Sentinel, Luminar's Full-Stack Autonomous Solution - VoiceTube 動画で英語を学ぶ

字幕表動画を再生する

(warm synth hum)
- Hello and welcome.
My name is Ed Niedermeyer.
I am the co-host of "The Autonocast."
and the communications director
for Partners for Automated Vehicle Education.
I'm thrilled to be here today
with two fascinating companies,
the leaders of two fascinating companies.
First up in Orlando, Florida, Austin Russell,
CEO and founder of Luminar.
Austin, how's it going?
- Hey, great to see you.
Thanks for having us too.
Yeah, I mean, here today
in our advanced manufacturing facility in Orlando,
the new building we just moved into here.
Exciting to have, you know,
the first Iris units roll off the pilot line here.
- Great, and we're also joined
by Ödgärd Andersson, CEO of Zenseact.
Ödgärd, welcome.
- Thank you, I'm down in our lab here in Gothenburg.
Apart from a little bit of a throat problem,
I'm doing quite well.
- Awesome, can we just start with you,
and just talk about just what is this announcement?
What are you announcing today?
- Yeah, so after, you know, eight, nine years
in Luminar's history, I think we've largely been known
for producing really, the highest performing, you know,
cost-effective Lidar for series production applications,
and progressing on that to a stage
of where we can see this commercialized.
But in parallel, you know, we've been working
through a number of different software developments as well.
And, you know, what we're here today,
as part of this announcement is in partnership
with Zenseact, we're launching this full stack solution
to be able to now enable, starting with a Volvo,
a holistic autonomous vehicle stack that
for the first time can actually make its way
into series production, into production vehicles.
And excited to be here and to get a chance
to converse with yourself and, you know, Ödgärd,
and the whole team at Zenseact,
they're certainly really strong leaders
within the broader industry.
To be able to have an opportunity to work with them
is fantastic, and really uniquely positioned,
but we, you know, can talk more about that.
Yep, Ödgärd, anything else there?
- Yeah, I mean, we're super excited about this partnership,
and we've known each other now for a few years.
And basically, we set out to find somebody
who could make the best Lidar that
we can put in all our cars.
We really believe this is like the next step of safety.
And bringing the partnership now to this new level,
I think we're really excited to have the full offering.
I think it's going to be greatly received
when we first deploy it in the Volvo,
but also of course, after that point,
expanding above and beyond just the Volvo cars.
- Yeah, Ödgärd, could you explain sort of what,
what is this product, right?
We hear a lot about sort of level four robo-taxis
that are operating in fleets.
We hear about driver assistance in privately owned vehicles.
Having actual autonomous drive capabilities
in a production vehicle that people can go out
and buy is incredible, but sort of, where will this be used?
What's sort of the domain,
and the sort of use case for this product?
- Yeah, I mean, what we develop is like the software,
one pilot, it covers everything from, I mean,
computer vision all the way through decision and control.
And basically, it's one solution that you can either use
as the driver's support in conditions
when the driver needs to be in charge,
but also developing towards a highway pilots,
which basically, unsupervised, can drive for you
when you're under certain conditions,
and in highway scenarios.
And of course, we won't be able to go
exactly everywhere from day one,
but we will start and then expand from that point.
Both in terms of geography,
but also in terms of conditions.
- So Austin, at OEM Partners,
it sounds like we'll be able to sort of
have a range of options of what to do
with this technology stack, right?
And Volvo's the first, but sort of,
can you talk a little more about sort of the range
of things that, you know, using a Lidar-based system
for these kinds of privately owned vehicle functions,
automation functions?
What all does that enable in terms of that user experience?
- In terms of the capabilities, at the end of the day,
as Ödgärd mentioned, it's all about safety,
it's all about autonomy, saving lives, saving time.
And it really just makes a huge difference
when you can have this hardware,
and software solution deeply integrated.
There's so many times where you see these
different components developed in a vacuum,
but if you can have something that's a holistic solution,
it really makes all the difference.
