AUTOMATIC HEADLAMP SWITCHING SYSTEM USING

ACCELEROMETERS

Kai-Chi Chan and Yiu-Sang Moon

Department of Computer Science and Engineering, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong

Keywords:

Automobile, Night vision, Accelerometer, Alpha beta ﬁlter.

Abstract:

This paper presents a two-sensor method to enhance the nighttime driving safety. It consists of two accelerom-

eters and an array of auxiliary swiveling headlamps. An alpha beta ﬁlter is proposed to stabilize the readings of

the accelerometers. With the kinematics of a turning car, the cars turning path is predicted based on the steer-

ing angle measured by the accelerometers so that the relevant auxiliary swiveling headlamps will be switched

on accordingly. In this paper, we will study the performance of the alpha beta ﬁlter. Test results demonstrate

that our angular measurement method is an efﬁcient way for proper road illumination along curved paths.

1 INTRODUCTION

Improving nighttime illumination is a crucial step for

trafﬁc safety (Plainis et al., 2006). Although the de-

velopment of headlamp system has drawn much at-

tention recently, many issues remain unsolved. The

direction of headlamps should be able to change ac-

cording to curved paths to enhance drivers visibil-

ity. When a car starts to make a turn, the focus

of the driver should be on a curved path. In addi-

tion to the front direction, the headlamps should be

able to illuminate along the curved road before the

cars turning to ensure that there is no obstacle along

the path. BMW also utilizes this kind of technology

called adaptive headlights (Adaptive Headlights); the

direction of headlamps is determined by the internally

measured front wheel direction. In this paper, the re-

lationship between the steering wheel angle and the

moving direction of a car is studied, so that the mov-

ing direction can be measured by vehicles other than

BMW. The cars moving path is then predicted by a

simple car model. The direction of the headlamps is

determined.

The major contribution of our work is to design

a method for determining the illuminating direction

without the knowledge of path history and digital map

data. Since the method does not depend on the path

history and digital map data, the system can respond

immediately even if the drivers do not follow the path

history or the digital map data is not up-to-date. It

is also more robust when compared with traditional

approaches because the data used to determine the il-

luminating direction is collected inside the car but not

from the road scenes or Global Positioning System

(GPS) signals which are unstable and easily affected

by environment. Moreover, the system can predict

and illuminate the cars turning path to provide enough

time for the driver to react and prevent accidents. It

emphasizes on a simple but robust approach to im-

prove curved path visibility in real-time.

2 RELATED WORKS

Predictive advanced front lighting system (P-

AFS) (Ibrahim, 2005) is proposed to determine

the path that a vehicle is most likely to drive. The

path is determined by using the current position of

the vehicle, a digital map and the current vehicle

data such as speed. After analyzing those data, the

headlight beam direction is found. The headlamps

will be swiveled accordingly to keep the kink point

of the beam in the center of the road. Clearly, the

performance of the P-AFS depends on the digital map

and GPS signals. If the digital map is not up-to-date

488

Chan K. and Moon Y.

AUTOMATIC HEADLAMP SWITCHING SYSTEM USING ACCELEROMETERS.

DOI: 10.5220/0002893704880491

In Proceedings of the International Conference on Computer Vision Theory and Applications (VISIGRAPP 2010), page

ISBN: 978-989-674-028-3

Figure 1: The circular motion of a turning car.

or the GPS signal is not stable, the performance of

the P-AFS will be affected.

(Morishita et al., 2007) investigate the drivers

preferences for headlight swivel angles on the

test track at the DENSO Abashiri Test Center in

Hokkaido. The preferences are then used to esti-

mate the optimal swivel angles by multiple-regression

model. Also, this study shows that the headlight

swivel should be predictive. This approach is based

on path history to determine the optimal swivel an-

gles. The system wont work if the driver does not

follow the path history or the path history is not avail-

able. Also, to locate the position of the car, a reliable

navigation system is needed.

