Q&A with Mobileye Vision Technologies

http://www.just-auto.com/article.aspx?id=95782
Q&A with Mobileye Vision Technologies
22 August 2008 | Source: just-auto.com editorial team
For some time, Mobileye has been pushing back the technical
boundaries of vision systems for intelligent transportation. The
company is recognized as a world leader in vision-based DAS. Its
technologies are sourced by a number of automakers, including BMW,
GM and Volvo. More specifically, Mobileye's AWS (Advance Warning
System), is being distributed and sold worldwide. Mobileye NV is
headquartered in The Netherlands, with a research and development
centre located in Jerusalem, Israel and sales and marketing offices in
Detroit, Michigan; Cyprus and Tokyo, Japan. Matthew Beecham talked
with Ido Amir, marketing manager, Mobileye Vision Technologies about the company's driver
assistance systems.
just-auto: Could you provide a brief update on Mobileye's fortunes with respect to
its vision-based DAS, i.e. on which new vehicles is it fitted to?
Ido Amir: Our technology is currently fitted to the BMW 5 Series, GM Cadillac DTS, STS, Buick
Lucerne, Volvo S80, V70 and XC70. In 2009 the list will expand.
just-auto: How is your aftermarket business shaping-up?
Ido Amir: Currently between 10 and 20 worldwide signed distribution agreements, including
US, UK, Chile, Holland, Turkey, China, Japan, Denmark and Australia.
just-auto: What is in the pipeline?
Ido Amir: Mobileye's future pedestrian detection technology, blind spot detection and lane
change assistance, automatic high-beam control, side-looking and rear-looking cameras are all
in store as next generation aftermarket products. Multiple feature consolidation is planned to
be achievable mainly thanks to the new Mobileye EyeQ2 chip, which represents another
technological breakthrough, being almost an order of magnitude stronger than its predecessor,
the EyeQ1, itself a leading chip of its kind.
just-auto: Conventionally, sensors such as LIDAR, millimeter-wave radar, and vision
sensors were used in driver assistance systems independently to recognize vehiclesurrounding
conditions. I guess the emphasis nowadays is on combining several
different sensors in order to achieve more accurate recognition. Would you agree?
Ido Amir: I would say that for functions that require a high level of confidence in detection,
such as automatic braking there is currently room for sensor integration. Consider a false
detection and consequently an emergency braking where there is actually no car in front of
you, or that there was really no need for braking, since the ahead vehicle was mistaken for
stopping. There, you would not prevent an accident, rather, you would create one. However,
as for the non-active, non-interfering functions, such as forward collision warning, which emits
some sort of feedback to the driver, leaving it in their hands to prevent an impending accident,
there is less damage in issuing false warnings, and even in a missed warning.
just-auto: To what extent is there a move to offer greater functionality and
integration of certain sensors for DAS applications?

