From Night Vision to LiDAR: An Automotive Perspective talk by Jan-Erik Kallhammer at FiO+LS 2018


[MUSIC PLAYING] So what do we do? Autoliv basically is evenly
split between Europe, Americas, and Asia. And China is the biggest
part of Asia there. And sell to basically
all the web. So I’m pretty
certain that you have Autoliv products in your car. If not, your neighbor has it. And the majority of the
sales was air bags, and then seat belts. And electronics was
only about one fifth. And it was spun off because it’s
getting into different sections of the market. Autoliv actually has the
strange problem that we often [INAUDIBLE] bankruptcy. They are going towards
50% of the world market. And it basically means you can’t
grow in automotive anymore. And the electronics
is in a gross phase, and the other one becomes
in a more mature phase. And the stock market analysts
find that a bit confusing. So the remedy is
then to spin it off. It’s about $2 billion
of electronics. Being a global
supplier, you have to be basically where cars are
developed, or cars are made. Specially developed. So that’s it. Currently, 22 OEMs
as a customer. And the products we make
is a bunch of stuff. So each product is maybe
not that expensive. I mean, most automotive
products is 100 bucks or less. But if you add them up, the
delivery content per vehicle is pretty good. So this can be a pretty good
business if you do it right. And also, if you multiply
it by lots of volumes. So that’s about that. This was actually my first baby
when I started with Autoliv. It was a project,
inflatable curtain. That’s for voluntary agreements. So all US cars have
it, for instance. And it’s very good if you have
a side impact, or rollover, or anything like that. But this was a product that
actually wasn’t believed that it was worthwhile
putting the money into it, because the initial estimate
was there was no mandate, no end cap, no Insurance
Institute of Highway rating or anything like that. So the initial volumes was
estimated at 100,000 a year. So it’s a lot of NRA. Is it affordable? Well, last year we produced– I say we– Autoliv
produced 50 million. It’s about 50% of
the world market. Accumulated production is
now approaching 1 billion. And even if the price has gone
down, so maybe 40 bucks apiece, it’s about $2 billion a year at
a very healthy profit margin. So you can make a lot of money. And then, of course,
it’s also a long program. You have over 20-some years. Basically had lots of
very safe revenue base. So if you good in automotive,
it’s very profitable. If you do bad, it’s not. So that’s really
the lesson here. So then going into night
vision development. And I think night vision
is interesting for LiDAR, because it shares
some common traits. Initially, it was considered
to be too expensive. I mean, the sensor
basically costs more than the car it’s put on. Bad equation. It also has a bunch
of new technologies that the automotive community
wasn’t accustomed to. And exact requirements
are unclear. Installation
questions, et cetera. And there are many
new components that are coming
from suppliers that are not used to automotive. That’s also bad news
from experience. And the timing, actually, when
this is going to be needed is unclear. Of course, you have
those that are– or depending on venture
capital, they say, tomorrow. Well, it generally takes a
bit longer time in automotive. And another thing is that
the first year of production, real zero production,
you may produce more than the accumulated historical
production before that. And that’s a big leap in– to make it happen. It doesn’t do it by itself. The driving for night vision
was pedestrian fatalities, although, well, in
the US, it’s only 12%. In Europe, a bit more. But the majority of
them are nighttime. The European number
47 basically means that the accident statistics
is not really accurate all over Europe. In Japan, 68%. So basically, the majority
of them happens at night. Of course. Then you won’t have a sense
that’s very good at night. Another good thing is the signal
to noise ratio was very good, so you had an easier
image processing when the [INAUDIBLE] power at
that time wasn’t this high. And this is just an
example of what you do. So the left is then
a normal camera, and the right is
that infrared view. It’s a long-wave infrared. 8 to 12 micron wavelength. But you can also– the other
thing is animal accidents. Animal accidents, well, they
are not dangerous, you think. Well, actually,
they may kill you. You don’t want to
be in that car. So even if the human
fatalities in the US, for instance, was
only 200, but there is more than 1
million accidents. And in many cases, you
