DE VELO PMENT S AFE T Y
Windshield Cleaning Technology for Improved Road Safety
The Ramp Integrated Blade unites wiper and washing systems in a single component. During the cleaning process the visibility is thus not impaired. Valeo launched a comparative usability study with Fraunhofer IOSB Institute Karlsruhe to give the scientific evidence for a safety gain of the system. The statistical analysis enables to demonstrate the safety benefit of the Ramp Integrated Blade.
28
REDEFINING WINDSHIELD CLEANING
Wiper systems, invented in 1903, ensure drivers visibility through windshield under various environmental conditions. To maintain the windshield transparency, washing systems spray liquids to eliminate pollutions. Washers were invented in in the early 1920’s. Wiper and washer systems are still separate devices, but they are functionally synchronised. Nozzles developments went from single to triple jets nozzles, disadvantage being that the fluid was limited to distinct spots. In the 1990’s so called Fluidics nozzles were used, which distributes washer fluid over a larger area to improve coverage. The limitation of this system is that the view of the driver is obliterated for a time period. Fluidics reached nevertheless industrial success being the standard solution today. Mid of the 1990’s Integrated Cleaning appeared (individual nozzles on wiper arms or blades) to improve efficiency at high speed, but still with visual imperfections. A breakthrough came in 2012 when Valeo introduced a new system integrating washing and cleaning functions to suppress totally the visual defects: the Ramp Integrated Blade (known under the trademark AquaBlade). The principle of this novel technology is that the liquid is driven through channels on the two sides of the blade where it is poured through laser drilled holes all along the blade.
Using measurements of an angular sensor in the wiper motor, a software synchronises the pump with the wiper movements, FIGURE 1. This enables to wet the glass approximately 10 ms before it is wiped, thus avoiding any visual disturbance while enabling a sufficient wetting. An effect of this efficiency (50 % less consumption of liquid) is that it makes possible to double the service interval time or to reduce the volume of the liquid reservoir by half, saving usually 2 kg per car. While the absence of visual disturbance when using the Ramp Integrated Blade had been observed on several production cars, scientific evidences for a safety gain were still missing. It remains challenging to make human feelings objective and to draw clear conclusions on data based on human opinions. For this reason Valeo launched a comparative usability study with Fraunhofer IOSB Institute Karlsruhe [1].
AUTHORS
Frederic Giraud, M. Eng. is Product Marketing Director of the Valeo Wiper Systems Product Group in La Verrière (France).
Dr. techn. Dieter Willersinn is Research Group Leader at the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation (IOSB) in Karlsruhe (Germany).
SAFET Y ASSESSMENT
The impacts of Fluidics and Ramp Integrated Blade on drivers have been evaluated following scenarios at low vehicle speed and urban environment. It has been paid attention to get a reference (meaning a perfectly clean windshield and no washing nor wiping actions, so no disturbance). The following perception performance parameters were measured:
Dipl.-Ing. (FH) Dietmar Baumgärtner is Laboratory Manager of Valeo Wiper Systems in Bietigheim (Germany).
Ph. D. (Eng.) Gilles Petitet is Systems Laboratory and Simulation Manager of the Valeo Wiper Systems Product Group in Issoire (France).
