Development of a Washout Algorithm for a Vehicle ...

Proceedings of ACMD 2004

Development of a Washout Algorithm for a Vehicle Driving Simulator Using New Tilt Coordination and Return Mode

Ki Sung You Graduate school of Mechanical and Intelligent Systems Engineering, Pusan National University San, 30, Jangjeon-dong, Kumjeong-gu, Busan, 609735, KOREA E-Mail: [email protected] Eugene Kang Advanced Engineering Center, Samsung SDI 575, Sin-dong, Paldal-gu, Suwon-si, Gyeonggi-do, 442-731, KOREA E-Mail: [email protected]

ABSTRACT A vehicle-driving simulator is a virtual reality device which makes a man feel as if he drove an actual vehicle. Unlike actual vehicles, the simulator has limited kinematical workspace and bounded dynamic characteristics. So it is difficult to simulate dynamic motions of a multi-body vehicle model. In order to overcome these problems, a washout algorithm, which controls the workspace of the simulator within the kinematic limitation, is needed. However, a classical washout algorithm contains several problems such as generation of wrong sensation of motions by filters in tilt coordination, requirement of trial and error method in selecting the proper cut-off frequencies and difficulty in returning the simulator to its origin using only high pass filters. This paper proposes a washout algorithm with new tilt coordination method which gives more accurate sensations to drivers. To reduce the time in returning the simulator to its origin, an algorithm that applies selectively onset mode from high pass filters and return mode from error functions is proposed. As a result of this study, the results of the proposed algorithm are compared with the results of classical washout algorithm through the human perception models. Also, the performance of the suggested algorithm is evaluated by using human perception and sensibility of some drivers through experiments. Key Words: vehicle driving simulator, washout algorithm, new tilt coordination, return mode

Min Cheol Lee School of Mechanical Engineering, Pusan National University San, 30, Jangjeon-dong, Kumjeong-gu, Busan, 609735, KOREA E-Mail: [email protected] Wan Suk Yoo School of Mechanical Engineering, Pusan National University San, 30, Jangjeon-dong, Kumjeong-gu, Busan, 609735, KOREA E-Mail: [email protected] which drivers feel as if they drive a real car. In order to reappear reality of driving situation, we analyze vehicle motion generated by operating a steering wheel and an accelerate pedal, and provide the motion situation such as the sense of motion, sight and hearing to driver according to the analysis. The functions make the efficiency of a vehicle-driving simulator high (Drosdol and Panik, 1985). A real vehicle has no limitation of motion range while a motion system which copies the real motion has a limitation to reappear the sensation of motion because it has a limitation in kinematical motion range and dynamic characteristic of the system. Therefore we need a washout algorithm which limits motion range of a vehicle within a physical limitation of the simulator (Freeman, et. al., 1995). A representative washout algorithm is a classical washout algorithm. The washout algorithm has advantages such as short analysis time and verified stability, while there are several problems: such that it is difficult to select cut-off frequency of used filters, the original signal is falsified on tilt coordination and there is a limitation to reappear a continuous motion and as only restoration by filter (Greenberg and Park, 1994). In order to overcome problem and a limitation of classical washout algorithm, we develop new tilt coordination algorithm and washout algorithm considering a return component. And we compare the proposed algorithm with classical washout algorithm quantitatively using human perception model.

2. CLASSICAL WASHOUT ALGORITHM 2.1 Human Perception Model

1. INTRODUCTION Vehicle driving simulator is a virtual reality device,

Human feels translation motion by detecting specific force in otolith and rotational motion by detecting

Copyright ⓒ2004 by KSME - 112 -

Proceedings of ACMD 2004

rotational angular acceleration in vestibular. The specific force detected in otolith is obtained by subtracting acceleration of gravity from translational acceleration:

uur r ur Fs = a − g

(1)

ur r where a is translational acceleration, g is acceleration of gravity. Young and Oman have proposed that human sensory organ is regard as spring, mass and damper system. The transfer function of otolith system is shown in Equation (2) (Peter and Burnell, 1981). The characteristic of frequency response of otolith system has good perception in the band of 0.01~0.5Hz.