can’t drive from the scene. And it usually happens, then,
in areas where you may not have cell phone
coverage, or you may be difficult to get away from it. And– but still, there
is a lot of injuries. And there’s a lot of
property cost, also, for it. So you have to have an– but
animals don’t look the same. You have to basically make
sure that your algorithm can handle a bunch of different
animals that are relevant. In the US, you have a
bunch of deer accidents. But if you have a camel
or something like that, you have to make sure
it works also for those. And wild boar. We have a lot of
wild boar in Sweden. They are difficult to detect. So I jokingly said that if
rock moves, then not rock. That’s the algorithm. [LAUGHTER] And he also added that– I think this is BMW. And then you see here, you
have on this screen the image, and you also have
the head up display to give a warning and this–
and a warning for the deer up front here. But there’s another
interesting thing, the HMI here is
they use a spotlight to actually spotlight. You don’t want the driver to
take his eyes off the road when something happens. You want him to
focus on the road. So why not– like in a theater. If you have a theater
without spotlights, it’s very difficult to
follow what’s going on. But you have a spotlight
on, here’s the attention. It’s the same principle. And in fact, it works very well. So give some example here. A deer and then. But there is another one
deer coming in there. And there in the
left image, you see that it’s flashing on
the animal so that you would be able to see it. And the driver’s put
his attention there. So it’s a pretty
straightforward system. In animal accidents, the
dark up in the lower center part, but also around
here, is high risk states. There’s a lot of deer accidents. That has to be handled. And another interesting
thing for LiDAR which is related to
automated driving, you have to make sure your
automated vehicles can also handle that if you’re going
to have level three or up. You cannot say that you’re
driving in automated mode and suddenly there’s a deer. It’s like, ooh, I
don’t know what to do. Please take over. Doesn’t work like that. They’re in a bunch of vehicles. They are so far only in
premium and luxury vehicles. We were not that successful
to get the cost down quickly enough to use
the momentum to get it into the volume market. It’s a big check change to get
it into the volume vehicles from the premium vehicles. And this is the kind of things
that the [INAUDIBLE] there. From the left, you
have the bolometer. You have the electronics. You have the shutter that
gives one point correction. And then you have
the house– oh, psh. Wrong button. OK. I should put the
right but– and then you have a ring here,
which is actually a heater. And then you put the
protective window, which is a germanium
piece with DLC coating. And then you have, in
that lower left corner, the complete assembly. And the heater is there to
make sure you take away ice or condensation in conditions. This is my former
car I had that then– well, sometimes, when you’re
driving in the winter, it’s a challenge
for the sensors. Neither the radar– you
see the radar over there– or the camera didn’t
really work very well. And if you do all your testing
in Arizona and California, you don’t encounter
this kind of problem. The mountain snow in
California Bay Area is limited. This is an example [INAUDIBLE]
in the instrument cluster, and then you have
where it’s located in the Cadillac, where
it’s located in a Audi, and in a BMW. And you also see where
they have the spotlight. Of course, when you have
developments of new headlights, you can include the
spotlight into the headlights in future versions. But there were
some issues that we had that we should
have done better. And if you had been very good
at requirements [INAUDIBLE].. We go back to this heater
in the protective window. Well, actually, it was
susceptible to abrasion. DLC, we thought,
was hard enough. But germanium is a
bit brittle, so even if you have a hard coating on
the more brittle surrounding, it didn’t take pinging from
sand particles very good. It passed testing, but
the field results– so we had specifications. Basically, military specs
and some automotive headlight specs. So we thought we were OK. We would be
[INAUDIBLE] according to all the qualification
that we thought was relevant. But we got, after
a while, some field returns where the DLC
basically was stripped clean from the protected window. So what really happened,
then, in the field? Well, this is a sandstorm. We sold cars to
the Persian Gulf. It was a BMW 7 Series initially. People that buy $100,000,
100,000 euro car is not the average one,