Arm position sensing
Arm position sensing
Hose Body controller Washing pump
FIGURE 1 Topology of the system 06I2015
Volume 117
29
DE VELO PMENT S AFE T Y
– The reaction time in [ms] for detection of a pedestrian approaching the road. – The time needed in [ms] to classify safety details into benign or critical traffic situations (missing of eye-contact when the pedestrian approach the road). – Finally a qualitative questionnaire has been filled by the testers. The cohort of persons has been designed to reach a confidence level of 95 % and an error probability of 5 %. A total of 204 persons participated, which means more than 5500 individual tests runs. The sample was constituted of 108 men, 96 women, between 18 and 80 years old, 80 % of the drivers having driven for more than 10 years. A Bernoulli process showed that 150 subjects would have been sufficient to stabilise and statistically validate results. TEST INFRASTRUCTURE
Two cars have been prepared for the test. These cars were at the origin exactly identical (same equipment, engine, colour), the original equipment was Fluidics nozzles and has not been modified, only checked, so that they were working at their best performance level. Then, on one of the two cars, we implemented the Ramp Integrated Blade system. The reference being realised by one of the two cars when neither the washing nor the wiping is in action. Both cars were equipped with sensors on pedals and with a radio link to the measurement centre. The study has been conducted by the Fraunhofer IOSB in its site of Karlsruhe. An area has been prepared on the track to contaminate the windshield with mud and salt. This has been done to put the driver in a realistic scenario of test. Some metres further on the track a display has been installed showing different scenes of pedestrians approaching the road, FIGURE 2. They appeared close to real size as the display diagonal was 2.04 m. Three video sequences represent three different situations with crossing pedestrians. An essential feature of all sequences was that the key images for pedestrian detection and the situations classification, FIGURE 3, were in each sequence in the same position.
Display
Bump
Contamination
Measurement centre
FIGURE 2 Test infrastructure; the dashed line from car to measurement centre represents a radio link
placed 24 m ahead of the display which reminds the driver, when running over, of starting the washing cycle. The event of pulling the washing switch is transmitted to the measurement centre to start the displaying of video sequences. For the reference runs however, where no washing is performed, the start of the video is triggered by a light barrier which is located at the bump. As the main goal is to measure reaction times, a particular attention has been paid on the evaluation of the latency between Ground Control station (GCS) displaying an image and the appearance of the image on the display, the temporal course of displaying the video sequence frames and the latency between the driver pressing pedals and the GCS recognising it. A careful analysis of the entire processing chain yielded a worstcase uncertainty of 3 ms, which is two
orders of magnitude below the observed reaction times. TEST PROCEDURE
The test procedure of the reaction time measurement consists of five stages. First the driver starts the car, and moves under the contamination gate (at this moment, for Fluidics and Ramp Integrated Blade tests, the car is covered with mud and salt). Then as a second the car approaches the display. It rolls on slight bump which gives the driver a haptic signal that he can start to wash. As a third, when the washing switch is activated the system starts to spray washing liquid on the windshield and to wipe, and a signal is sent from the car to the measurement centre, which starts to send the video flow to the display. This
DATA ACQUISITION
The driver triggers the wiping and washing system. To this end, a bump was
30
FIGURE 3 Example of critical and benign scenes appearing on the display
Distribution [Number of people]
35
TEST RESULTS
30
Results for detection time and classification time have been analysed as histograms, FIGURE 4, from which average values were computed and then analysed. The evaluation of the detection time shows that from a general point of view the time needed for the detection task is between 800 and 1200 ms, which, compared to previous studies [2], is quite logical for drivers knowing that “something will happen”. In the experiment, the reaction time difference between the reference case (when nothing is disturbing the driver view) and the Ramp Integrated Blade is of 48 ms only. So it has been demonstrated that this innovative technology of cleaning almost doesn’t interact with the driver vision and ability to manage dangerous situations. The difference between the Detection times with Ramp Integrated Blade and the Fluidics nozzles is of 315 ms, which is strongly impacting the ability to react quickly and efficiently (Roughly an increasing of 30 % of reaction time). Translated in estimated stopping distance at 50 km/h for a typical car [3], these 315 ms gives the situation in FIGURE 5. In addition, the analysis shows an increasing reaction time of approximately 6.2 ms per year of age of driver which is rather close to values seen in others studies where testers had to drive being disturbed [4]. It is