And there's no question this is what a lot of folks
in the larger industry have been trying to do.
I think the key distinction here is actually though,
starting with the hardware, you have a foundation,
you have something that works, that delivers,
that isn't just, you know, one day theoretically trying
to make the physics work to have a product
that meets a spec, but having a real product, you know,
in an auto grade system that already meets this spec,
that's being deployed into production vehicles,
and then powering that, and supercharging it
with this software, that makes the vehicles
be able to have substantially greater safety capabilities,
as well in parallel, enabling autonomy,
starting on highways.
And that constrained highway autonomy problem,
as Ödgärd mentioned, highway pilot, is key to being able
to have a solvable solution in the relative near term.
You know, it's gonna take a long time by the time,
as you mentioned, robo-taxis are gonna be deployed
everywhere throughout the broader industry.
That's a decade-plus type problem.
This is something that's being solved for the next,
just to the next couple of years.
I mean, this is a deployed, this is gonna be available
on vehicles that you can buy, starting with Volvo,
and expanding outward.
So that's the distinction.
- Yeah, and just to clarify, so, at the high end,
this is, we're looking at like SAE level three,
conditional autonomy, is that right or?
- I think, you know, we've,
and Ödgärd and I are generally aligned on this, you know?
We largely try and avoid using the SAE, you know,
terminology just because, you know, there's,
it's kinda funny, you can see almost like,
level inflation going on, you know?
In the industry at this point right now,
and confusion about what does level three really mean?
What does that?
At the end of the day, there's really two modes of driving.
One is driver in the loop,
the other is driver out of the loop.
And the whole point here from an autonomous perspective
is to develop driver out of the loop capability.
So this would be the first time that someone
has a real, truly autonomous car,
much less something that you can buy.
But at the same time, while everyone is gunning for,
you know, those types of applications,
the safety side and the assisted driving side
should not be overlooked,
and that's where we're building these, you know,
what we're calling proactive safety features,
you know, into the broader system
to be able to actually start preventing accidents,
preventing collisions, preventing deaths,
even when the human driver is in the loop.
So, that's the story.
- Absolutely, and as someone who works
in an automated vehicle education,
I really appreciate your simplifying it, you know,
I think the SAE levels are great for what they are,
but sometimes for consumers,
it's just important to break it down,
and make it a little more simple.
- Totally, and by the way, on that note,
technically speaking, the kinds of capabilities
from a safety standpoint that we're talking about
are like level zero, you know?
Like, it's kinda interesting, like that,
which gets no attention, you know?
Normally in the industry.
But like, even just advancing these basic
safety capabilities that, you know,
people still get into accidents all the time,
even with these advanced driver assistance systems.
They help incrementally,
but they don't prevent accidents altogether.
And that's what makes all the difference.
And then, you know, obviously it goes all the way
up to level four, but again, yeah,
glad we can simplify it and clarify it for everyone.
- Yeah, and Ödgärd, can you talk a little bit
about why, sort of, what Lidar changes
in terms of what you're able to do on your end,
and maybe even sort of what co-development
sort of work, you know, doing this work together
in this kind of tight partnership.
What does that enable in terms of the final product?
- Yeah, I mean, if I starts with the Lidar,
we're basically trying to solve a quite difficult problem,
and you want to have the absolute best eyes
and ears on the car in order to do that.
And since we're aiming for the highway scenario,
we need to drive quite fast,
and then you need to see quite far.
And Lidar really fundamentally,
and especially working with Luminar,
and the performance of the Luminar Lidar,
we're able to see 250 meters,
which is really what you need in order to make this happen.
- Yeah, absolutely.
- So that's really the basic idea
behind starting to work with Austin,
and the team in the first place.
- And does doing the development work together,
once you have that, that partnership, you know,
what does that allow in terms of, you know,
really being tightly integrated in the development process?
- I mean, for us, it's a lot,
of course, learning between the teams,
but it's also about getting speed in the development loop.
