Placeholder Image

字幕表 動画を再生する

  • - [Jason] Hi, my name is Jason Wojack,

  • I'm the senior vice president of product development,

  • here at Luminar.

  • Today. We're going to talk about Lidar design,

  • integration and performance.

  • The evolution that we're going through

  • in autonomous vehicles

  • is like a metamorphosis for the driver.

  • They're going to change the way

  • in which they interact with the car.

  • Human drivers are pretty amazing.

  • They have the ability to have situational awareness

  • that autonomous cars don't have right now.

  • Specifically ADAS especially.

  • So if you look at the,

  • the sign here on the left,

  • you can tell, hey,

  • there may be a deer coming soon down the road,

  • and a driver may, you know, change the way they're driving

  • based off of that.

  • The picture below, there's an icy bridge,

  • and that may affect

  • how somebody is going to approach that bridge,

  • or how they're going to make the turn on that bridge.

  • Predictability is very difficult in autonomous cars as well.

  • So if you look at the picture on the right,

  • you see that car has a, a mattress that may come flying off,

  • and it may affect where you drive,

  • or how you drive around that vehicle.

  • The user experience transformation that's going to happen

  • is going to be pretty dramatic.

  • It's going to be one from

  • hands on, eyes on,

  • focused,

  • you know, paying attention to everything that's happening to

  • hands off, eyes off, relaxed,

  • and more enjoying the experience.

  • So, when it comes to autonomous driving,

  • you have to put safety first.

  • So we're going to walk through some details around safety.

  • This is a study that was done by AAA a little while ago.

  • Basically went through, you know,

  • how current ADAS systems perform

  • when it comes to pedestrians.

  • And we're going to talk about

  • how we can progress and improve on that.

  • In current ADAS systems,

  • you have camera and you have radar.

  • But you need additional modalities

  • to solve some of these corner cases

  • that we're going to walk through.

  • This chart kind of overviews

  • some of the areas in which Luminar's Lidar,

  • helps with different areas where cars can't see right now.

  • So in this video that we're gonna walk through,

  • you can see there's something that shows up in the road,

  • but it's difficult for the camera to pick it up,

  • because of the overhead lighting,

  • and because of the shading in the entire area.

  • When you, when you look at this scene with Luminar's Lidar,

  • immediately you can see that's a ball in the road,

  • there's actually a small child off to the side,

  • that's hidden in the shade.

  • And you can see the major advantage that's given to you

  • by using Lidar in this situation.

  • In this case, we're going to look at

  • Lidar's role in different weather conditions.

  • here in the upper left-hand corner you can see

  • what the camera sees in fog.

  • The issue with cameras and fog here

  • is they mostly work off of contrast.

  • So the contrast that you see here,

  • it's difficult to make out

  • what's happening in the road very far up,

  • as you look through the fog.

  • However in the picture on the right,

  • you can see the division in the road,

  • this, this, the sign that's there

  • that's identifying the division in the road.

  • And lidar can kind of punch through that fog,

  • and see all that detail,

  • so the car can do its path planning.

  • In this next situation, we've got

  • some specific corner cases we're going to walk through.

  • So first we've got an image that shows

  • a couple of small objects in the road, a muffler, a tire,

  • some lamp version targets that we'll walk through.

  • The issue here is a small object, in a road,

  • at a very far distance.

  • This is something difficult for cameras to pick up right now

  • and radar.

  • So, for example, take that tire.

  • It's a dark color,

  • it's dark against the road,

  • and if it's at night, at a far distance, 150 meters,

  • the camera's not going to be able to see it,

  • and the car is not going to be able to see it.

  • But if you look at the, the point density,

  • and the distance at which Luminar's lidar can see

  • based off this chart,

  • you can see that all the way out to 150 meters,

  • we can still put points on that target,

  • identify that targets out there,

  • and start to make path planning decisions

  • based on identifying that there's an object out there,

  • and that we may have to watch out for it.

  • These are the kind of details that are required

  • to get true safety in autonomy.

  • Let's talk about field coverage a little bit,

  • in relation to different sensors and their locations.

  • So first, there's many different sensor configurations

  • that you can put on a car.

  • In these two examples, we have Highway Pilot,

  • where there's configurations with lidar looking forward.

  • On the right-hand side we've got full self-driving vehicles,

  • which need, you know, modalities all around the car.

  • In this example,

  • we're going to talk about the backup camera.

  • It's placed behind the car,

  • typically down near the license plate.

  • And it's there very specifically

  • because that's where the driver can't see.

  • It's a blind spot.

  • So the camera is able to pick up

  • whether there's a bike, or a small child,

  • or some somebody, somebody directly behind it.

  • And that's the reason why it's placed in that location.

  • In this next example, let's talk about ultrasonic sensors.

  • So typically they're placed

  • all around the skirt of the vehicle.

  • And this is for

  • situations like is shown here in parallel parking.