important to observe that the safety benefit of the Ramp Integrated Blade system (difference of average detection time between Fluidics and Ramp Integrated Blade) is increasing with the age of the driver as shown on the graph, FIGURE 6. The classification task, whom the object is to discriminate, as quickly as possible, two different situations, is a very important one as lots of governments are recommending to recognise these situations during driving license lessons. For the needs of the tests we asked to the drivers to press throttle for benign situations and brake pedal for critical. The earliest reaction occurs at 1530 ms, the last reaction for the critical scene is at 3600 ms. The quantitative results show that the gain of classification time with the new system versus Fluidics is about 278 ms. The questionnaire is part of the safety assessment and was the opportunity to validate the interest of the drivers for
25 20 15 10 5 0 0
200
400
600
800 1000 1200 1400 1600 1800 2000 2200 2400
Time after the event [ms] FIGURE 4 Example of detection time (ms) distribution for benign scene, Ramp Integrated Blade
FIGURE 5 Estimated and rounded stopping distances of the car after an event, for different situations
sequence ensures that the movies characteristic frames (the moment when a pedestrian appears) are always shown when the wiper blades are in the same position on the windshield. As a fourth is detection time test: A scene showing pedestrians approaching the road is shown. If a pedestrian is appearing the driver is instructed to release the throttle pedal, there we measure its detection time. If nobody is appearing the driver must remain on the throttle pedal, this run is then finished. In the final and fifth stage the classification time test takes place: In the case when a pedestrian has appeared on the display there are two possibilities, FIGURE 3. First case the pedestrian has eye-contact with the driver, in this case the situation is benign and the driver is instructed to push on the throttle pedal. 06I2015
Volume 117
In the second case there is no eye-contact and the situation is critical, the driver must push on the brake pedal. There we measure the classification time. Then this run is finished and the driver can go to the next run. Testers had to do 27 runs as they had three types of systems to test (Ramp Integrated Blade / Fluidics / reference) and three situations (nobody, a pedestrian with eye-contact, a pedestrian without eye-contact). For each situation there were three representative video sequences. The video sequences were shown in random order. However, care was taken that each of the six possible orders in which the three systems were tested was equally represented in the test plan. Then the drivers were asked about their preference for one of the systems.
31
DE VELO PMENT S AFE T Y
400
Gain on average detection time [ms]
360
320
280
240
200 18-30
31-50
51-80
Age of the driver [years]
this type of technology. Asked to say if they actually encountered such visibility problems during a washing cycle, 86 % of the respondents confirmed. Asked which system was the safest, 82 % said Ramp Integrated Blade (8 % don’t know). 70 % are convinced that this type of technology will prevent accidents (22 % are not sure). PRECONDITION OF CAMERA-BASED ASSISTANCE SYSTEMS
A large study has been carried out including an unprecedented large number of testers and combining objective and subjective approaches. The statisti-
32
cal analysis enables to demonstrate the safety benefit of the Ramp Integrated Blade (AquaBlade system of Valeo was used for the test). Generally it shows the interest to integrate the nozzle function as close as possible to the windshield. In the future the need for perfect cleaning of forward-facing cameras (under the windshield) will make this kind of technology even more necessary, as autonomous cars will have to maintain a perfect camera vision quality to avoid any corruption of the video flow and so danger for the passengers. Further than this, from an Innovation point of view it shows the possibilities to reinvent, thanks to electronisation of
FIGURE 6 Difference of detection time between Ramp Integrated Blade and Fluidics nozzles over the driver’s age
functions and softwares, and for a shared progress in safety, such mature systems than wiper systems. REFERENCES [1] Willersinn, D.; Manger, D.; Erdnüss, B.; Petitet, G.; Baumgärtner, D.; Giraud, F.: Safety Assessment of a washer jets integrated wiper/washer system. In: Proceedings of the 2015 JSAE Annual Congress (Spring), Yokohama, Japan [2] Bernardina, F.; Bremond, R.: Measuring the effect of the rainfall on the windshield in terms of visual performance. Elsevier, Vol. 63 [3] Stopping sight distance and decision sight distance, prepared for the Oregon Department of Transportation by the Transportation Research Institute of the Oregon State University, 1997 [4] Hansman, R. J.: Experimental Studies of Driver Cognitive Distraction Caused by Cell Phone Use. Transportation Research part F, MIT
Lightweight and more efficient: Developed by FTE automotive
FTE automotive – Innovation drives While working with a customer on an innovative gearbox we achieved significantly reduced CO2 emissions with this electrically powered oil pump made of high-performance plastic. We could be taking the next step with you.
www.fte.de 06I2015
Volume 117
33