3 KINEMATICS OF A TURNING

CAR

A simple car can be considered as a rigid boy. Adopt-

ing the symbols from (LaValle, 2006), the angular ve-

locity of the car is

θ =

tanu

(1)

When the car is turning around a corner, it per-

forms circular motion. In Figure 1, if the car is turn-

ing from point A to B in the period of time t, the angle

between point A and B at the center O is

a =

θt (2)

Assume the angles of the two front wheels are the

same. The ratio between the angle of the front wheels

and the angle of the steering wheel is ﬁxed in most

vehicles. If the ratio is r and the angle of the steering

wheel is u

, the angle between the tangent at point A

and the line joining point A and B is

b =

tan(u

r) (3)

4 HEADLAMP DIRECTION

PREDICTION

As discussed in section 3, the position of the car on

a curved path can be predicted by Equation (3). The

point A and B in Figure 1 correspond to the current

and predicted positions of the car on a curved road.

The predicted position is a place where the driver

should pay most attention to. It should be as close

to the car as possible so that any unexpected incident

near the car can be discovered, reacted and avoided.

The distance between the current and predicted posi-

tions turns out to be the car stopping distance. If the

driver uses time t

to react and the car undergoes con-

stant deceleration (d) along the curved path to avoid

accident, the predicted direction of headlamp is

b =





tan(u

r) (4)

Typically, the deceleration rate is 3.4m/sec

and

human reaction time is 2.5 seconds (Transportation

Research Institue, 1997). The general speed limits

in build-up areas are around 40-60km/h (Legislative

Council Panel On Transport, 1999). Based on the car

conﬁguration, the possible range of headlamp direc-

tions can be calculated by Equation (4). Then, an ar-

ray of headlamps, in which each headlamp has a con-

tinuous illuminating angle, can be installed in the car.

Each headlamp can be switched on according to the

instantaneous speed u

and the steering wheel angle

of the car.

The instantaneous speed can be easily obtained

from the speedometer of the car. To measure the steer-

ing wheel angle of the car, two accelerometers are

used to provide an accurate and stable angle.

4.1 Steering Wheel Angle Measurement

To measure the angle of the steering wheel, a two-

axis accelerometer (A1) is attached to the center of

the steering wheel with the x-y plane parallel to the

steering wheel plane. Another two-axis accelerome-

ter (A2) is installed such that one axis is parallel to

the line of intersection of the ground plane and the

steering wheel plane.

Initially, the X-axis and Y-axis of the accelerom-

eter A1 are placed horizontally and vertically on the

steering wheel surface respectively. The horizontal

acceleration A

of the car is measured by A2. If the

X-axis and Y-axis acceleration rates measured by A1

are A

and A

respectively, the steering wheel angle

= arctan





− arcsin





+ A





(5)

4.2 Steering Wheel Angle Stabilization

To stabilize the sensor readings measured by A1 and

A2, a running average algorithm with T data points is

AUTOMATIC HEADLAMP SWITCHING SYSTEM USING ACCELEROMETERS

489

Figure 2: Sensor readings from A1 are stabilized by the

running average algorithm.

Figure 3: Steering wheel angle after smoothing using alpha

beta ﬁlter.

adopted. The running average of a sensor reading A(i)

at time i is calculated by Equation (6). Subsequently,

the rotational acceleration of the steering wheel gen-

erated by the driver which is treated as a short-term

ﬂuctuation is removed by applying the alpha beta ﬁl-

ter (Penoyer, 1993) on the steering wheel angle. The

alpha beta ﬁlter is employed because it is fast in com-

putation and requires only the previous state for esti-

mating the next state. As described in Figure 2, the

sensor readings from A1 are stabilized by the running

average algorithm. Figure 3 shows the steering wheel

angle which is smoothed by the alpha beta ﬁlter.

A(i) =

∑

k=i−T +1

A(k) (6)

5 EXPERIMENTS

In the proposed approach, the predicted direction of

headlamps is calculated by Equation (4). In Equa-

tion (4), there are typical values for all the variables

except the steering wheel angle. So, the accuracy of

the direction mainly depends on the steering wheel

angle. To test the accuracy of the measured steering

wheel angle, several experiments were conducted in a

simulated car environment.

(a) Front view (b) Side view

Figure 4: Experiment setup.

5.1 Experiment Setup

A circular disc was afﬁxed on an inclined plane at

an inclination angle of 60

◦

. Two accelerometers

(ADXL322) were attached to the center of the disc

and the ground plane respectively. The experiment

setup was shown in Figure 4.

