Impact of Variable Luminous Intensity in the Low Beam Section

To investigate the influence of the maximum intensity for the low beam section of the light distribution, a test is performed, in which the detection distances for different low beam settings are recorded.

test setup

The test is performed at theAugust Eulerairfield in Griesheim, Darmstadt. The airfield offers the opportunity to set up real life driving tests on a 1.2 km straight road in a controlled en-vironment with similar conditions on each test day and no unwanted traffic or light sources.

The testing area is shown as an OpenStreetMap (OSM) image in figure5.1. The main straight, the southern part of the airfield, is the 1.2 km straight on which the test is performed.

Figure5.1– Testing Area on theAugust EulerAirstrip as anOSMimage.

Since current state-of-the-art headlamps are mainly using LEDs as light sources, as de-scribed in chapter3, this test will also be based onLEDheadlamps. Since the variation of the low beam intensity is only possible in a very limited way for different production cars, the headlamps in use, need to be modified. For this reason, the low beam LEDs need to be ad-dressed manually and can not be driven using the vehicles’ headlamp control. Since the test vehicle is a OpelInsigniaA, from 2008(BuickRegal, VauxhallInsignia), it is equipped with HID headlamps. These are replaced with Audi A8 D4/H4 headlamps from 2009. The light distribution is shown in appendix B.1in figureB.1. The main reason for choosing this headlamp set is, that the LEDs are addressable via the standard input pins. This is used to limit the maximum luminous intensity in the headlamp hotspot. Both headlamps are limited to 50 % for the reduced setting. This leads to 10 000 cd in the hot spot for each headlamp under 50 % and 21 000 cd in the normal setting for each headlamp.

5.1 investigation of detection distances under varying light conditions 87 To measure the resulting detection distance for both settings, two human like test dummies, similar in shape and colour to the ones used in the studies shown in chapter 4, are placed at both sides of the road. These dummies are coated in a matte grey paint with a reflectance of ρ = 5 % to simulate dark clothing and therefore a worst case scenario. Measuring their reflctance also shows, that they are nearly perfectly lambertian. x1They are motorized and can be erected automatically in a randomized order with the auxiliary condition, of at least three test runs per setting, where the dummy on the right side is erect. The randomized order for both sides is used to avoid, accustoming of the participants and thereby increasing the detection distance for each run. Both test vehicle and detection dummies are equipped with General Positioning System (GPS). This setup allows for synchronization between the vehicle and the dummies positions through the time stamps. The time of detection is then recorded, by the press of adetection buttonand the resulting detection distance can be measured.

test process

The detection dummies are located around 500 m behind the starting point of the run way and 1 m besides the road markings. The setup is schematically shown in figure5.2where the starting position is shown as a red circle on the bottom left, the end position is marked by the red circle at the right side, the test dummies are shown the middle and test area is marked in red. The drivers are asked to drive the straight line with a velocity of 50 km h⁻¹ using cruise control. The cruise control is pre-set by the test coordinator and the test subjects only acceler-ate to about 45 km h⁻¹ and then reactivate the cruise control which will accelerate up to the pre-set of 50 km h⁻¹. The rest of the course, marked by the grey arrows, the participants are allowed to drive according to their personal preferences since no data is recorded at those positions.

B

A v^Test=^50km /h

Figure5.2– Schematic setup for the test on the influence of luminous intensity in the low beam area. The grey human like figures indicate the possible positions of the detection dummies. The red circles mark the start end the end of the test section.

To enable a proper adaptation time for each setting, the light distribution is changed at the end point. This leads to about 2 min time for adaptation to the new brightness setting, depending on the driven speed of the participant in this area, but is never below 1.5 min.

During the complete test, the test coordinator is seated on the co-drivers seat to instruct the participant and control the electronic recording of theGPS data and the correct recording of the detection times. Before the actual test runs start, each participant takes 5 runs to get accus-tomed to the vehicle and the course. This test set-up is also chosen to further investigate the influence of oncoming traffic and wet road surfaces on glare. Therefore, an additional vehicle

88 analysis and optimization of light distributions

is placed on the testing area. For both vehicles the intensity in the low beam is then varied.

However, since this thesis only focuses on the influence of different light distributions for the driver, all test drives done with the additional glare vehicle and the wet road surface are not discussed here. Furthermore, the data with dummies on the left side (from the driver’s view) are not evaluated since the recorded data for this setting is not sufficient since the main emphasis is set upon the detection of objects closest to the driven vehicle.