G ( s )otolith =

k (τ A s + 1) (τ L s + 1)(τ S s + 1)

(2)

The transfer function of the modeled vestibular system is shown in equation (3) and the characteristic of frequency response has good perception in the band of 0.01~0.5Hz.

G ( s )vestibular

TLTA s 2 = (TL s + 1)(TS s + 1)(TA s + 1)

Translational acceleration of vehicle

(3)

2.2 Classical Washout Algorithm Classical washout algorithm is mainly composed of three parts. One is translational washout, another is rotational washout, and the other is tilt coordination. And the layout of the algorithm is shown in Figure 1. The translational acceleration and rotational acceleration of a car derived from analysis result of vehicle dynamics are inputted to a washout algorithm, then the inputted signal is outputted to a transformed translational acceleration and rotational acceleration which can be reappeared in simulator through washout algorithm. Translational motion washout algorithm consists of high pass filter separated into two steps and coordinate transformation. The algorithm removes a low frequency element using the first step high pass filter and generates high frequency acceleration on the center of an upper plate of motion manipulator. The acceleration on the center of the upper plate is converted to the acceleration of the center coordinate system of the under plate through coordinate transformation. We cannot reappear the motion within the kinematical limitation of the motion manipulator with only the first step high pass filter because there is incidental specific force according to Bryant angle of motion manipulator in coordinate transformation process. In order to solve the problem, we generate high frequency translational acceleration on the center of the under plate of the motion manipulator using the second step high pass filter.

Translational washout algorithm 1st step high pass filter

Transform of coordinate

2nd step high pass filter

Translational acceleration of simulator

Tilt coordination algorithm Low pass filter

Angular velocity of vehicle

Tilt coordination

1st step high pass filter

+ +

Transform of coordinate

2nd step high pass filter

Angular velocity of simulator

Rotational washout algorithm

Figure 1. SCHEMATIC DIAGRAM OF CLASSICAL WASHOUT ALGORITHM

The rotational motion washout algorithm consists of high pass filter of 2 steps and coordinate transformation. The algorithm removes a continuous low frequency rotational angular acceleration inputted to coordinate system of a car seat by using the first step high pass filter. The algorithm is converted to rotational angular acceleration which is represented to Bryant angle by coordinates transformation. The desired rotational angular acceleration in high frequency is generated by passing through the 2nd high pass filter, where the generated acceleration can return to the origin while the acceleration can restricts a gimbals angle of motion manipulators (Reid and Nahon,1988). We cannot generate a low frequency specific force with only washout algorithm of translational motion because this algorithm only generates high frequency specific force using the high pass filter. However, if a tilt coordination algorithm is used, the specific force that is removed in the translational motion washout algorithm can be generated as tilt angle to follow rotational motion. As driver feels a specific force in the opposite direction to the amount of gravity, we can generate a desired specific force by tilting a motion manipulator in the opposite direction to the gravity vector. The classical washout algorithm has advantages of fast computation, stable performance and etc. However this algorithm has two problems. First, the phase and magnitude of the desired reappearing specific force in frequency domain are distorted from the ones of the specific force in a vehicle dynamic with a high and low frequency because the desired filter in tilt coordination are not an ideal filter. Second, the classical washout algorithm generates high frequency motion which can reappear with limitation

Copyright ⓒ2004 by KSME - 113 -

Proceedings of ACMD 2004

Translational aP acceleration

Translational specific force transform

FPtS

wP

Rotational specific force transform

S FPr

Rotational angular velocity

+

FPS

+

wP

Human sensation model otolith system

Translational ap accelerations

Human sensation model vestibular system

1st step high pass filter

of kinematical motion using high pass filter. However, the signal, which passed through these high pass filters, generates motion in the opposite direction to the movement of original signal according to cut off frequencies. Since then the motion that is generated by high pass filter return to the origin, the return motion remained after a regular time. Therefore continuously generated motion can’t be reappeared satisfactorily because there is a limitation to reduce the time which it takes to return to the motional origin completely only force of restitution by high pass filter.