  • So the vehicle needs to see all around its edges.

  • Now let's talk about lidar,

  • specifically long range lidar like Luminar creates.

  • If that lidar is placed in the bumper

  • as is shown in one of these images,

  • you can see in the simulation here

  • that a lot of the view of the lidar

  • is obscured by the vehicles or other objects in front of it.

  • So a lot of the sensing that you need to drive

  • isn't really available.

  • In this next example,

  • if the lidar is placed at the roof of the vehicle,

  • you can see the dramatic difference in this simulation.

  • That even when there's vehicles directly in front of it,

  • it, it's able to see not only the cars around it,

  • and the cars in front of it,

  • but even the road in front of those cars.

  • So if you really want to get to true L3 level,

  • hands-off driving, you're going to need this kind of

  • ability to see out at a distance.

  • Kind of equate this to, you know, in this example, you know,

  • an ant, you know, down in the grass, versus a giraffe,

  • being able to see way above it.

  • You know, animals have eyes at the top of their head

  • for a reason, it's to be able to have that perspective.

  • It's also probably the,

  • the difference between driving on the highway in a go-kart,

  • or, you know, in a semi-truck.

  • You need that, that point of view.

  • This example shows

  • Luminar's lidar integrated into a roof line,

  • and the view at which it actually sees out on the road.

  • You can see how, how far the lane markings are marked off,

  • how much of the, the road surface that we can actually see,

  • and the distance at which we can see the vehicles.

  • Now let's talk about the sensors need to see.

  • And what, what's going to be required

  • to ensure that the lidar can properly see.

  • So there's many different design considerations

  • that have to be taken into account for sensor availability.

  • These are a few examples that radar and lidar in ADAS cases

  • have to already contend with.

  • So you've got rainy roads, snowy roads, gravel roads,

  • you know,

  • all kinds of different wet and inclement conditions.

  • Lidar has to deal with those same conditions,

  • and you have to design for it.

  • In this case, we're talking about

  • both the placement considerations

  • for where you're going to place a lidar,

  • as well as, you know, where the lidar can stay clean,

  • you know, the most easily.

  • So in the, the image on the left here,

  • you can see the laminar flow that's happening over the car

  • allows for the top region of the vehicle

  • to stay relatively clean and free from debris.

  • Down here at the grill, the vehicle near the front,

  • you can see where contaminants, and snow,

  • and other things are going to pile up,

  • and that'll cause you to use a lot more fluid

  • to keep the sensors clean and freely operating.

  • The end goal for lidar as is with other sensors

  • is for a touchless,

  • you know, cleaning and maintenance free solution.

  • We don't want the end user to have to worry about

  • whether the sensor is going to be available all the time.

  • There's many different things that we have to do

  • to ensure that that's going to happen.

  • The first one is heating.

  • We have to be able to heat up the window

  • to make sure it can defog, as well as, you know,

  • take ice or snow off of the window.

  • The next one is if you get dust, or dirt,

  • or other things on the window, you want to have, you know,

  • a nozzle for spraying air and water to clear those off.

  • And finally, potentially coatings indoor vibration to keep

  • the different contaminants, water droplets,

  • moving and off of the sensor.

  • So let's talk about form and function relative to the lidar.

  • Design integration of, of lidar and sensors in general,

  • you know, really started back for autonomous vehicles

  • in the DARPA challenge.

  • Back then it was, you know, engineers getting together,

  • throwing sensors on vehicles, and you know,

  • building the robots.

  • That's progressed slowly over time.

  • And you can see in the, in the TRI example,

  • they've taken the Luminar lidars,

  • integrated them into the roof,

  • and created a dramatically different

  • really designed in or scaled in

  • sensor package,

  • that looks dramatically different than it did back in the,

  • the DARPA days.

  • In 2022, Luminar is going to release

  • an integrated lidar solution with Volvo.

  • This, this lidar solution is at the roof line,

  • it's integrated in, and it's becoming relatively sleek.

  • I think the thing we want to highlight here is that

  • safety, you know, placing the,

  • the lidar at the roof of the vehicle,

  • and designed, don't necessarily have to be

  • mutually exclusive.

  • As the sensor design gets smaller and smaller,

  • we'll be able to integrate this in better and better.

  • And you'll still have your point of view,

  • and the design will become more and more seamless.

  • This is one of the main factors

  • that's going to enable a transformative experience

  • for drivers to be able to take their hands and eyes

  • off the road and enjoy the drive.

  • Thank you for coming to this special event.

  • If you'd like to learn more,

  • please visit luminartech.com for more information.

  • Thank you.

- [Jason] Hi, my name is Jason Wojack,

字幕と単語

動画の操作 ここで「動画」の調整と「字幕」の表示を設定することができます

B1 中級

The Crown | Presented by Jason Wojack

  • 1 1
    joey joey に公開 2021 年 05 月 24 日