Wiki Consumer AV - Status Tracking Thread

[EVNow]

2-4 months to map every road in the US would not be bad IMO.

AFAIK, lidar mapping would be similar to camera mapping.

I've not found good independent articles on what it really takes to do HD mapping. We need to understand how much manual labeling and labor is involved.

Anyway, @Bladerskb clarified that for UltraCruise GM would not be using HD maps. GM has also categorically stated doing HD Maps for all roads is impractical ...

 

[Bladerskb]

I've not found good independent articles on what it really takes to do HD mapping. We need to understand how much manual labeling and labor is involved.

Anyway, @Bladerskb clarified that for UltraCruise GM would not be using HD maps. GM has also categorically stated doing HD Maps for all roads is impractical ...

Not an independent article but here Mobileye shows us what HD map crowdsourcing from 1.5 million cars look like over a 12 months period.

https://twitter.com/x/status/1558446914365947904

 

[EVNow]

Not an independent article but here Mobileye shows us what HD map crowdsourcing from 1.5 million cars look like over a 12 months period.

https://twitter.com/x/status/1558446914365947904

No - not talking about MobilEye map - I mean lidar generated hd maps Waymo etc use.

BTW, what kind of map does Super Cruise use and how do they generate that ?

 

[Bladerskb]

No - not talking about MobilEye map - I mean lidar generated hd maps Waymo etc use.

BTW, what kind of map does Super Cruise use and how do they generate that ?

Dynamic Map Platform North America | DMP North America

We provide high-definition maps for advanced driver assistance, semi-autonomous and autonomous driving applications, and geospatial measurement solutions.

www.ushrauto.com

Ushr Powers HD Map Data for Cadillac Super Cruise™ — Ushr Inc.

Ushr Powers HD Map Data for Cadillac Super Cruise™ With 2018 Cadillac CT6, Ushr Becomes the First-to-Market HD Map Provider in a Production Vehicle DETROIT (Oct. 18, 2017) — Ushr Inc. , creator of high-definition (HD) mapping technology and software for autonomous and semi-autonomo

www.ushrauto.com

Also is there a reason why you are segmenting camera based hd map from lidar based hd map?

 

[EVNow]

Dynamic Map Platform North America | DMP North America

We provide high-definition maps for advanced driver assistance, semi-autonomous and autonomous driving applications, and geospatial measurement solutions.

www.ushrauto.com

Ushr Powers HD Map Data for Cadillac Super Cruise™ — Ushr Inc.

Ushr Powers HD Map Data for Cadillac Super Cruise™ With 2018 Cadillac CT6, Ushr Becomes the First-to-Market HD Map Provider in a Production Vehicle DETROIT (Oct. 18, 2017) — Ushr Inc. , creator of high-definition (HD) mapping technology and software for autonomous and semi-autonomo

www.ushrauto.com

Also is there a reason why you are segmenting camera based hd map from lidar based hd map?

MobilEye (and Tesla) can crowd source camera data ... Not lidar.

 

[diplomat33]

Nice look at XPENG's XNGP:

https://twitter.com/x/status/1656967452792684548

XNGP is basically a Chinese copy of FSD beta.

 

[diplomat33]

MobilEye (and Tesla) can crowd source camera data ... Not lidar.

But is Tesla actually crowdsourcing maps from the fleet? I think AI Day showed a clip of creating maps from crowdsourcing but I think they said it is for simulation and training. I could be wrong but I don't think Tesla is using crowdsourcing maps for driving like Mobileye is. Any info or correction would be appreciated. @Bladerskb

 

[EVNow]

But is Tesla actually crowdsourcing maps from the fleet? I think AI Day showed a clip of creating maps from crowdsourcing but I think they said it is for simulation and training. I could be wrong but I don't think Tesla is using crowdsourcing maps for driving like Mobileye is. Any info or correction would be appreciated. @Bladerskb

As I said Tesla can crowdsource. We don't know whether they do or not - see @verygreen tweets about extra map metadata Tesla sends to the cars once a destination / route is entered. Its a low hanging fruit and my bet is Tesla is atleast working on it (if not already doing something about it).

 

[EVNow]

XNGP is basically a Chinese copy of FSD beta.

LOL - copy indeed

Tesla settles lawsuit against engineer who it claims stole Autopilot source code for Chinese competitor

Tesla has settled a lawsuit against a former engineer who the company claims stole its Autopilot source code before going...

electrek.co

Behind Tesla's stunning suit against Xpeng · TechNode

Tesla is asking for an awful lot of information from Xpeng, from a total look at its source code to the hard drives of its top executives.

technode.com

 

[diplomat33]

I think this is the right thread for this.

