You are on page 1of 6

SICE Annual Conference 2008 Aug. 20-22, 2008, Univ.

of Electro-Communications, Japan

Proposal of Lateral Force Observer with Active Steering for Electric Vehicle
Yuya Yamauchi1 and Hiroshi Fujimoto1
Department of Electrical and Computer Engineering Yokohama National University 79-5 Tokiwadai, Hodogaya-ku, Yokohama, 240-8501 Japan Phone: +81-4-5339-4262, Fax: +81-4-5338-1157 E-mail: yamauchi@h.dnj.ynu.ac.jp E-mail: hfuji@ynu.ac.jp Abstract: The motion stability of vehicle depends on yaw-rate and side-slip angle. These variables inuence each other, so we applied decoupling control for yaw-rate and side-slip angle for electric vehicle (EV) with active steering and torque difference between left and right in-wheel motors. However the robustness of decoupling control is fragile for road condition. In this paper, we propose lateral force observer (LFO) which is more robust against variation of cornering stiffness. Then we apply LFO and yaw-moment observer (YMO) to our EV with active steering and two inwheel motors. The results of simulations and experiments show that the proposed method achives better performance than our conventional method. Keywords: electric vehicle, yaw-rate control, side-slip angle control, active steering, disturbance obserber
1

1. INTRODUCTION
As a solution of energy and environmental problems, electric vehicles (EVs) is paid to attention. In addition, from the point of view of control engineering, EVs have very attractive potential. Since electric motors and inverters are utilized in drive system, they have great adventages over internal combustion engine vehicles (ICVs)[1]. These advantages are 1) Quick torque response The electric motorss torque response is 100-500 times as fast as that of ICVs. 2) Motor torque can be measured easily In ICVs, it is difcult to acculately measure their output. On the other hand, the output torque electric motor can be measured easliy from current. 3) Individual wheels Electric motors like in-wheel motors are very small. Therefore a motor can be attached to each wheel. These advantages of electric motor enhance vehicle motion control in the EVs. Our research group proposed traction control which prevent wheel slip on slippy road and yaw-rate control which nominalize vehicle yawing dynamics simply for EVs which have in-wheel motors[2]. Electric power steering (EPS) which is operated by the motor also has the advantage of the motor drive. Using this advantage, the new steering system has been investigated so that senior drivers can treat easily[3]. In addition, the rear wheel steer can be used as a control input by using 4WS. The paper [5] reports that the performance limitation of vehicle stability is improved by using 4WS. Moreover, because the degree of freedom in the control system design is increased, the automatic lane following control can be achieved in 4WS vehicle in which the stability is higher than 2WS vehicle[6]. In recent years, the vehicle which is installed four-wheel active steer is made on the market, and it will be thought that the range of use

expands further in the future[7]. In this paper, lateral force observer (LFO) is proposed and applied with yaw-moment observer (YMO) to a small-scale EV with active steering and two in-wheel motors. First, for comparison, decoupling control for yaw-rate and side-slip angle which was proposed by authors in [8][9] is shown as a conventional method. The robustness of decoupling control is fragile for road condition. Second, LFO is proposed based on theory of disturbance observer. The robustness of LFO is much higher than decoupling control against variation of cornering stiffness. Moreover, control method with LFO and YMO is applied to a small-scale EV. Finally, the results of simulations and experiments show that the proposed method achieves better performance than our conventional method.

