CES 2020: Robosense Says It's Got The Secret Sauce For Autonomous Driving LIDAR (reupload)

Edit: This video is a reupload due to a strange video processing error on YouTube's end.

This video was made possible thanks to the kind donations of our supporters on Patreon and Ko-fi. Join them in supporting us.

-------
Continuing our CES 2020 coverage, we're sharing with you an interview with Dr. Leilei Shinohara, Vice president of R&D at Robosense.

Robosense is a Chinese-based company focusing on minituarizing and refining LIDAR and other sensor technology used in cars today for active safety feature systems and, eventually, autonomous driving systems.

We learn what's special about Robosense's latest new sensors, as well as the basics of how LIDAR works -- as well as get a ride in Robosense's demonstration vehicle. We also discuss why Tesla's visual system is currently being used over LIDAR -- and how Robosense hopes to change that.

Watch the video above, support us with the provided links, and leave your comments below -- but remember to keep your comments civil!

Presenter: Nikki Gordon-Bloomfield
Director of Photography: Brandon Yates
Camera Operator: Brandon Yang
Audio: Erin Carlie
Produced: Transport EvolvedCES 2020: Robosense Says It's Got The Secret Sauce For Autonomous Driving LIDAR (reupload)

-~-~~-~~~-~~-~-
Please watch: "2023 Kia Niro EV: Why You'll Want To Drive This"
-~-~~-~~~-~~-~-

[Wookey.]

Also at 7:36 he included bicycles in the list but you missed them out. It's annoying when people just forget about the bicycles in a transport list!
That'll let you update the captions.

[karlp]

A very good point was made: LIDAR is not trying to emulate the human eye or camera in terms of visual quality, it provides thousands of data points (shape, range, relative velocity etc) for a computer. A computer struggles to process a visual image - this provides what a computer needs to know. Cameras freeze over, get blinded by strong light, LIDAR doesn't have any of these problems because it's not "visual".

[NeilBlanchard]

I think that *full* autonomy is a long time into the future, and may not be feasible, or actually be needed.

Autonomy on marked route - sure.

[andrewpaulhart]

I’m with Elon. he goes back to the basics of physics. If we can determine objects and judge distances without lidiar there is no reason to believe that an autonomous system can’t do the same. It’s a more difficult engineering problem but you can see that Tesla is well on its way to solving it. They may be off in their timeline predictions by a few years, but legislation allowing autonomous vehicles won’t be put in place for years yet, so Tesla have plenty of time. Regarding night driving, this is a silly point. Passive cameras have way better night vision than human eyes. Also the point about lidiar systems being able to operate with cheaper computer systems is patently false because Tesla has already shown that it is possible to process camera input with their own very affordable FSD computer.

[CabrilloTV]

Nice interview with good array of questions, based on your interview I would take away this. These systems are getting better, but until the system can operate in all conditions with consistent reliability, they are not at a point where I could put the safety of my family in the hands of this technology as it is today.

[Follow_thedopamine]

Vision Systems are not held back by the lack of light, Hey can we pitch dark out and the vision system can utilize a different frequency other than visual light to see at night.

[davidarf]

You made a very important point in this report. Autonomy comes down to trust. Can we trust the mechanical systems - possibly and can we trust the software creators - unlikely. To back this up, one only needs to watch the NTSB Board Meeting on the accident involving the testing of an autonomous vehicle, or look at the reports on the design of the MCAS software on the Boeing 737 Max. In both cases the software design and implementation was a significant factor in the fatalities. It is currently an inevitability that corporate greed at the senior levels will result in systems being designed and implemented down to a price and safety will inevitably suffer. Look at the reports on the Boeing situation for indications of this. I think autonomy has a very long way to go to achieve trust and to prove that it can cope with any driving environment eventuality.

[toptobottom]

I'm personally not a fan of LIDAR unless you can make all the components disappear so I can't see them. Also, Nikki's last question wasn't accurate to what Elon of Tesla quoted. He said, " Lidar is only good right now in specific applications like Autonomy in a geo-fenced environment. He also said it was a bad idea for Tesla but he could use it for SpaceX applications.

[gregandkaruna]

They said a few years ago these solid state lidar were coming and here there are! And costs are falling fast.

[pjdambra]

Great Interview. Thank you for the subtitling. It sounds to me Elon may have to Re-evaluate his opposition to LIDAR. The cost is the game changer.

[BlazeCollie]

Smart LiDAR with YOLO object identification sounds AWESOME.

[CSHarvey]

So when there's a dozen lidar enabled vehicles at the same intersection.. Will they read each others laser illuminations?

[justinmallaiz]

It doesn't take an expert to notice the low resolution output, and lack of data displayed on screen compared to a cheap camera.. Perhaps you should of recorded this interview using lidar to prove how good it is...

[mrcet]

One of the advantages of traditional LIDAR systems in autonomous-vehicle applications has been that they have been mounted up high. That higher vantage point gives it a considerable advantage in detecting hazards, so mounting it unobtrusively could reduce its effectiveness!

[alaneasthope]

Interesting development. However, cameras are sensitive to infra red and have filters to use only visible light. If they utilise the infra red wavelengths, the cameras could see in the dark. I think this would be a cheaper option, but it's good to see people pushing the envelope.

[PepperPengu]

I think the best way to implement self driving vehicles is definitely using visual and LIDAR sensors in conjunction. And asolid state LIDAR system greatly improves the viability of these vehicles. Because I think even though cameras are “good enough” it cannot do a lot of things LIDAR can do

[truhunk]

Interesting, but full driverless won't happen for at least a few decades of longer. Sorry people, it will go from hype to bust. At least until computers can reason. Progress will be made in self driving, but not driverless.

[miketrebert]

I thought that Tesla's objection to LIDAR is that it's too expensive.

[patreekotime]

I'm still really not convinced Lidar is that great. The range isn't as far as a normal camera during the day, and about the same in the forward direction as a camera illuminated by street lights or headlights at night. It has the same, if not worse problems as cameras in bad weather and wet conditions. So the only winning edge case is that it can see further off of the sides of vehicles at night.

I've never really understood why traditional 'night vision' systems aren't in use more? Digital cameras can natively see in IR, and an IR 'headlight' isn't especially expensive, so the hardware costs are negligible, while the spatial detection software is virtually identical. And when the night-vision camera is blinded, it means the normal camera can see. The backup camera on most cars already does IR "night vision" mode with an IR emmiter array and automatically switches between modes. Again, it has the same downsides as Lidar in inclement weather... But it seems like it would offer the same upsides for dramatically cheaper? 🤷

[mgherdach]

Does it mean all taxi drivers are going to loose their job?