So basically, now we're deploying with Volvo
as the first customer, we have a really
speedy development loop, collecting data,
getting kind of improvements in,
redeveloping, recollecting data.
And that just is unbeatable
when you can get that machine to work quickly.
- Yeah, and that speed is especially important when,
you know, there's a big difference
I think a lot of people don't always understand,
between sort of, you know, being out in development,
which a lot of autonomous drive technology is,
and having sort of test vehicles,
versus having something that is really,
and Austin, you mentioned the term automotive grade, right?
And I think sometimes people don't fully,
a lot of people don't fully appreciate
sort of how dramatic the difference is
between something that is automotive grade qualified,
and can be sold to private consumers,
versus something that's out on the streets, you know,
with trained engineers behind the wheel, you know,
ready to take over if anything goes wrong.
Can you just talk a little bit
about that distinction and sort of, you know,
what sets these two things apart?
- Yeah, I mean, we have, of course,
fundamental experience from our heritage within Volvo,
working with the active safety systems and deploying them.
And it's everything from how you get the software
into the car, and make sure it's safe
for all kinds of production vehicles.
But it's also about the user interface,
and how do you avoid any kind of mode confusion
for the user?
So they fundamentally know, as Austin mentioned,
is the car driving now, or am I driving now?
- Yeah, such an important thing.
And Austin, sort of on the, sort of under the skin, what is,
how different is an automotive grade, you know,
system like this from something maybe more experimental
that a partner who might be just sort of
in long-term development would be able to get together?
- Yeah, so, I think the key distinction here,
it is, you know, ultimately the holy grail
in this larger industry is to be able
to get into series production with a product, you know?
We've successfully done that on the on the Lidar side
in terms of what we have lined up, you know, with OEMs.
But now, you have to have the software side come together,
and work to really have this all play out.
And developing an auto grade solution
for both the software side equivalently
is absolutely critical, of being able to, you know,
have everything, it starts with everything
from the fundamental aspects of the organization,
and how you can, you know, get this, you know,
get your whole company, you know, IATF qualified
all the way to all the different specific standards,
functional safety systems, you know, ISO standards.
And, you know, you can just goes on and on and on
about what what you have to do to build this.
And the thing is is that this is a total transition
from kind of the traditional Silicon Valley hacker mentality
of let's throw something together, let's make it work,
let's just put a bunch of different sensors
on a roof rack here, who cares if it costs
a hundred thousand dollars or isn't you know, scalable,
let's just make it work at all costs.
And that's great for the development cycle of what's needed,
but the whole point here, what we're developing
is not a development vehicle.
This is a real, this is a real production product,
it's production hardware, it's production software.
It has to go into a vehicle then,
something that people's lives depend on.
So that's why it's so important.
And this is where in addition to our team of experts here,
where Zenseact really comes into play and shines, you know,
I mean, they have over 500, you know, different experts
on their team that have deployed software products
in the series production before.
I don't think there's anyone in the autonomous industry
that has anything close to that as a track record.
You know, generally it's much more focused
around the robo-taxi side specifically,
and not having something that
goes into a production vehicle.
And it really makes all the difference in terms
of actually getting something out there.
And by the way, you know, it's a complete solution.
If you're missing any single, like,
no puzzle piece in complete isolation, you know,
whether it be hardware, whether it be software,
whether it be Lidar, whether it be, you know, perception,
the rest of the full stack here for controls
and planning is useful in isolation, it has to be together.
So, that's what it takes.
- Yeah, absolutely, and Ödgärd, you know,
Austin just sort of referenced the fact
that there are so many levels to this,
that really, it becomes a question almost really of culture,
and, you know, your company, you're not just
have worked with automotive clients,
you've worked with, you know, an automotive client
that really sort of safety is at the core of their brand,
right, in Volvo.
Can you just talk a little bit about culture,
and sort of how that plays a role
in developing this kind of product?
- I mean, I really think a lot of people actually join us,
because, of course, the tech is cool.