19 test subjects participated at the test, all participants are either students or follow an office job. The subjects were asked to imagine a night drive on country roads, but not further intro-duced to the goal of the study. The mean age of the test subjects is 28 years and the complete information on all participants can be found in figureB.2and in tableB.1in appendix B.1.

To determine the exact influence of the low beam intensity on the detection distances, lu-minance pictures are recorded for all light distributions in 10 m intervals starting at 100 m.

As already described in chapter 2.3 these luminance recordings need to be done stationary since integration times of up to 10 s are needed due to the low surrounding luminance at the August EulerAirstrip. Therefore, it is important, that the lighting functions and settings are available during complete still stand. The results of these luminance recordings are summa-rized with the results from the following two studies in section5.1.4.

detection distances with variable low beam intensity

The raw detection distances for both, the 50 % and the 100 % low beam settings are set into 5 m bins and the histogram data for the detection frequency is calculated for both data sets.

This is shown in figure 5.3. Since the distance is calculated from the press of the button, the reaction time is already deducted from the data. With a reaction time of 500 ms, the raw detection distance would be increased by ∆d = 7 m when driving at 50 km h⁻¹. This detection data shows, that the data sets are highly overlapping and a normal distribution is not necessarily given.

0 20 40 60 80 100 120

0 1 2 3 4 5

Detection Distance in m

DetectionFrequencyina.u.

50% Low Beam 100% Low Beam

Figure5.3– Detection distances grouped into 5 m bins for the two low beam set ups. The dimmed 50 % low beam detection data is shown in blue, and the standard low beam (100 % is shown in red. The mean detection distances are calculated to be 52.6 m for the 50 % low beam and 62.8 m for the standard low beam.

5.1 investigation of detection distances under varying light conditions 89 However, normalizing the data according to equation5.1, where x is the normalized data, d is the measured data set of the detection distances, µ is the mean detection distance for the light distribution andσis the standard deviation for this data, this data can be tested for normal distribution using a OneSampleKolmogorov-Smirnov.

x = (d − ^µ)/σ (5.1)

This test shows, that both data sets are indeed normal distributed, however while the 100 % low beam leads to a very good representation of a normal distribution with a p = 0.98, the 50 % low beam only leads to a distribution similar to a normal distribution with p=0.3. The distribution form is visualized with Cumulated Distribution Function (^CDF) plots in appendix B.1 figure B.3. With both data sets originating from normal distributions, the significance between the two data sets can be tested using the TwoSampleKolmogorov-Smirnov. This test shows, that the data sets from 50 % and the 100 % detection are indeed not from the same distribution. However, with p = 0.08, the 0.05 threshold for significant differences in data sets is only slightly crossed. This overlap of the data sets is visualized in figure5.4 as box plots.

50% Low Beam 100% Low Beam

30 40 50 60 70 80 90

DetectionDistanceinm

Figure5.4– Box Plots for 50 % low beam (left) and 100 % low beam intensity (right). The overlap of both data sets show, thatp=0.08 only slightly crosses the threshold for significant differences.

The mean detection distances for both distribution are measured at 52.6 m for the 50 % low beam and at 62.8 m for the standard low beam with 100 % of luminous intensity. This leads to a difference of 10.1 m or a reduction in detection distances by only 16.2 %, when decreasing the luminous intensity by half. The detection probability over the distance is shown in figure 5.5, where the probability to detect an object with 50 % low beam is shown by the blue line and the detection probability with fully activated low beam is indicated by the red line.

90 analysis and optimization of light distributions

0 10 20 30 40 50 60 70 80 90 100

0 0.2 0.4 0.6 0.8 1

Detection Distance in m

DetectionProbability

50% Low Beam 100% Low Beam

Figure5.5– Detection probability for 50 % low beam intensity (blue) and 100 % low beam intensity (red).

The 50 % thresholds are recorded at 52.6 m for 50 % and at 62.8 m for 100 %. The 95 % threshold is measured at 29.5 m and 36.9 m respectively

.

In this figure, the 50 % probability marks the mean detection distance. Since 50 % detection probability means, that only half of the test subjects are able to detect an object at this distance, and half of the participants did not, the mean detection distance is not a good indicator for traffic safety. For this reason, additionally the 95 % probability distances are calculated, analogous to Adrian. This results in distances of 29.5 m for 50 % low beam and 36.9 m for the standard low beam, a difference of 7.4 m. The effect on traffic safety for this will be discussed later.

5.1.2 impact of variable luminous intensity in the high beam section