. Also since sliding mode control is used chattering effect can be seen in all graphs on magnication. VI. V ELOCITY
VS TIME

It can be seen from the graph that the vehicle, to track the path and attain the desirable result, slows down to a lower constant velocity.
V vs time 20

18

16

14

V (in m/s)

12

10

0.2

0.4

0.6

0.8

1 time (in sec)

1.2

1.4

1.6

1.8

Fig. 2.

Velocity vs time

VII.

VS TIME

We see, here, from the graph that the system is able to attain the desirable ( = 0), once the system is able to reach the desired path to be tracked. It is desired for beta to go to zero because once the vehicle is on the path, the sideslip angle should now go to zero, that is, the angle between the velocity of the vehicle and the body-xed coordinates should go to zero. VIII.
VS TIME

Through the graph of versus time, we can see that once on the path, gamma for the vehicle attain a constant value. This can easily be understood physically. , the yaw rate of the car, changes its value accordingly such that the motion of the car is directed towards the path that we need to track. Once on the path, the car now needs to move on the path which, in our case, is circular with a non zero ref . Now, to further move on such a path (a path with non-zero ref ), the car should have a constant yaw rate that would allow the car to stay on the path. 9

Beta vs time 1.2

0.8

Beta (in rad)

0.6

0.4

0.2

0.2

0.2

0.4

0.6

0.8

1 time (in sec)

1.2

1.4

1.6

1.8

Fig. 3.

Beta vs time
gamma vs time 0.4

0.2

0.2 gamma (in rad)

0.4

0.6

0.8

1.2

0.2

0.4

0.6

0.8

1 time (in sec)

1.2

1.4

1.6

1.8 x 10

2
4

Fig. 4.

gamma vs time

IX. j

VS TIME

It is observed from the graph that the wheels of the vehicle attain a constant angular velocity once they reach the desired path to be tracked. This can be understood physically in the sense that once the vehicle is on the path, we already saw that the velocity of the vehicle attains a constant value and to be in agreement with the vehicle of the car, the angular velocity of the wheels of the car attain a constant value. X. C ONCLUSION This study presents an approach to implement sliding mode control in the autonomous controlling of a 4WD4WS vehicle. Equations relating the kinematics of the vehicle are derived (the rigorous derivation of which is shown in the report) and the derived complicated equations are then used to implement the sliding mode control. The theory involving implementation of sliding mode control is described and then used. The simulation shows the effectiveness of the approach. A physical realization into the graphs obtained from the simulation is also given. The study, thus, focuses on using an effective method in the autonomous control of 4WD4WS vehicle. R EFERENCES
[1] Willie D. Jones. Putting electricity where the rubber meets the road. IEEE Spectrum, pages 1213, July 2007. [2] Kazuya Kitajima and Huei Peng. H control for integrated side-slip, roll and yaw controls for ground vehicles. In Proceedings of the 5th International Symposium on Advanced Vehicle Control (AVEC 2000), 2000. [3] U-Sok Chong, Eok Namgoong, and Seung-Ki Sul. Torque steering control of 4WD electric vehicle. In Proceedings of the IEEE Conference on Power Electronics in Transportation, pages 159 164, 1996. [4] Farzad Tahami, Reza Kazemi, and Shahrokh Farhanghi. A novel drive assist stability system for all-wheel-drive electric vehicles. IEEE Transactions on Vehicular Technology, 52(3):683 692, May 2003. [5] M. Parent. Distributed motion control of an electric vehicle. In 5th International Workshop on Advanced Motion Control, page 573, Coimbra, Portugal, 1998. IEEE. [6] Shin-ichiro Sakai, Hideo Sado, and Yoichi Hori. Motion control in an electric vehicle with four independently driven in-wheel motors. IEEE/ASME transactions on mechatronics, 4(1):9 16, March 1999. [7] Maxim Makatchev, John J. McPhee, S.K. Tso, and Sherman Y.T. Lang. Modelling and control of a four-wheel-steering mobile robot. IEEE Transactions on Automatic Control, pages 100 108, 1999. [8] Shou-Tao Peng, Jer-Jia Sheu, and Chau-Chin Chang. On one approach to constraining wheel slip for the auto nomous control of a 4WS/4WD vehicle. In Proceedings of the 2004 IEEE International Conference on Control Applications, pages 1254 1259, 2004. [9] Johan Boot. ATV control: regulating a 4WD/4WS autonomous guided vehicle. Masters thesis, Eindhoven University of Technology, 2005. [10] Dragos Maciuca. Design, modeling and control of steering and braking for an urban electric vehicle. Technical report, University of California at Berkeley and INRIA, 1996.

10

[11] Chengliang Liu, Mingjun Wang, and Jun Zhou. Coordinating control for an agricultural vehicle with individual wheel speeds and steering angles. IEEE control systems magazine, 28(5):21 24, October 2008. [12] Shou-Tao Peng, Chau-Chin Chang, and Jer-Jia Sheu. On robust bounded control of the combined wheel slip with integral compensation for an autonomous 4WS4WD vehicle. Vehicle System Dynamics, 45(5):477 503, May 2007. [13] Shou-Tao Peng. On one approach to constraining the combined wheel slip in the autonomous control of a 4ws4wd vehicle. IEEE Transactions on Control System Technology, 15(1):168 175, May 2007. [14] G. Ishigami and K. Yoshida. Steering characteristics of an exploration rover on loose soil based on all-wheel dynamics model. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2005), pages 3099 3104, 2005. [15] J. Borenstein. Control and kinematic design of multi-degree-of-freedom mobilerobots with compliant linkage. IEEE transactions on robotics and automation, 11(1):21 35, February 1995. [16] Jean-Jacques E. Slotine and Weiping Li. Applied nonlinear control. Prentice Hall, 1991. [17] Manavaalan Gunasekaran and Ramprasad Potluri. Typographical corrections in slotine and lis sliding mode control algorithm for mimo systems. Private discussion, IIT Kanpur, 2010. [18] Shou-Tao Peng, Chau-Chin Chang, and Jer-Jia Sheu. On robust bounded control of the combined wheel slip with integral compensation for an autonomous 4ws4wd vehicle. Vehicle System Dynamics, 45(5):477 503, May 2007.

11

w1 vs time 90

80

70

60

w1 (in rad/s)

50

40

30

20

10

0.2

0.4

0.6

0.8

1 time

1.2

1.4

1.6

1.8

(a) w1 vs time
w2 vs time 90

80

70

60

w2 (in rad/s)

50

40

30

20

10

0.2

0.4

0.6

0.8

1 time

1.2

1.4

1.6

1.8

(b) w2 vs time
w3 vs time 90

80

70

60

w3 (in rad/s)

50

40

30

20

10

0.2

0.4

0.6

0.8

1 time

1.2

1.4

1.6

1.8

(c) w3 vs time
w4 vs time 90

80

70

60

w4 (in rad/s)

50

40

30

20

10

0.2

0.4

0.6

0.8

1 time(in sec)

1.2

1.4

1.6

1.8

(d) w4 vs time

12

V d_/dt + ` Ycd = 0 T=0

2x2 system

Yc T

wjd Beta

7x7 system

7 states

V,Beta,gamma,w1,w2,w3,w4

Fig. 5.

Cascade control of the 4WD4WS vehicle.

13

Вам также может понравиться