FOR

3.1 New Tilt Coordination Algorithm For translational acceleration and rotational angular acceleration which are analyzed from car dynamics, the motional perception which is recognized from a human perception model without signal processing such as washout algorithm is a motion perception in a vehicle. A schematic diagram for sensations of motion in actual vehicle is shown in Figure 2. The specific force in equation (1) is translated to coordinate system of the center of seat so that we can get equation (4)  α Px + g sin β  uur uur ur FPs = aP − GP R(α , β , λ ) g =  aPx − g sin α sin β   aPz + g cos α cos β 

F Spa,High

Low pass filter +

+

ap,Low

+

Rotational wp angular velocity

wr,Low

MOTIONS IN ACTUAL VEHICLE

ALGORITHM

Translational specific force transform

S F pa,High

Human sensation model otolith system

Tilt coordination

Figure 2. SCHEMATIC DIAGRAM FOR SENSATIONS OF

3. NEW WASHOUT RETURN MODE

a p,High

(4)

+

+

Specific force transform

FpwS

+ FSpw,Low

wp+ wr,Low

Human sensation model vestivular system

Figure 3. SCHEMATIC DIAGRAM FOR SENSATIONS OF MOTIONS BY CLASSICAL TILT COORDINATION

uuur ur rotational motion FPrs for − GP R(α , β , λ ) g . Then specific

force in coordinate system of seat is represented such as equation (5). uur uuur uuur FPs = FPts + FPrs

(5)

A schematic diagram for sensations of motions by classical tilt coordination is shown in Figure 3. In order to remove continuous translational acceleration using high pass filter and reappear a specific force, we generate low frequency translational acceleration by passing same translational acceleration through low pass filter and then generate specific force through tilt coordination. The generated specific force using classical tilt coordination is expressed as uuuuuuuuur uuuuuuur uuuuuuur uuur FPs,Classical = FPts , High + sFPr,s Low + FPrs , (6) uuuuuuur where FPts , High is a specific force which is generated by

high frequency translational acceleration passing through uuuuuur high pass filter and FPr,s Low is a specific force which is

generated after doing tilt coordination of low frequency

uur where aP is translational acceleration in the coordinate ur system of the seat, − GP R(α , β , λ ) g is acceleration of

gravity in the coordinate system of the seat. We substitute specific force by translational motion uuur uur FPts for motion aP and specific force according to

translational acceleration passing through low pass filter. However, because real high pass filter and low pass filter are not ideal filers, the sum of the specific force which is applied to low and high pass filter gives rise to distort the original one. Figure 4 represents a schematic diagram for sensations of motions by a new tilt coordination algorithm. In this paper, we propose a new tilt coordination algorithm which

Copyright ⓒ2004 by KSME - 114 -

Proceedings of ACMD 2004

1st step high pass filter

+

a P,High

Translational specific force transform

FSPt,High 16

-

+

a P,Diff

+

S

FP,New

Human sensation model otolith system

Tilt coordination Rotational wP angular velocity

wP,Diff

+

+

Rotational specific force transform

FSPr

S + FPr,Diff

wP+ wP,Diff

Human sensation model vestibular system

Yaw directional angular velocity (deg/sec)

Translational aP acceleration

Without washout Classical washout

14 12 10 8 6 4 2 0 -2 -4 -6 -8 0

20

40

60

80

100

Time (sec)

Figure 4. SCHEMATIC DIAGRAM FOR SENSATIONS OF MOTIONS BY NEW TILT COORDINATION

coordination, is defined as uuuuuur uur uuuuuur aP , Diff = aP − aP , High

(7)

uur where aP is a translational acceleration of the seat which uuuuuur is acquired from analysis result of car and aP , High is

translational acceleration passing through high pass filter. The specific force generated by translational acceleration is defined as uuur uuuuuuur uuuuuur FPts = FPts , High + FPr,s Diff

Figure 5. YAW DIRECTIONAL ANGULAR VELOCITY THROUGH HIGH PASS FILTER

120

Without washout Classical washout

100

Yaw directional angle (deg)

uses the difference between the translational acceleration acquired from car dynamics and filtered acceleration passing through high pass filter. In order to generate additional specific force, uuuuuur acceleration component ( aP , Diff ), which is inputted to tilt

80 60 40 20 0 -20 0

20

40

60

80

100

Time (sec)