so they were rich kids. And they realize that,
long-wave infrared, you can see through sand. So they were doing
racing in sandstorms. Of course, you didn’t see
anything through the window, but you saw through the camera. So they were racing each other. Kids. Of course, if you do racing,
because you’re sandblasting the whole $100,000,
100,000 euro vehicle, you are not the
poorest kid around, which meant that they had
some influential parents. So basically, when
the DLC coating drops and your [INAUDIBLE]
reflector, your transmission drops by more than 30%. So the emission is poor. You couldn’t race anymore. That’s major problem. So it was a warranty claim. And these kids had
influential parents. You couldn’t say that you
should have known better. It didn’t work out like that. You can probably imagine. I’m not sure if you give your
kids a brand new BMW 7 Series as your play car, but they had. But they should have
known, because this is an extreme transmutation. Actually, people protect
the car from sand. We didn’t do enough
field research to understand the
market where it’s going to happen in all places. So we had to look for
additional protection. [INAUDIBLE] grids in
front of the window do not work, because
the grids could not made– be narrow enough to catch
the rather small particles. And if you made
them narrow enough, you’d lost too
much transmission. So we try the harder way. Even went to try to
do diamond windows. Very expensive. Difficult for that
construction in automotive. It work better, but
you still had problems. This is similar problems
that the Air Force found when they did the
Persian Gulf War, when they had to suddenly replace
a lot of infrared cameras when they flew over
the desert for it. We should have
understood better. So we try boron sulfide. Toughest coating we knew about. But limited capacity,
and basically [INAUDIBLE] for
military programs, so it wasn’t very good. And it’s dangerous, expensive,
and dangerous to treat. So eventual solution is to
get it moved from the bumper. But initially– we were
late into the program, so the bumper
where it was placed was actually the only space
we could have at that time. And we didn’t know, really,
a good justification why it had to be moved. But then to put it into grill in
a BMW, it’s a styling element. Then you get styling involved. If you don’t know what I’m
talking about when styling is involved, you have a problem. Because they are very powerful. So when they have an issue,
that means go home and redo, as you know. They don’t say it’s wrong. They say, I have an issue,
and then you say, yes, sir. [LAUGHTER] Heat management. Infrared camera measures heat. The camera can flow
the temperature. You just have to make
sure it’s uniform. And the mounting location
was mounted too heavy. There’s a crossbeam there
for crashing purpose. Very heavy aluminum,
I think, it’s in BMW. Very heavy one, very sturdy
one to take your crash forces. So a lot of inertia. A lot of terminal inertia. Should be good. But during systems test, we
had some strange gradations. Almost like Newton rings, if
you understand roughly what I’m talking. And then– so it
occurred intermittently, and we were very confused
what was going on. Had extensive search for cause. And we added the
temperature sensors then, and we didn’t
believe our results. We had 20 degrees per second in
thermal increase at the mount. 20 degrees a second. It’s like a [INAUDIBLE] gun. And the whole camera
increased six degrees a minute inside where the sensor is. Bad news for a detector. It meant it also become
uneven, so that’s why we got those gradients. And the cause was
actually– after you had been driving with a trailer,
or in heels, or anything like that [INAUDIBLE] very hot,
you know high speed driving, and you come to a
stoplight, or you actually get into very slow traffic,
you actually got backflow. So the air was not going
from the front to the back, but actually all that
heat was actually blowing past the camera. So you’ve got 100 degree
centigrade boiling air, basically– boiling
water temperature– blowing onto the camera. And it wasn’t designed for that. Our mistakes. So we had to
redesigned the camera. Basically make it more
like a thermos flask to slow down the thermal– I mean, we had to
handle the move, but we had to make it slower. But of course, this came
during system tests, so we actually had a
pretty mature product. I think it was B prototypes. And then looping
back is bad news. Other things like rain drops. Well, of course, we know
there were rain drops. But you also had this heated
window and things like that, and you start to see
it in circumstances. When the water hits, it