It's like, you get to work on the coolest thing there is.
But I think most of the people that are here
are here because they know they're saving lives,
and it really, it's everything from what we talk
about in meetings, and reviews, to what we actually talk
about in the lunch room, or by the coffee machine.
It's really an enormous
sense of pride in what we're doing,
it's saving real people's lives out there.
And I think that's why we believe in AD so much,
it's because it's actually bringing all that tech,
that both in the unsupervised where the car drives,
but also adding all of that safety when somebody is driving,
it's going to be the next big step in safety.
And that's why we're here.
- Absolutely, and also, I mean,
do you have a sense of sort of,
is it possible to quantify what the potential
safety impact here is?
You know, what is your hope, at least,
for so what you can do with this
in terms of improving freeway level safety?
- Yep, yep, so, and I'd say there's
two different aspects of it.
One is kinda safety in a broader capacity,
actually beyond, you know, highways and freeway scenarios.
And then there's autonomy, which is starting out
applied specifically to the highway and freeways.
But you know, by our simulations, you know,
in terms of what can be enabled with this Lidar,
and this level of performance
that hasn't been seen before in the industry,
in an automotive series production application,
you're talking up to seven-X, you know, safety improvement.
That's not seven percent, which would actually be huge
on its own, but that's like seven-X.
I mean, there's just, there's still 1.3 million lives
lost on the road, you know, every year, thereabouts,
due to accidents that are totally preventable, you know?
It's like 94 or 95% of them are just
caused by pure human error, you know,
that's totally preventable.
And that's the thing, you know,
is that it doesn't actually take getting
to the proverbial drive anywhere anytime level five autonomy
to be able to start saving so many lives.
At the same time, you know, for the highway scenarios,
of course, in addition to improved safety, you know,
this is really the focus around the autonomous launch,
and the autonomous application,
and everything that we have here.
And with us, and then in partnership with Zenseact here,
is delivering this highway pilot system,
this autonomous stack really does make a huge difference
in terms of, frankly, just recovering time for the driver.
Right now, you take a look at, you know,
the existing quote unquote "full self-driving"
or equivalent type systems out there,
and they're not actually self-driving at all.
It's really just a lane keep assist system within, you know,
cruise control that adjusts depending
on the length of vehicle in front of you.
The whole point with this is that
you don't have to be constantly paying attention,
or ready to take over the wheel at any given moment
in a split second's notice.
You can actually take your hands off, eyes off,
relax, read a book, use your phone, work on your laptop,
watch a movie, take a nap, you know,
whatever it may be for a period of time.
And this is what is possible, this is what is enabled.
And again, starting on the highway scenarios,
expanding outwards over time.
But, you know, it's important to be realistic
about how you deploy this, where you deploy this
off the bat, and then expand from there.
- Absolutely, and Ödgärd, you know, one of the other
sort of issues we've seen with some of the more popular
sort of freeway driver assistance systems too
is something you touched on earlier, which is HMI.
And, you know, and particularly,
I think driver monitoring has been
a really important part of that.
What, you know, you mentioned that this
is sort of integrated in what you're doing.
How much of it are you all doing on the HMI side,
versus how much are you leaving to your OEM partners?
And, you know, is it important to have that,
that piece of it sort of at the really
fundamental system level?
- Yeah, I mean, of course, we work closely now
with our first customer, Volvo.
And they have quite a lot of data on this,
because they've done research for quite a long time.
And we do know that this is quite important.
So, while the HMI is always a bit different
between different OEMs, it's really important
that you don't end up in this kind of confusion
about what's going on.
- Absolutely, and Austin, I mean, so we're, you know,
we're announcing this partnership,
and the first partner is, or OEM,
who's going to be deploying it is Volvo.
But in terms of other, you stated you have aspirations
to get others on board as well.
Sort of what is your pitch to them in terms of, you know,
why should they come to you for this?
- To be frank, Volvo has really demonstrated
some very clear leadership, you know,
holistically in this space.