Figure 6. YAW DIRECTIONAL ANGLE THROUGH HIGH PASS

(8)

FILTER

uuur where FPts is the generated specific force by translational uuuuuuur acceleration, FPts , High is the generated specific force by

translational acceleration passing through high pass filter uuuuuur and FPr,s Diff is the generated specific force by proposed tilt

coordination algorithm. The specific force which a driver on seat recognizes is represented sum of the specific force by a high pass filter, tilt coordination and rotational angular velocity. And the specific force which is generated by the proposed algorithm is defined as uuuuuur uuuuuuur uuuuuur uuur uuur uuur uur FPs, New = FPts , High + FPr,s Diff + FPrs = FPts + FPrs = FPs (9)

The specific force is defined as the same force with the one being recognized in a real car.

3.2 Washout Algorithm Considering Return Mode

3.2.1 Return of the Motional Classical Washout Algorithm

Origin

in

In classical washout algorithm, we generate a high frequency motion that can reappear within limitation of kinematical motion using high pass filter. After the motion which is inputted to high pass filter have been returned to the origin of motion, the filter output motion is returned to the origin after taking a regular time. Figure 5 and Figure 6

Copyright ⓒ2004 by KSME - 115 -

Proceedings of ACMD 2004

Dynamic analysis results

Dynamic analysis results

Mode selector

High pass filter

Onset motion

Switching function Error function

Low pass filter

Difference function

Absolute function

Integrators

Mode selector

Motion cues Mode analysis

Return motion

Washout filter

Onset motion

Error function

Return motion

Switching function

Figure 7. SCHEMATIC DIAGRAM OF WASHOUT ALGORITHM USING RETURN MODE

show a yaw directional angular velocity and a yaw directional angle using high pass filter. Cut off frequency of the high pass filter is 0.01Hz. A solid line is a rotational angular velocity and angle of a car when driver turns left. A dotted line is a rotational angular velocity and the angle that are transformed by high pass filter. The angular velocity, which is opposite to motional direction, is generated as a low frequency component is removed by high pass filter such as Figure 5. A rotational angle of a car without washout preserves 90 degrees after left turn. However a rotational angle from classical washout returns to 0 degree of motional origin after increasing to 60 degree. That is, a rotational motion is on onset mode to peak angle. Then return mode starts. The motion rotates toward opposite motional direction and slowly returns to origin. If a continuous motion generates before completely returning to the origin or in moving on an opposite direction of the motion, motion manipulator may deviate from position limitation. Because various motions continuously generate in a real car driving situation, the return of the motional origin is done as fast as possible and the opposite motion has to be removed.

3.2.2 Washout Algorithm Considering Return Mode In this study, we proposed the algorithm that compensates return mode by using an error from the motional origin to overcome limitation of returning time to the motional origin in classical washout algorithm. Figure 7 shows a schematic diagram of washout algorithm using return mode. Washout algorithm using return mode consists of high pass filter, error function, mode selector, switching function and integrator. An analytical result of vehicle dynamics is inputted to the proposed algorithm. After then motion cues that can generate within motional origin is outputted. A low frequency element is removed by using high pass filter and a high frequency element passing through the filter is selected as a beginning motion which moves from the motional origin to a peak displacement and rotational angle. We generate a return motion using

Integrators

Figure 8. SCHEMATIC DIAGRAM OF MODE SELECTOR

integrator, error function and motional cue, which can be implemented in simulator and always, return to the origin. Figure 8 shows a schematic diagram of mode selector that analyzes motion mode and controls a switching function. Mode selector consists of low pass filter to remove a noise, absolute function to consider return element in bidirectional motion, difference function to find time variable rate, and mode analyzer which analyzes a preprocessed signal and determines motion mode. We select a motion from motion origin to a peak angle is selected as a beginning mode and a motion from the peak angle to the motion origin is selected a return mode as shown in Figure 6. However we use the result of vehicle dynamic analysis in analyzing motion mode because there is a problem such as a phase lead when we use a signal passing through high pass filter in analyzing mode.