splashes onto the window that’s slightly higher in order
to keep the humidity at just about freezing. You have to keep it warm so you
don’t get condensation on it. But that also meant that this
surface window was higher. And when the water
splashes into thin, it became actually a hot
spot that reflected back into the sensors. You got flashes. And the final one is that
installation tolerances. We had initially–
in order to reduce cost– put the bolometer in
form factor eight to three. Basically, you’re looking
at a horizontal scene. And it reduced
the detector cost. There’s a rule of sum. Basically, if [INAUDIBLE] was
half the area of the detector, I’d pay in volume half the cost. So that was a good incentive. At least I thought so. But what I didn’t
consider was the time it needed to be installed
at the assembly point. The end point where it
was married to the car. Because there as the
whole car comes down, you have basically a minute
to get it there, and align it. And that meant your vertical
alignment had to be very good. And we couldn’t do it. And we would have to slow
down the car assembly line. Won’t happen. And the final
[INAUDIBLE],, we actually had to increase the form
factor to four to three. We did discover it pretty early,
because we did the validation process. We understood it very quickly. But we actually had to double
the vertical resolution in order to then do
electronic adjustment, which meant you didn’t have to do the
vertical alignment very good. You instead just
looked at the target and then set your center
line electronically instead. But that, of course,
increased cost. But it didn’t slow
down the assembly line. So this one, we did catch
early, but it’s [INAUDIBLE].. And then, export. It was a US-made detector. And we made a camera in the US. We had our assembly
plant in Canada. I thought Canada is
like a 51st state. No offense to Canadians here. But that was not what the
US government considered. And– well, friendly nations. Well, nowadays, I’m not
sure who is friend anymore. But there are
strict regulations. We had to put our assembly
plant in the US, then. A new assembly plant. And there are some strict
requirements there. That also is a big difference. US supply basically
say, well, let’s all sign the Wassenaar Agreement. All Western countries have
done it, so it’s the same rule. Well, that’s in theory,
but the interpretation is done by the
individual nationals, and there is a big
difference in practice. So that means that,
if I can choose today, I won’t use a US supplier,
everything else equal, because it’s just too much
hassle to get the product. We have to produce worldwide. And you can, of
course, get there. But nowadays, you
also have to have an [INAUDIBLE] to even
start to change design specs and things like that. And if you can get
commercial assignment, then it’s the
Department of Commerce. It’s basically good. They are in favor of commerce. I think it comes from the name. But if it’s believed to
be military technology or military use of this estate,
that covers it, basically. They think that any
technology that goes away, it may put US
servicemen at risk. So therefore, it’s bad. So that’s a bit tougher. So getting the export license
to German employees [INAUDIBLE] took two years. And we had to
implement something which we call tamper
protection, but actually meant that the camera
had to be married to an electronic unit that was
married at the assembly line. And if the camera was
detached, it didn’t work. It had to work with a
specific [INAUDIBLE].. There’s been a
number of cases where they bought scrap
from a scrap car, and they can’t get to work. Well, it’s not
supposed to be work. So if you’re going to use
it for any illicit purposes, you basically have to
drag the whole car along. And that was the purpose. But you have to plan for that. Most– if you’re coming from
another area– don’t, probably. Now I’m going to
leave that thing and then go into
LiDAR specifications. LiDAR has to handle all
kinds of disturbances. Retro-reflect
[INAUDIBLE] and sunlight. But you also occasionally–
at least outside California– have rain, fog, dust, smog. China, for instance, you
have a lot of smog problems. And snow. And it will hamper you a lot. So the question is, how much
regulation can we allow? That’s really important,
because it determines how much power base do you
have to throw at the problem, because it’s more or less
that’s the only remedy we have. Throw more power at the problem. Which means you may not
be eye safe anymore, so you may have to
switch wavelengths. I wanted to have a [INAUDIBLE]
equation in the end. But if you put it
up to there, there is a small, usually
referred to as K. And it’s not the