2. CORNERING FORCE AND VEHICLE DYNAMICS


In this section, the cornering force and vehicle dynamics are shown. Two equations which represent vehicle dynamics are used for plant model in the control system design simulation. 2.1 Cornering force The force in the orthogonal direction to the wheel sideslip angle is called the cornering force. If the wheel side-slip angle is small, the cornering force increases in proportion to the wheel side-slip angle. If the wheel side-slip angle is big, the cornering force has nonlinear characteristic. Here, assuming that the wheel side-slip angle is small, the relation between the cornering force and the wheel side-slip angle can be described as follows Yf = Cf f = Cf lf f V , (1)

Magnitude (dB)

f f

0 -10 -20 -30 -40 -50 -60 0 Phase (deg)

f
Nz

2Yf

f
Nz

V y

lf

-30 -60 -90 0 10

r 2Yr

10 10 Frequency (rad/sec)

10

lr

Fig. 2 Frequency response (yaw-rate)


0 Magnitude (dB) -20 -40 -60 -80 -100 -120 -140 180 135 90 45 0 -45 -90 0 10 f
Nz

Fig. 1 Two wheel model.

f
Nz

Yr

= Cr r = Cr lr , V (2)

Phase (deg)

10 10 Frequency (rad/sec)

10

where Cf and Cr are the cornering stiffness, V is the vehicle speed, is the vehicle side-slip angle, is the vehicle yaw-rate, lf and lr are the distance from body center of gravity to steering knuckle spindle and rear wheel axle. Yf and Yr are the cornering force of front and rear wheels which depend on tire characteristics. 2.2 Vehicle dynamics Fig. 1 is called two-wheel vehicle model. This model is assumed that the roll of vehicle is ignored and the vehicle speed is constant. From two-wheel vehicle model, the linearized dynamics of the lateral motion is derived as
MV d + dt + = 2Yf + 2Yr lf f V 2Cr lr , V (3)

Fig. 3 Frequency response (Side-slip angle). 1 IV . M lf V 2 2llr Cr 1 2llr Cr

Transfer function P2 from the Nz to the vehicle side-slip angle can be written as P2 (s) = G z (0) N G z (0) N = 1 1+
2s n

s2 2 n

(6)

M V 2 + 2(lf Cf lr Cr ) , 4Cf Cr l2 (1 + AV 2 )

= 2Cf

d = 2lf Yf 2lr Yr + Nz I dt lf = 2Cf + f lf + 2Cr V

where n is the character frequency of the vehicle to the steer angle, is the damping ratio. Coefcient of A is called stability factor, which is dened as follows A= M lf Cf lr Cr . 2l2 Cf Cr (7)

lr l r + Nz , V (4)

where M is the vehicle mass, I is the inertia around vehicle yawing axis, Nz is yaw-moment generated by the torque difference between left and right in-wheel motors.

In the same way, transfer function P3 and P4 from the front steering angle and the Nz to the vehicle yaw-rate can be described as P3 (s) = G (0) = G (0) 1 + T s , s2 1 + 2s + 2 n
n

3. CONSIDERATION OF THE CONTROL INPUT


3.1 Transfer function of vehicle motion From the (3) and (4), each transfer function from control input to output can be derived. Transfer function P1 from the front steering angle to the vehicle side-slip angle can be described as 1 + T s , (5) P1 (s) = G (0) s2 1 + 2s + 2 n
n

(8)

V M lf V , , T = l(1 + AV 2 ) 2lCr 1 + TNz s , s2 1 + 2s + 2 n


n

P4 (s) = G z (0) N G z (0) N TNz = =

(9)

G (0)

1 1

M lf V 2 2llr Cr + AV 2

lr , l

V (Cf + Cr ) , 2Cf Cr l2 (1 + AV 2 ) MV . 2(Cf + Cr )

c ______ s+c ref ref + _ C(s) c f + _ G1(s) G2(s) _ Nz Plant + _ Controler + + + _ Plant -1
*

input

Plant

output + _ C(s) in + Nin + + + Nz Plant _ + f

MnVns _____ 2Cfn

ref

+ _

C(s)

Nc +

c _____ s+c

+ _

C(s)

Ins

(a)Decoupling control.