4. QUANTITATIVE ESTIMATION THROUGH SIMULATION 4.1 Virtual Driving Circumstance Virtual driving circumstance is composed of a scenario, which drives an asphalt road in the city. In order to estimate reappearance of sensations of motions variously, a road map includes general road conditions such as a straight road, curve, a change of lane, a bumper passage, a variable speed range and an uneven road as shown in Figure 9.

4.2 Quantitative Estimation for Sensations of Motion By New Tilt Coordination Figures 10 ~ 15 show sensations of motions in virtual driving circumstance. We select sensation of translational motion of surge and sway direction and sensation of rotational motion of roll and pitch direction, because heave and yaw directional element don’t have effect on tilt

Copyright ⓒ2004 by KSME - 116 -

Proceedings of ACMD 2004

Surge motion sensation error (m/sec)

0.4

Classical tilt coordination New tilt coordination

0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 0

10

20

30

40

50

60

70

80

Time (sec)

Figure 9. ROAD MAP FOR SIMULATOR Figure 10. SURGE MOTION SENSATION

0.4

Without washout Classical tilt coordination New tilt coordination

0.3

Sensation of surge motion (m/sec)

coordination. Cut off frequency of high pass filter and low pass filter is set to be 0.01Hz. In the figures which represent all directional sensation of motion, the solid line is sensation of motion which is response of human sensation model about vehicle analysis without passing through wash out algorithm, a long dotted line is sensation of motion in simulator by classical tilt coordination, and a short dotted line is a sensation of motion by new tilt coordination. Error of sensation of motion is defined as a difference of directional sensation in vehicle in simulator. Figures 10 ~ 11 show translational sensation of motion of surge direction and its error. Sensation of motion by classical tilt coordination shown in Figure 10 can be considered that it is falsified by characteristics of magnitude and phase compared with sensation of motion by new tilt coordination. In Figure 11, error of sensation of motion by classical tilt coordination is represented within about ±0.1m/sec of 30% of total sensation of motion of surge direction, and its error by new tilt coordination represents within about ±0.02m/sec. Figures 12 ~ 13 represent translational sensation of motion of sway direction and its error. Same as result of surge direction, characteristics of magnitude and phase by new tilt coordination are less falsified. And error of sensation of motion is decreased within ±0.02m/sec range. Figures 14 ~ 15 represent rotational sensation of motion of roll and pitch direction. Additional translational sensation of motion is generated. And its magnitude is ±0.4 degrees compared with that without washout algorithm. However, considering a maximum magnitude of ±10 degrees of pitch directional sensation of motion, we can assume that additional rotational sensation of motion doesn’t have an effect on total rotational sensation of motion

0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 0

10

20

30

40

50

60

70

80

Time (sec)

Figure 11. SURGE DIRECTIONAL SENSATION ERROR

4.3 Quantitative Estimation for Algorithm Considering Return Mode From Figures 16 ~ 19 represent sensation of motion by human perception model about rotational angular velocity of yaw direction when left turn and rotational angle which is a value of the first integral of rotational angular acceleration in driving seat’s center. A solid line is a result without washout algorithm, a long dotted line is a

Copyright ⓒ2004 by KSME - 117 -

Proceedings of ACMD 2004

0.4

Without washout Classical tilt coordination New tilt coordination

Without washout Classical tilt coordination New tilt coordination

10

Sensation of pitch motion (deg)

Sensation of sway motion (m/sec)

0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3

5

0

-5

-10

-0.4 0

10

20

30

40

50

60

70

0

80

10

20

30

40

50

60

70

80

Time (sec)

Time (sec)

Figure 15. PITCH DIRECTIONAL SENSATION

Figure 12. SWAY DIRECTIONAL SENSATION

Without washout Classical washout Proposed washout

120

Classical tilt coordination New tilt coordination

0.3

100

0.2

Angle of yaw motion (deg)

Sway motion sensation error (m/sec)

0.4

0.1 0.0 -0.1 -0.2

80 60 40 20 0

-0.3 -20

-0.4 0

10

20

30

40

50

60

70

0

80

5

10

15

25

30

35

Figure 16. YAW MOTION ANGLE

Figure 13. SWAY DIRECTIONAL SENSATION ERROR

12

1.2

Without washout Classical tilt coordination New tilt coordination

0.8

Without washout Classical washout Proposed washout

10 8

Sensation of yaw motion (deg)