constant people believe. That obscures the atmosphere. That dampens the– And that’s easily to
get, that way, below 50%. And that, of course, means that
the number of photos you get is less than 50%. And you have to increase
your power by 100%, double your power,
to come to that. And that’s not always easy. Or if you see the
distance, this [INAUDIBLE],, then we have a
70% range instead. Bad news. Of course, there
is some solutions. You can then go up
into [INAUDIBLE] gas. Over 1,400. Eye safety is so much better. If you go up here, you can
have 50 times more power, or something like that. You’ll have to check
your optical link budget. You have to be, of
course, eye safe. Class 1 is the only thing
considered in a car. Not 1M. Nothing class 3 or any
other thing like that. But you also have to make
sure that you have, also, reasonable power. Even if you can send
it out, you have to make sure you
have the capability to send out that much power. And that you have an
EMC that’s acceptable. Your scanning
mechanism, and you have to make sure it’s affordable. Suddenly, it became
a bit difficult. But when you consider
specification, you also have to say that
range resolution around the car is not the same. It’s basically higher
in the past [INAUDIBLE].. It’s pretty unnatural,
but some of them, I have had the same resolution. And that may be unnecessary
cost if you do that. But range and resolution
requirements are connected. So range, basically,
determine how fast you can drive in automated mode. But you have to have
resolution to determine small objects that you have
to see and maybe brake for. A range without
resolution is not enough. And there’s a lot
of things that, oh, you have to have 200 meters. Well, first of all, for
what, and for what object? Minimum resolution
determined at target range. As an example, not
promoting anything. If you take the rotating
[INAUDIBLE] which is used for many test
vehicles, they claim– from their spec– 120 meter range. And a vertical field of
view just under 27 degrees. That gives a vertical
resolution of 0.42 degrees. If I just measured that, that
means that, at range, it’s 0.9 meters between each spot. If you have a total equal,
you get maximum two spots, and I leave it to you to
accurately classify that as a vehicle and nothing else. Because if you need to brake,
you have to be very certain. Because if you
brake unnecessarily, you will be hit from the rear. And that can be very bad. I actually know from
personal experience that you don’t want
to go it again. There was a reason
[INAUDIBLE] so sensitive make the [INAUDIBLE]
very much sense. But what does it
mean for resolution? They claim that you need
to have, at minimum, 150 meters range. And you have to
see 10 centimeters. Tall objects. Agree with many
[INAUDIBLE] specs. So I agree there. But what does it mean
in terms of resolution? I think that’s a good spec. But can we do that? So if I do a vertical
field of view, then assuming 128 points–
which is the top of the line of the current
[INAUDIBLE] line– I get a vertical field
of view 4.9 degrees. And that is to see
objects like road debris. If you hit those things,
you won’t be killed, but the risk that
you will damage your front face of the
vehicle is pretty high, so you will have to go to
the shop and repair it. And you don’t want to pay for
it if it was in automated mode. So around 10 centimeters
is what you have to see. But can you live
with a vertical field of even less than five
degrees in, for instance, San Francisco? I’m not sure. So if you have 150
meters, then you have five degrees
vertical field of view. If you say that,
OK, I can lower. I only need to see at 50 meters. Well, you still
only get 15 degrees. And you have to handle the
front loading of the vehicle and [INAUDIBLE] and so forth. So this is, I would say,
at least borderline. So in case they
use cases correct, we must increase the
number of vertical points or reduce the speed
you’re allowed to do in automated driving. But I cannot see that you can
drive it AD mode slower than the rest of the traffic. So that’s really a difficulty. So if you want to increase the
thing– and then also somebody says that you have to
do it for 100 bucks. And then, suddenly, things
start to become very difficult. And then transition
into automotive supply that we had some issues. This may be the
boring part for you, but it’s necessary
if you want to become an automotive supplier. But first, I want to take
this poem by Piet Hain. Danish theoretical physicist. “Put it up in the
paper it’s easy to see the cryptical admonishment,
T, T, T. When you feel how depressingly slowly you
climb, it’s well to remember that things take time.” And that’s something to remember