(b)Disturbance obserber. Fig. 4 Block diagram of control system.

c ______ s+c

(c)LFO+YMO. Table. 1 Cut off frequency. f Nz 21.8 rad/s 22.9 rad/s 27.1 rad/s 16.0 rad/s Nz = Nc G2 (s)f , (12)

where c is the control steering angle, Nc is the control moment, G1 (s) and G2 (s) are the decoupling controllers. By substituting (11) and (12) for (10), we obtain = = P1 (s)(c + h ) + (P2 (s) P1 (s)G1 (s))Nz , P4 (s)Nc + (P3 (s) P4 (s)G2 (s))f .

3.2 Investigation of control input to yaw-rate and vehicle side-slip angle It is ideal to control both yaw-rate and the vehicle sideslip angle to stabilize the vehicle motion. In this paper, the front steering angle and the Nz are used as these control inputs. We decided the combination of the control inputs to the plant by the frequency response. However, the front steering angle and the Nz cannot be simply compared because they have a different dimension. So, to normalize each variable, front steering is divided by the maximum front steering angle and the Nz is divided by the maximum value. Fig. 2 and 3 are the frequency responses of the yawrate and the vehicle side-slip angle at 12 km/h. Table. 1 represents the cut off frequency. About only 2rad/s has the difference by the cut off frequency of yaw-rate in the Nz and the front steering angle. As for the cut off frequency of the vehicle side-slip angle, the front steering angle is higher than the Nz by about the twice as high and the gain decrease is also small in high frequency band. Therefore, the front steering angle is selected as the control input of the vehicle side-slip angle, and the Nz is decided as the control input of yaw-rate.

Then the decoupling controller G1 (s) and G2 (s) are given as follows G1 (s) = G2 (s) =
1 P1 (s)P2 (s), 1 P4 (s)P3 (s).

(13) (14)

The inuence of the coupling terms are conceled. Fig. 4(a) is the block diagram of the decoupling control method. 4.2 Yaw-moment observer(YMO) It is very difcult to detect the V directly because we can measure not the vehicle velocity but the wheel speed. Moreover, the lateral forces, Yf and Yr , are difcult to measure directly. The paper[10] assumed that affect of these variables are disturbance moment and proposed compensation algorithm by disturbance observer. By considering disturbance (4) is rewritten as follows I d = 2lf Yf 2lr Yr + Nd + Nz . dt (15)

4. CONTROL METHOD OF VEHICLE MOTION


4.1 Decoupling control method The vehicle side-slip angle and yaw-rate can be described as follows by the use of transfer function P1 (s) P4 (s). P1 (s) P2 (s) f = (10) P3 (s) P4 (s) Nz From this matrix, it is understood that two control input interferes in each of the vehicle side-slip angle and yawrate. This interference inuences the vehicle motion control. Then, the controller which takes into consideration this interference beforehand is designed. Two control inputs are give as f = c + h G1 (s)Nz , (11)

Here Nd is the disturbance moment such as the road condition variation and the affection of crosswind. When excluding the control input Nz dened as Ndt = Nd Nt , (15) is simplied as I d = Ntd + Nz . dt (16)

By using the moment Nz as control input and yaw-rate as measured signal, the disturbance observer can be designed as g. 4(b). The authors named this specic disturbance observer as yaw-moment observer (YMO) in [10]. This YMO can compensate the lumped disturbance and nominalize the system as = 1 Nin . In s (17)

(17) is valid less than the cut-off frequency c .

0.04 0.15 0.03 gamma [rad/s] delta [rad] beta [rad] 0.1

0.2

200 LFO+YMO decoupling

0.15 Nz [Nm] LFO+YMO decoupling 5 0 0 1 2 time [s] 3 4 5

0.02

0.1

200

0.05 LFO+YMO decoupling reference 0 0 1 2 time [s] 3 4 5 0.01 LFO+YMO decoupling reference 1 2 time [s] 3 4 0.05 400

0 0

600 0

2 time [s]