They're paying attention to the right things,
they're paying attention to safety,
they're paying attention to getting something
into real world series production.
This is an actual, not just branded as autonomy,
but actually can be driver out of the loop.
And that's what's possible to have enabled,
and is in series production, relatively speaking,
right around the corner.
So, it makes a huge difference.
And I think, I mean, when it comes down
to what's the pitch here,
first and foremost, let's state the obvious, you know,
do you want to have autonomous capabilities?
And do you want your customers to get into less accidents?
I think the answer on the consumer side
is absolutely to both of those,
which of course naturally should flow to the OEM side.
But the key distinction here, again,
is that when thinking about, you know,
different partnerships in the space,
there's no question there,
there's a whole landscape of AV companies.
The Waymos, the Cruises, the Argos, the Auroras,
like, all of these guys out there.
But the distinction, again, there's the focus
of robo-taxis with those kinds of partners,
not for production vehicles for OEMs.
And that's why we don't really, us,
or Zenseact here, as part of this solution,
we don't compete directly with those companies.
This is an entirely different dimension here
that we've been working through,
to actually build a real business behind this.
You know, this is something that's really materializing.
Again, it's not a hundred thousand dollar roof rack
full of sensing systems, and setups, and software
that's in R&D stage.
This is a production product that
we're going to like a series production with,
starting with Volvo again, very shortly here.
So, I think that's the key distinction.
And when it goes to other OEMs, there's a couple things.
One is, you know, having an opportunity
to have a lead partner for this,
that's taking the charge and really investing.
I mean, this is billions of dollars
worth of combined investment between, you know,
Zenseact and us to deliver this,
you know, this is serious stuff.
And to be able to have that backer,
to be able to have that put into series production
is key off the bat.
At the same time, again, you know, these are people that,
you know, they can actually trust.
Building trust in the OEM world,
and in the landscape is absolutely critical.
And you need folks that have
delivered stuff before, you know?
We're the first to recognize that, you know,
I think having the kind of skill sets that
the folks at Zenseact have is hugely complementary
to what we have, and, you know,
I'm sure Ödgärd can speak to it,
but you know, vice versa as well.
I'm getting a little bit of Silicon Valley DNA,
as well as, you know, of course, the hardware side of this,
and the Lidar, and combining that.
It makes all the difference.
So, that's the core of that, that's why OEMs, you know,
will adopt this holistic solution.
And being able to have it proven out,
and being able to really be the only one in the industry
that's capable of doing this.
You know, there are a few other in-house
developments at OEMs, but again, they're largely
entirely focused around robo-taxis,
not around production vehicles.
And that's why it's just so unique
to have this combination together.
And there's no one else in the industry like Zenseact,
and there's no one else in the industry like Luminar.
- Yeah, and Ödgärd, can you talk about what that's like?
I mean, I've certainly seen, we've seen this narrative
of Silicon Valley versus Detroit,
and it really seems like the future
kind of belongs to the folks
who get beyond that kind of confrontational approach,
and really try and bring the both, best of,
sorry, the best of both worlds together.
You know, can you just talk about what that's like
to sort of, you know, 'cause there is,
there is some cultural divide there, right?
Can you talk what it's like to sort of overcome that,
and sort of learn from each other?
- Well, I never lived in Detroit,
but I used to live in Silicon Valley for a while.
So, I think there are differences,
and there were a lot of similarities as well.
I think, obviously, we are learning quite a lot
from the Luminar team, and that I mentioned.
I believe we're also bringing some of our culture,
as we talked about before.
How do you think about actually getting a vehicle out?
What do you need to do in a certain order?
How do you make sure you have all the functional
safety systems built in from start?
Things like that.
But then, of course, this idea of continuously upgrading,
using data, iterating on the solution,
I think very much comes from the Silicon Valley heritage.
And I think that's obviously very good
if you combine the both.
- Absolutely, I think we're probably just about out of time,
but Austin, and I understand if you can't, you know,
talk specific price points, that's obviously
for your OEM partners to determine, right?