if you want to do automotive. It won’t be tomorrow. Automotive supply is basically
you have a bunch of component suppliers into a tier two. Maybe a model integrator, and
that supplies to tier one. And then it feed to the OEM. That’s the phase. If you want to
break that line, you are going to have difficulties. But you’re actually not
the component supplier. You’re part of a system. It’s part of a chain. So technology, your
matters and processes may be suitable for
small or modest volumes, but they may be
completely unsustainable at automotive scale. Your equipment may not scale. Your design tolerances are
not achievable at scale. Your specialized
suppliers may not have the means or
the willingness to transfer into automotive. And you may not get
the high volume yields that you hope to get in order
to get your cost perspective. And some automotive–
temperature, for instance. Usually minus 40, or plus 85. It can get worse. And that’s ambient
temperature, not component. And the vibration input. Some have compared
a vibration to– we are not mounting it next to
a jet engine, but it’s hard. EMI or EMC. And if you’re going to
have a lot of that thing and then you put
optics on top of that, it’s very difficult to contain
it with some metal enclosure, because you have a lens that
you want to see through. And then you have the
conductive disturbance. If you compare, for
instance, if you run a laser diode at high power, you
are pulsing high amps at megahertz rate. I call it a transmitter. The channel is pretty
boring, but it’s a transmitter in my view. And you have to handle
substance immunity. All kinds of things [INAUDIBLE]
high pressure cleaner. You have to be able to take
your high pressure cleaner and put it directly
onto the sensor. People don’t understand
that you should not do that. And if it fails, they go to
the shop and say, it’s faulty. Financial. You get a nomination, but
it’s not a formal order. It’s for cost. And then you actually get,
with short notice, then, when you’re going to ship. The OEMs see that I have the
right to review everything, and they do. Potentially, all
the high volumes will come with a
modest profit margin. It’s not pharmaceutical, or it’s
not Apple selling cell phones. And then they ask for price
reductions every year. They assume that you will
be more and more efficient, and they want to have
their cut of that. So if you’re unlucky,
your price may be lowered, and your total
cost of goods sold. And so you have to advance– what they call advanced pricing. And capital equipment
is amortized. And it’s also pretty
long programs, which means cash flow and
other things like that. So my suggestion is you
focus on your core strengths. Auto industry is very excellent
at making parts a commodity. And several suppliers can
provide similar parts, highly competitive. There are often competent
OEMs that actually scrutinize what you’re doing, and can see
that, yes, you can probably check this supplier instead. And why are you wanted to have
so much for this part when I think you can [INAUDIBLE]? So IP is a differentiation to
protect your profit levels. Many think that it’s better to
push higher up into the food chain because there’s more
money to be made there. But that also means
that your [INAUDIBLE] is going to be eaten up
by the commodity pricing that you probably are
not as good at handling. Big tier one suppliers
generally get the best prices at the same product,
if you produce it. Or if you produce
it, we are probably going to have a lower cost. So it may be more advantageous
that you license your IP, or you sell your
core technology. What are you really good at? Sell that one, and then
leave it to somebody else to produce the rest. But of course, it will probably
generate less profit total, but at a lot less risk and a
lot less capital requirement. As an example or qualification,
digital light recently came by Audi and Daimler. This is Daimler’s pictures. And you can see that you
basically have a camera, and you control the
light, and [INAUDIBLE] you block out [INAUDIBLE]. You basically drive on
high beam all the time. If you’re in the US, you
don’t know about this product, but it’s great. I have it in my third car,
and I wouldn’t even consider. I always drive on high
beam and let the camera take care of blocking
where it’s not needed, like approaching vehicles. And you can also
block out their face so you don’t blind
the pedestrian. And you can even predict
information on the road. So it’s a great product. And it’s based on
the [INAUDIBLE],, which has been in
production for– well, why not? OK, it’s been in production