1.0

Sensation of roll motion (deg)

20

Time (sec)

Time (sec)

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8

6 4 2 0 -2 -4 -6 -8 -10

-1.0

-12

-1.2 0

10

20

30

40

50

60

70

0

80

5

10

15

20

25

30

35

Time (sec)

Time (sec)

Figure 14. ROLL DIRECTIONAL SENSATION

Figure 17. YAW MOTION SENSATION

Copyright ⓒ2004 by KSME - 118 -

Proceedings of ACMD 2004

Without washout Classical washout Proposed washout

120

Angle of yaw motion(deg)

100 80 60 40 20 0 -20 0

5

10

15

20

25

30

35

Time(sec)

Figure 18. YAW DIRECTIONAL ANGLE

12 10

Without washout Classical washout Proposed washout

8

Sensation of yaw motion(deg)

6 4

Figure 20. ROAD MAP FOR EXPERIMENTS

2 0 -2

and returns to the motional origin completely about 18 seconds. However, shown in Figure 19, we expect that there generates a big bouncing part and driver feels wrong sensation of motion.

-4 -6 -8 -10 -12 -14 -16 -18 0

5

10

15

20

25

30

35

5. REAL TIME DRIVING AND QUALITATIVE ESTIMATION

Time(sec)

Figure 19. YAW DIRECTIONAL SENSATION

a result in the classical washout and a short dotted line is a analysis result in the proposed algorithm. And cut off frequency of high pass filter is selected by 0.02Hz. The results which remove opposite directional motion generated by high pass filter using proposed algorithm and reduce a final return time to the motional origin are shown in figures 16 ~ 17. A rotational angle by classical washout algorithm shown in Figure 16 increases to about 40 degrees after starting motion then the first return occurs. After that, opposite rotation is generated. Compared with that, rotational angle by proposed algorithm returns to the motional origin without opposite direction. In general rotational sensation of motion which driver recognizes shown in Figure 17 is represented similarly in general though phase lead is generated a little in return motion. Figures 18 ~ 19 are results when a return time to the motional origin by using proposed algorithm is more reduced. Shown in Figure 18, the rotational angle by proposed algorithm removes opposite directional rotation

5.1 Survey Plan In order to estimate validity of washout algorithm and performance of vehicle driving simulator developed by Pusan National University (Park, et. al., 2001), it is important not only quantitative estimation using human perception model but also qualitative estimation of reappearance of sensation of motion of driver. In order to estimate validity of proposed washout algorithm and investigate the influence of sensation of motion by washout algorithm and each module, we prepare a standard of qualitative estimation by dividing a question narrowly. Survey consists of estimation by driving course and motional component. Estimation by driving course consists of comparing questions of feeling of a real car driving with sensation of motion which is embodied in Pusan National University’s driving situation such as a straight section, a left turn section, a bended road, a slope section and passing through an uneven road. In order to estimate according to motional direction, it consists of sensation of motion and feeling about each 6 Freedom of degree.

Copyright ⓒ2004 by KSME - 119 -

Proceedings of ACMD 2004

Table 1. AVERAGE VALUES OF MOTIONS AND FEELINGS AS

Table 2. AVERAGE VALUES OF MOTIONS AND FEELINGS AS

DRIVING COURSES

EACH DIRECTION

Estimation Item

Motion Sensation

Feeling

Estimation Item

Motion Sensation

Feeling

Straight Section

2.72

2.72

Surge Direction

2.92

2.96

Left Turn Section

2.44

2.56

Sway Direction

2.74

2.82

Bended Road

2.56

2.56

Heave Direction

3.37

3.46

Slope Section

2.40

2.56

Roll Direction

2.71

2.78

Uneven Road

3.36

3.28

Pitch Direction

2.88

2.76

Yaw Direction

2.88

2.6

6. CONCLUSION

5.2 Driving Experiment In order to estimate vehicle simulator qualitatively, we drive virtual situation of Pusan National University. Figure 20 shows driving path of car according to a scenario. A driving scenario is organized to experience a left turn, a straight section, a bended road and a climbing section from bumper 1 to bumper 8. It starts 2m back of bumper 1 and driver drives subjectively with 0~35km/h. Then it stops in front of bumper 8.