(a)Yaw-rate.
0.2

(b)Side-slip angle. (c)Front steering. Fig. 5 Simulation results (CS with true value).
0.5 0.06 LFO+YMO decoupling reference delta [rad] 0.4 0.3 Nz [Nm] 0.2 0.1 0 0.1 0 5 0 1 2 time [s] 3 4 5 0.2 0 1 2 time [s] 3 4 5 1000 1500 0 1 500 0 500 LFO+YMO decoupling 1500 1000

(D)Nz .
LFO+YMO decoupling

0.15 gamma [rad/s] beta [rad] LFO+YMO decoupling reference 1 2 time [s] 3 4 0.04

0.1

0.02

0.05

0 0

2 time [s]

(a)Yaw-rate.

(b)Side-slip angle. (c)Front steering. Fig. 6 Simulation results (CS with parameter error).

(D)Nz .

4.3 Lateral force observer(LFO) The lateral vehicle motion equation that considers disturbance is described as follows d + = 2Yf + 2Yr + Yd MV dt lf lr = 2Cf + f 2Cr + Yd , V V (18) where Yd is the disturbance lateral force such as road condition variation and affection of crosswind. Here, Ytd is dened as lf lr Ytd = 2Cf + 2Cr V V M V + Yd . Then, (18) is rewritten as M V = 2Cf f + Ytd (19)

placed to -1.0rad/s for nominal plant (17). The cut off frequency of LFO and YMO is set to -10rad/s. In the decoupling control, C (s) is designed by PID controller in which the poles of the closed loop are placed to -6.5rad/s for transfer function P1 (s). C (s) is also designed by PID controller with -5.5rad/s closed-loop pole for transfer function P4 (s). The simulation parameters are set to vehicle speed is 12km/h, reference of vehicle side-slip angle is 0.035rad, and reference of yaw-rate is 0.15rad/s. 5.1 The case of nominal cornering stiffness with true value Cornering stiffness (hereafter, CS) is one of the variable parameters that inuence the vehicle motion. This parameter is frequently changed by the state of the road condition. CS of the plant model is set to front wheel 2270 and rear wheel 6800 in this simulation. These values are assumed to be the true value. Figure. 5 shows the simulation result in which the true value of CS is given to the controllers.. As for the proposed method that used the observers, it is understood that both yaw-rate and the vehicle side-slip angle follow to the reference values with the response of the rst-order system. In the decoupling control method, both yaw-rate and the vehicle side-slip angle follow to the reference values as well as the proposed method. Thus, assuming that CS of controllers are the true value, the proposed method and the decoupling control method achieves the almost same performance. 5.2 The case of cornering stiffness with parameter error Because CS frepuently changes, CS has parameter error in actuality. For this case, the variation was given to the nominal value used in controller design to match the experimental conditions although the plant parameter is

By using the front steering f as control input and as measured signal, the disturbance observer can be designed as g. 4(b). The authors named this specic disturbance observer as lateral force observer (LFO). This LFO can nominalize the system as 2Cf n = (20) in . Mn Vn s

5. VERIFICATION BY SIMULATION
The decoupling control method shown in g. 4(a) and the proposed control method combined YMO and LFO shown in g. 4(c) are compared by simulation. In the proposed method, the controller C (s) is simply desighned by propotional control in which pole of the closed loop is placed to -2.0rad/s for nominal plant (20). The controller C (s) is also designed by proportional controller in which the pole of the closed loop is

acceleration a x ,a y sensor Motor for SBW yaw-rate sensor steering angle sensor

,Treal Li-ion Battery 15V 10 300V L inverter Chopper R inverter

R motor

Controller AUTOBOX DS1103

) 150V

Motor for SBW

T*

L motor

Fig. 7 FPEV2-kanon.

Fig. 8 Conguration of experimental system. usually given to the variation. In this simulation, front and rear CS in controller is set to 1890, 4760, respectively. Fig. 6 shows the simulation result. Because the decoupling controllers (13) and (14) needs exact values of CS, the response has terrible oscillation by the CS variation. On the other hand, the proposed method has almost same response with g. 5 even if the values of CS have big variation. From this results, it can be said that the robustness is improved to against the variation of CS by using LFO and YMO.