for over 20 years, so it was a very mature product. So it’s probably a piece of cake
to transfer it into automotive. Well, they did a lot of
tests, but they actually used years and some iterations
before they actually mapped automotive qualification. I think [INAUDIBLE] has
the resources, experience, and other things like that
that produce other automotive. Finally, you have to have a
structured development plan. We call it [INAUDIBLE]. It’s risk management tool. It’s a workflow tool. And it’s also a way, by
the different functions, to check that you
meet your milestones and things like that. It’s a complex product,
but basically, OEMs, all the customers who will
monitor your progress based on this type of model. So if you don’t
like it, it’s tough. There are some
major deliverables. I won’t go into all of them. And they are required. But you also have to make
sure you have quality. Quality is a process. You design robust products. You buy flawless components, you
manufacture flawless products, and you verify conformity. And levels is a– you can do it by inspection,
but then you won’t get very far. You can then be reactive,
but you have a problem. You look at the root
cause, and you fix those. And you will get at
fairly decent levels. You may have good
problem-solving skills, and you share lessons
learned, and et cetera. But if you want
to go further, you have be proactive
and figure out where the quality problems may come. And then implement those. I have been with
the [INAUDIBLE].. They talk about
the quality issue, and I think that’s good, until
I heard the number percent. I thought we were talking PPM. We have a disconnect
here when they talk quality levels in percent. We are not really
accustomed to that. So problems is a– you have to visualize
them and find them. So when the water– basically, we had
to go to learn us how to do according to
other production assistants. When the water level is
high, problems are hidden, so you have to lower the water. And then you find the problems
and you take care of them, because they want to have
smooth flow and high harmony. That’s good. But then you lower
the water level, and then you find new problems. And you fix them. And then you lower
the water again, and fix the rest of them. Each problem is going to
be successively smaller, but they will always be there. And so you actually will
never be mission accomplished. Too bad for George W. Bush,
[INAUDIBLE],, by the way, declared mission accomplished. But actually, you’re never done. So the tale of the tail is
also get early production. At startup, you probably
don’t have much volumes that you can do production
statistics from. You expect some
[INAUDIBLE] incurred, and hopefully you can
live with that one. But look at the tail. That’s disaster
waiting for happen. And it may actually
happen then 100% of the time it happened that. So we actually idle
BMW for two months because the process of
making the package actually turned out to be
getting poor resistance. It shouldn’t have
been any changes, they said– the supplier–
but suddenly it was porous, and then the vacuum lost. So build sample size as
quickly as possible to try to find the corner cases. And we recommend that
cooperate with a tier one. It’s a good choice to
help you transition. And you improve your
efficiency immensely, and I think that’s
important, because you will be so much more competitive
in your other markets. And your competitors should
really worry about that. We use a Toyota
production system. And it may be the gold
standard, but it’s not that easy to implement. You can’t just read it up. There is books about it, but. And finally, a suitable
tier one is likely beneficial for both parties. So in summary, in conclusion,
lessons learned from [INAUDIBLE] is that
maybe relevant for LiDAR. Pay a lot of attention
to requirements, engineering early on. What do you really
want to accomplish? What is the core issues? Pay special attention to
things that you may not think is core functions,
like mounting, ambient exposure, vibrations,
EMC, or export control, or anything like that. You think that it’s your
sensor and your functions inside the box that’s important,
but many times, the problem is coming from the outside. So for LiDAR requires, make sure
you really address the LiDAR use cases of the [INAUDIBLE]. It will otherwise catch
up with you later on. And what it means
to automotive supply is that developing the
processes are essential. I recommend that
you find somebody to cooperate with to transfer
into automotive supply. And I’d like to thank
the number of people that have helped
me, also, to discuss our mutual experiences here. And thank you for
your attention. Any– [APPLAUSE] [MUSIC PLAYING]