5.3 Estimation Result of Survey 5.3.1 Motion Sensation And Feeling For Each Driving Course The result of survey which compares motion sensation and feeling for 6 DOF motion of simulator with those of a real car measures such as ‘very difference (0), a little difference (1), normal (2), similarity (3), equality (4) and very equality (5)’ and evaluates the numerical mean. Then we arrange them such as Table 2. Estimation for each item represents a value of ‘similarity(3)’ and is estimated that motion sensation of surge and heave direction is similar to a real car. Generally Table 1 shows that average value of estimation for motion sensation as motion direction is lager than average value of motion sensation for driving course and handling. If we consider that washout algorithm is the most influential element for estimation as motion direction, proposed washout algorithm provides drivers with translational and rotational motion sensation which is similar to a real car. When distribution of drivers is represented to percentage of total persons, there is evaluation of ‘very difference’ (4%). However over 50% persons response positive evaluation for each item.

In this paper, we develop washout algorithm which can operate vehicle motion within limitation of motional range. A new tilt coordination algorithm is proposed for compensating the falsified specific force that is generated by tilt coordination of classical washout algorithm, which can reappear specific force more completely. And we analysis a problem which is generated on producing return component using only high pass filter in classical washout algorithm. In order to improve the problem, an algorithm of considering return component is proposed. And it is verified validity of proposed algorithm. A quantitative estimation is performed through quantitative estimation of simulation, experiment and survey. Tilt coordination algorithm by quantitative estimation generates somewhat additional rotational sensation of motion of roll and pitch direction compared with classical tilt coordination. However, we can confirm that it reappears well sensation of motion which feels in a real car. Besides proposed algorithm considered return component returns faster than general washout algorithm, it removes opposite direction of motion to original signal’s motional direction. We confirm that it can embody motion which always returns to the origin and is continuous. In qualitative estimation, sensation of driving which drivers in simulator recognize is represented over ‘normality’ compared with a real car, and sensation of motion and feeling according to motional direction which has large effect on washout algorithm. Proposed algorithm doesn’t reappear of wrong sensation of motion to driver in simulator and provides similar driving situation with a real car. In future work, we will have to improve algorithm which consider not only return component by including human sensation mode to proposed algorithm but also sensation of motion recognized by driver when develop algorithm with return component.

Copyright ⓒ2004 by KSME - 120 -

Proceedings of ACMD 2004

7. ACKNOWLEDGEMENTS The authors would like to thank the Ministry of Science and Technology of Korea for financial support in the form of grant (M1-0203-00-0017-02J0000-00910) under the NRL (National Research Laboratory)

REFERENCES J. Drosdol and F. Panik, "The Daimler-Benz Driving Simulator, A Tool for Vehicle Develoment," SAE paper 850334, 1985. J. S. Freeman, G. Watson, T. E. Papelis, T. C. Lin, A. Tayyab, and R. A. Romano, "The Iowa Driving Simulator: An Implementation and Application Overview," SAE paper 950174, 1995. J. A. Greenberg and T. J. Park, "The Ford Driving Simulator," SAE paper 940179, 1994. T. Suetomi, A. Horiguchi, T. Okamoto and S. Hata "The Driving Simulator with Large Amplitude Motion System," SAE Paper 910113. 1991. C. C. Peter, and T. M. Burnell, "Analysis Procedures and Subjective Flight Results of a Simulator Validation and Cue Fidelity Experiment," NASA Technical Memorandum 88270, 1981. L. D. Reid and M. A. Nahon, "Response of Airline Pilots to Variations on Flight Simulator Motion Algorithm," AIAA, Vol. 1, pp. 639-646, 1988. M. K. Park, M. C. Lee, K. S. Yoo, K. Son, W. S. Yoo and M. C. Han, "Development of the PNU Driving Simulator and Its erformance Evaluation," Proceedings of the IEEE International Conference on Robotics & Automation, Vol. 3, pp. 2325-2330, 2001.

Copyright ⓒ2004 by KSME - 121 -