6.2 Experiment result In this paper, experiment vehicle runs on nonpavement road where the grip is lower than the asphalt road at V =12 km/h. The reference values are same with the simulations. The vehicle side-slip angle is measured with optical sensor CORREVIT S-400 of the DATRON Co., Ltd. Fig. 9 is the experiment result only with LFO. The vehicle side-slip angle follows to the reference value 0.035rad with the rst-order system response. As a result, it is able to be conrmed that the dynamics of sideslip angle is nominalized using LFO in experiment. However, yaw-rate has error to the reference value 0.15rad/s only by LFO because it is not controlled. On the other hand, g. 10 shows the result only with YMO. The vehicle side-slip angle has steady-state error although yaw-rate follows to the reference value. Therefore, it is understood that YMO can control only yawrate. Then, the proposed method that is applied both LFO and YMO is experimented. This system control the vehicle side-slip angle and yaw-rate at the same time. Fig. 11 shows the experiment result. Thus, it can be conrmed that the vehicle side-slip angle and the yaw-rate can follow to the reference value by applying LFO and YMO with the rst-order system response. Finally, the proposed method is compared to the decoupling control method by experiment. In this experiment, front wheel CS of controller is 1890, rear wheel of CS is 4760. Fig. 12 shows the result. Though yaw-rate and side-slip angle with the decoupling control method are oscillated, with the proposed method they follows the reference value. By this result, the proposed method achieves better performance than the decoupling control method.

6. EXPERIMENT
6.1 Experiment vehicle An original electric vehicle FPEV2-Kanon developed in our laboratory is used for the experiment vehicle. As for the main feature of the experiment vehicle, the following four points are enumerated.

The outer rotor type in-wheel motors are installed. Active front and rear steering system is implemented by electric motor. The Li-ion battery is used for the battery. AUTOBOX is used for the vehicle controller.

The outer rotor type in-wheel motor made by Toyo Denki Seizo K.K., Ltd. is installed in two rear wheels as driving power. The maximam torque of this in-wheel motor is 340Nm. Because this motor is the direct drive system, the reaction force from the road is directly transfered to the motor side without gear reduction and backlash. The steering wheel mechanism adopts Steer-by-Wire (SbW) and active front and rear steering system, by using two 250W DC motors for EPS. Moreover, in order to switch SbW and EPS, the steering wheel shaft has a removable structure. The Li-ion battery that is able to charge in a short time is used for the energy storage. In this vehicle, ten batteries of 15V per module are installed. In addition, the battery voltage 150V boosts to 300V by using chopper circuit, and it is fed to power to the inverter. AUTOBOX-DS1103 of the dSPACE Co., Ltd. is used for the vehicle control. Fig. 7 shows the experiment vehicle, and g. 8 shows the conguration of experimental system.

7. CONCLUSION
Lateral force observer (LFO) based on the disturbance observer was developed in this paper, the control method combined with yaw moment observer (YMO) was proposal. This proposed method and the decoupling control method were compared by the simulations. It was shown that robustness improved by using the proposed method.

0.04

0.25 0.2 gamma[rad/s] beta[rad] 0.15 0.1 0.05 experiment reference 0 0 2 4 time [s] 6 8 10

0.05 0.04 0.03 0.02 0.01 experiment reference 0 0 2 4 time [s] 6 8 10 gamma[rad/s]

0.2

0.03 beta[rad]

0.15

0.02

0.1

0.01 experiment reference 0 0 2 4 time [s] 6 8 10

0.05 experiment reference 0 0 2 4 time [s] 6 8 10

(a)Side-slip angle. (b)Yaw-rate. Fig. 9 Experiment results (LFO only).