But you know, you sort of mentioned that, you know,
this is not a a hundred thousand dollar rack of sensors
like you see on top of a robo-taxi,
but, and oftentimes products start,
obviously, Volvo's a premium automaker, a premium brand.
And obviously, things start, tend to start premium,
new technologies, and work their way down.
Is this fundamentally a premium thing?
Is there sort of a path to getting the cost down,
where this could be something
that could be in just about any car?
- Yup, yup, so I'd say that the biggest cost barrier
to a lot of this stuff, you know,
when it comes to the hardware side, you know,
it really comes down to the Lidar.
We've solved that problem.
That's what makes a huge difference.
Then there's also getting the right setup,
getting the right sophistication
around all of these different things,
and having the economies of scale
to actually be able to make all of this happen,
'cause there's no way you're gonna get the cost
if you don't, you can't make anything work
if you don't get it into series production,
at the end of the day, economics-wise.
And then there's the software side of it too, you know,
there's, this is the kind of business
where there's huge upfront development costs,
but of course, huge backend returns that correspond to that.
So, that's that, when it comes down to it, you know,
in terms of the economics for this,
this is a solution that's, you know,
this is not hundreds of thousands of dollars.
This is not tens of thousands of dollars.
This, holistically, you know, all in,
this is single-digit thousands, you know,
for what you're talking about
for these kinds of autonomous capabilities.
Now, of course, you know, it really just depends
on what level of capabilities that you successfully enable.
And there's gonna be different price points accordingly to,
for example, if you have a base case of you,
where you're standardizing it across, you know,
a given OEM for a basic safety level capability,
it's gonna have a different kinda price point,
and pricing requirements, than it will for, you know
full stack all-in autonomous capabilities for everything.
And I think if there's anything that's been shown,
it's that there's huge consumer willingness to pay,
you know, for these features as well.
I mean, we've seen that there are serious dollars
that are commanded for just basic assisted driving features
that are out there today.
And this is taking it entirely to the next level,
actually recovering driver time, and at the same time,
making the vehicle substantially safer.
And that's where real value is unlocked,
where it's a strong value proposition, you know, for us,
and Zenseact, for the OEM,
for the consumer, really all around.
So, that's how we structure it.
Is that fair to say, Ödgärd?
- [Ödgärd] Yes.
- Great, well, Ödgärd, any final thoughts
that you'd like to leave people with?
I really appreciate you both taking the time to do this.
This has been absolutely fascinating,
I love learning more about how this all works,
but Ödgärd, any final thoughts about this
that people should take away?
- Yeah, I think one cool thing we're doing now
is actually building a really big brain
in terms of a big computer in the middle of the car.
And really, the best kind of sensors around it.
I think that's also an innovation system
that we've only started to scratch the surface
what we can tap into.
So that's, I think, the next step on this
is also enabling a lot of other innovation
on top of the system.
- Absolutely, and Austin, any final thoughts?
- Yeah, no, I think that's great.
The ecosystem is a huge part of all of this.
And you know, I mean, of course, the highlights are around
some of the key components around the Lidar system,
and the software associated with it, you know,
for the perception and the controls, planning, actuation,
some of the other stuff Zenseact is working on here.
But at the end of the day,
it is a broader ecosystem that's being provided,
and something that we get to leverage for what's
in play with Volvo here, now out to the broader industry.
I mean, so this is effectively an open invitation, you know,
for every OEM and frankly, a challenge to be able to,
can you have something that is dramatically improved safety?
Can you deliver autonomy?
And now they know where to go.
- Absolutely, well, I think that's a great,
great place to leave it.
Thank you both, Ödgärd Andersson, Austin Russell.
Again, thank you for talking me through this,
it's been really fascinating.
I've learned a lot,
and I hope everyone watching has learned as much as I have.
- [Ödgärd] Thank you, Ed.
- [Austin] Thanks so much, Ed.