There were two sets of experiments. In the ﬁrst

set, the system was tested under 4 different angles

(−45

◦

, −90

◦

, 45

◦

, 90

◦

). In each test, the disc was

rotated from 0

◦

to a speciﬁc angle ﬁve times. During

the test, the plane was vibrated manually to simulate

the car vibration on a rough and curved road. In the

second set, the disc was ﬁrst rotated to one of the 4

different angles (−45

◦

, −90

◦

, 45

◦

, 90

◦

). The plane

was then vibrated manually to test the stability of a

ﬁxed steering wheel angle.

The best results were achieved when the parame-

ter T in Equation (6) was set to 10. The parameters

alpha and beta in the alpha beta ﬁlter were set to 0.2

and 0.01 respectively.

5.2 Evaluation Results

The angular transitions of one trial in each test are

shown in Figure 5. Among the tests, our system

can successfully remove the noise and ﬂuctuations

from the sensor readings. Table 1 shows the root

mean square deviations (RMSD) at speciﬁc angles

while vibrating. The results are reliable and ro-

bust to car vibration on uneven and curved roads.

There is a real time demonstration available at

http://www.cse.cuhk.edu.hk/˜kcchan/stabilization.avi

6 CONCLUSIONS AND FUTURE

WORK

We have presented a novel method to determine the

headlamp movement direction when the car is turn-

ing. Without using any path history or digital map

data, the directions of headlamps on curved paths are

predicted based on the steering wheel angles. Two ac-

celerometers are coordinated to measure the steering

VISAPP 2010 - International Conference on Computer Vision Theory and Applications

490

(a) The ﬁrst trial rotating from 0

◦

to −45

◦

(b) The ﬁrst trial rotating from 0

◦

to −90

◦

to 45

◦

(d) The ﬁrst trial rotating from 0

◦

to 90

◦

Figure 5: The angular transitions of one trial under speciﬁc

angles.

wheel angles. The sensor readings can successfully

be stabilized using the running average algorithm and

alpha beta ﬁlter. The evaluation results under a sim-

ulated environment are promising. In the future, the

system will be implemented on a vehicle to evaluate

the illumination effectiveness under corners with dif-

ferent radii. The simple turning car model can be im-

proved later for better modeling of car turning.

Table 1: The root mean square deviations (RMSD) at spe-

ciﬁc angles.

−45

◦

−90

◦

Overall

No. of 214 229 226 257 926

Samples

RMSD 4.66 4.50 1.51 1.07 3.30

(degree)

ACKNOWLEDGEMENTS

The work described in this paper was substantially

supported by a grant from the MPECENG(SEEM).

REFERENCES

Plainis, S., Murray, I.J., Pallikaris, I.G. (2006). Road traf-

ﬁc casualties: understanding the night-time death toll.

Injury prevention journal of the International Society

for Child and Adolescent Injury Prevention. BMJ Pub-

lishing Group Press.

Ibrahim, F., Schmidt, C., Klein, M. (2005). Predictive Ad-

vanced Front Lighting System In Proceedings of the

ISAL 2005 Symposium. ISAL Press.

Morishita, M., Horii, Y., Hagiwara, T. (2007). Investigation

of Headlight Swivel-angle Preference at Curves on

Rural Two-lane Highways In the 86th Annual Meeting

of the Transportation Research Board.

LaValle, S.M. (2006). Planning Algorithm. Cambridge Uni-

versity Press, Cambridge, U.K.

Transportation Research Institue, Oregon State University.

(1997). STOPPING SIGHT DISTANCE AND DE-

CISION SIGHT DISTANCE Oregon Department of

Transportation, Salem, Oregon.

Legislative Council Panel On Transport. (1999). Leg-

islative Council Panel On Transport: Speed Limit

in Hong Kong. http://www.legco.gov.hk/yr99-

00/english/panels/tp/papers/a886e03.pdf

Penoyer, R. (1993). The alpha-beta ﬁlter. In The C Users

Journal. R & D Publications Press.

Adaptive Headlights. http://www.bmw.com/com/en/newvehi

cles/x3/x3/2006/allfacts/ergonomics ahc.html

AUTOMATIC HEADLAMP SWITCHING SYSTEM USING ACCELEROMETERS

491