0.2 0.04 0.15 beta [rad] 0.03 beta[rad] 0.04 0.03 0.02 0.05 experiment reference 0 0 2 4 time [s] 6 8 10 0 0 2 4 time [s] 6 experiment reference 8 10 0.01 0 0 0.06 0.05

(a)Side-slip angle. (b)Yaw-rate. Fig. 10 Experiment results (YMO only).


0.25 gamma [rad/s] LFO+YMO decoupling reference 2 4 time [s] 6 8 10 0.2 0.15 0.1 0.05 0 0 2 4 time [s] 6 LFO+YMO decoupling reference 8 10

gamma[rad/s]

0.02

0.1

0.01

(a)Side-slip angle. (b)Yaw-rate. Fig. 11 Experiment results (LFO+YMO). Moreover, we conrmed that the proposed method is able to be applied to real vehicle by the experiment, and showed the effectiveness. The future work is to improve the response speed of the proposed method.

(a)Side-slip angle. (b)Yaw-rate. Fig. 12 Experiment results (comparison). JSAE Annual Congress, No. 1107, pp. 16, 2007 (in Japanese) [6] P. Raksincharoensak, H. Mouri, M. Nagai: Investigation of Automatic Lane-Keeping Control of Four-Wheel-Steering Vehicle(Theoretical Discussions on Lateral and Yaw Dynamics), Transactions of the Japan Society of Mechanical Engineers. C, vol. 69, No. 681, pp. 12541259, 2003 (in Japanese) [7] T. Katayama, Y. Yasuno, T. Oida, M. Sao, M. Imamura, N. Seki, Y. Sato: Development of 4 Wheel Active Steer, JSAE Annual Congress, No. 1107, pp. 712, 2007 (in Japanese) [8] H. Nagase, T. Inoue, Y. Hori: Decoupling Control of and for high performance AFS and DYC of 4 Wheel Motered Electric Vehicle, in Proceedings of the International Symposium on Advanced Vehicle Control(AVEC), Hiroshima, Japan, 2002 [9] Y. Yamauchi, H. Fujimoto: Fundamental Experiment on Yaw-rate and Body Slip Angle Control with Active Steering for Electric Vehicle, The 16th Transportation and Logistics Conference 2007, No. 0751, pp. 139142, 2007 (in Japanese) [10] H. Fujimoto, T. Saito, T. Noguchi: Motion Stabilization Control of Electric Vehicle under Snowy Conditions Based on Yaw-Moment Observer, in Proc. 8th IEEE International Workshop on Advanced Motion Control (AMC04) , pp. 3540, Kawasaki, 2004.

ACKNOWLEDGMENT
This study was partly supported by Industrial Technology Research Grant Program from New Energy and Industrial Technology Development Organization (NEDO) of Japan.

REFERENCES
[1] Y. Hori: Future Vehicle Driven by Electricity and Control-Research on Four-Wheel-Motored: UOT Electric March II, IEEE Trans. IE, Vol. 51, No. 5. 2004 H. Fujimoto, K. Fujii. N. Takahashi: Motion Control and Road Condition Estimation for Electric Vehicle, in Proc. The 22nd International Battery, Hybrid and Fuel Cell Electric Vehicle Symposium & Exposition, Yokohama, pp. 18061817, 2006. M. Udagawa, J. Ishino, Y. Knou, M. Abe: Development of planetary gear system with power assist using torque split control of traction motors and its application to aged drivers, JSAE Annual Congress, No. 1107, pp. 2328, 2007 (in Japanese) M. Kubota, H. Mouri, N. Horiguchi: An investigation on the effect of transient assistant torque input on steering effort, JSAE Annual Congress, No. 7505, pp. 914, 2005 (in Japanese) E. Ono, Y. Hattori, S. Monzaki: Improvement in Safety and Comfort by Active Four Wheel Steering-Improvement in Critical Performance by Optimal Distribution of Four Wheel Tire Forces-,

[2]

[3]

[4]

[5]

You might also like