Ido Amir: This is actually one of the main points of strength a camera has over its competing
sensors. OEMs would like, naturally, to spend as little as possible on hardware, and receive as
many functions as possible. This being said, it is understandable why OEMs, and especially the
ones who are the leading edge of implementing technology into their vehicles, place feature
consolidation very high up the ladder when considering both cost reduction, and innovation,
which is a main selling tool. A digital camera has a very strong position in this head-to-head
sensors competition. Neither radar nor LIDAR have the capability to perform a combination
such as the following: lane departure warning, forward collision warning, headway monitoring
and warning, pedestrian protection (forward, sides, and rear), intelligent high beam control,
rain sensing, day/night sensing (for automatic headlamp control, also called automatic lowbeam
control), obstacle detection (rear looking), traffic sign recognition, and the list goes on.
just-auto: Could existing reverse parking sensors play a greater role in partial and
full parking assistance systems?
Ido Amir: Actually, the trend is clearly to assist the driver actively in as many ways as possible,
and the first ones who will be able to relieve drivers of the need to park their cars will
capitalize on that. To predict whether or not the existing, ultra-sonic, parking sensors will be
relevant for a 'full parking assist' function is not an easy task. Ultra sonic sensors have the
advantage of sonar, i.e. the ability to tell there is an obstacle in a certain direction. However,
it's been noted that these sensors also have 'holes' in their detection capabilities. Moreover,
they are not intelligent and accurate enough to 'tell the whole picture" Some car OEMs are
already equipping their high-end vehicles with an all-around-view set of cameras, which gives
the driver something like a 'bird's eye view' of their own car when they are attempting to park.
This allows the driver to park their car without looking sideways or backwards, which is inching
toward the next step: automatic parking. The technological step that needs to be surmounted
is to add more artificial intelligence to those systems, and voila! I'm not suggesting that the
required AI (software) is easy to create, of course.
just-auto: Could car navigation systems play an important role in detecting vehiclesurrounding
conditions?
Ido Amir: Not only could they, but they probably should, and the sooner the better. No invehicle
sensor can see what is not in their line of sight! For example, driving around the bend
and straight into a line of cars that have rear ended each other is not going to be solved by
anything that is not able to see what's happening beyond the next curve you are approaching.
For this, intelligent traffic systems that have 'eyes from above', are a feasible, and an
excellent solution. Naturally, a better-connected road scene (WiFi, Bluetooth, or some other
type of wireless communication) would make a huge difference in road safety. Terrible
accidents that occur in tunnels, sometimes involving tens of vehicles, and always deadly, can
be prevented altogether, if all cars 'know' that one car is now stuck in the tunnel ahead.
just-auto: Flashing lights and video screens on the dashboard and audible bleeps
from the car's loud speakers all suggest distracting the driver too much. How will
the driver respond? Will they use the information correctly? Will it elicit the correct
response?
Ido Amir: In fact, many studies were made on these exact questions. Since an alert is futile if
it does not elicit an effective driver response. The National Highway Traffic Safety
Administration (NHTSA) has issued a long list of recommendations, based on such studies,
which define the do's and don'ts of driver feedback. Whoever adheres to those
recommendations should expect to have an effective DAS, that has a good chance, statistically,
to prevent and mitigate road accidents. These NHTSA comments take into account
disturbances in the cabin, and indicates the correct measures that should be taken in, for
instance, a noisy cabin, e.g. they define the type of sounds that should be used as an alert,
but also the volume range that would make an effective warning.
just-auto: Is there a danger that some drivers will actually drive faster or pay less
attention to the possibility of a hazard ahead if they have faith that a gadget will
alert them if needed?
Ido Amir: The danger is always there. Therefore companies like Mobileye make sure to state
very clearly that having the device in your car is in no way replacing the need for safe and

alert driving at all times, according to the local laws. To put it in a less formal way: you can't
prevent people from doing stupid things, You can only try your best, and make sure all the
reasonable warning signs are put up. My personal view on this is, that in the short term,
people 'test' new systems (which is not advisable, but happens). In the long term, however,
they get used to it, and continue to drive according to their natural driving style; this is when
DAS really become the most effective. Also, as long as the DAS does not take real action, the
driver is eventually aware of it, so this is another force that acts in the direction of more
sensible driving.
Another point is that having a DAS in your car subconsciously raises the issue of driving safety,
and the fact that the road is a dangerous place. This, in turn, can and does raise the general
sense of alertness - at least from some checking and surveys I've done. Just noting that LDW
systems were able to reduce between 65 and 85 percent of all lane departure accidents in a
few large trucking fleets in the US gives you a strong indication of the 'net potential of DAS.
To summarise, the overall expected result of widespread implementation of DAS, is a
considerable drop in crashes.
just-auto: Some people refer to driver assistance systems as the 'second revolution'
after ESP. How and where do you see the technologies evolving?
Ido Amir: This is a long discussion. In short, I would say that there is room for technologies
inside and out of the car that will make the roads of tomorrow safer. Starting with
'smart infrastructure', through inter-vehicle wireless communication, and sat-nav based
systems, and ending with both active and passive in-vehicle DAS. The rule of thumb is always:
the less drivers will need to be responsible for driving, the fewer accidents we will have.
just-auto: As the algorithms for doing image processing is continuously evolving and
more and more computer power enters the vehicle, could you foresee that in the
next decade, every new car will have a video camera?
Ido Amir: You have outlined the issue correctly, and in fact, Mobileye has just come out with
its second generation vision-system-on-chip, the EyeQ2, which is positioned as a global
leader of its kind, having six times more computational capabilities compared to its
predecessor, the EyeQ1. The EyeQ2 is a low cost, extremely powerful video processing chip
that Mobileye expects to be installed and used by a considerable percentage of the cars that
will be manufactured in the upcoming years. There are quite a few OEMs and Tier1 companies
that target the EyeQ2 as their future chip for DAS applications, and the process is constantly
ongoing and catching speed. This being said, and taking into account the value-to-cost and
feature consolidation discussion above, it can be well expected that in ten years from now,
most of the cars around the world will have a camera-based DAS in them.