Journalists in China did a comparison test of several consumer L2 systems. Mobileye's SVO says SuperVision "won". It is in chinese. The first column that says 001 is the Zeekr001 with Mobileye's SuperVision. The first row seems to be compute power since it says TOPS. The little steering wheel is probably driver interventions and the last one is trip time. Not sure what the other ones are. I am thinking the arrow pointing to the line is typically the symbol for downloading so it might refer to downloading maps or going mapless as Johann mentions.

https://twitter.com/x/status/1687571937885552640

@Bladerskb can you help? You seem knowledgeable of these systems in China. Thanks.

 

[diplomat33]

I got answer from reddit.

The rows from top to bottom are:

car model
Compute
Lidar
Radar
Cameras
Downgrades
Interventions
Trip time

Here is the chart in English. Trip was 66 km in Shanghai.

Car Model	Zeekr001 (Mobileye SuperVision)	XPeng G9 570 Max	NIO ES6	Li Auto L8 Max
Compute (TOPS)	48	508	1016	508
Lidar	0	2	1	1
Radar	1	5	5	1
Camera	15	14	13	13
Downgrade(s)	1	1	2	5
Intervention(s)	1	1	2	7
Trip Time	1:06:51	1:12:31	1:21:06	1:21:53

Video with clips from testing: 微博

 

[scottf200]

I got answer from reddit.

The rows from top to bottom are:

car model
Compute
Lidar
Radar
Cameras
Downgrades
Interventions
Trip time

Here is the chart in English. Trip was 66 km in Shanghai.

Car Model	Zeekr001 (Mobileye SuperVision)	XPeng G9 570 Max	NIO ES6	Li Auto L8 Max
Compute (TOPS)	48	508	1016	508
Lidar	0	2	1	1
Radar	1	5	5	1
Camera	15	14	13	13
Downgrade(s)	1	1	2	5
Intervention(s)	1	1	2	7
Trip Time	1:06:51	1:12:31	1:21:06	1:21:53

Video with clips from testing: 微博

It'd be hilarious to see Tesla FSD in this test. ie. it'd fail a lot based on my experience.

 

[diplomat33]

https://twitter.com/x/status/1689350928296718336

 

[scottf200]

https://twitter.com/x/status/1689350928296718336

I wonder how Tesla will transition to this "Elon Mode" when they decide to (ie. some limited roads). Based on my road trip this summer, they 100% canNOT do it with the current FSDbeta on highways (it made many mistakes and some dangerous for me). However, for most interstates, Tesla likely could go hands-free.

 

[diplomat33]

 

[scottf200]

Everyone use to be in awe of Tesla displays. Other companies are doing it and it looks maybe better. Lot of complex tracking going on in this area of the video.

 

[diplomat33]

Good article in MIT Tech Review about the different Chinese versions of "City NOA".

A race for autopilot dominance is giving China the edge in autonomous driving

Electric vehicle makers and AI companies are taking Tesla FSD-like systems to China, but it’s still out of reach for most consumers.

www.technologyreview.com

 

[scottf200]

Good article in MIT Tech Review about the different Chinese versions of "City NOA".

A race for autopilot dominance is giving China the edge in autonomous driving

Electric vehicle makers and AI companies are taking Tesla FSD-like systems to China, but it’s still out of reach for most consumers.

www.technologyreview.com

High level of the auto companies and their software nomenclature.

"... Chinese car companies have given their NOA products all kinds of misleading or meaningless names:

• Li Auto follows Tesla’s tradition and calls it NOA
• NIO calls it NOP (Navigate on Pilot) and NAD (NIO Assisted and Intelligent Driving)
• XPeng calls it NGP (Navigation Guided Pilot) and more recently XNGP (the “last step before full autonomous driving is realized,” the company says)
• Huawei calls it NCA (Navigation Cruise Assist)
• Haomo.AI, an AI startup, calls it NOH (Navigation on HPilot)
• Baidu calls it Apollo City Driving Max

... Some might only cover the major beltways in a city, while others go into smaller streets; some use LiDAR (a laser-based sensor) to help improve accuracy, while others only use cameras. And there’s no standard on how safe the tech needs to be before it is sold to consumers."

 

[Bladerskb]

High level of the auto companies and their software nomenclature.

"... Chinese car companies have given their NOA products all kinds of misleading or meaningless names:

• Li Auto follows Tesla’s tradition and calls it NOA
• NIO calls it NOP (Navigate on Pilot) and NAD (NIO Assisted and Intelligent Driving)
• XPeng calls it NGP (Navigation Guided Pilot) and more recently XNGP (the “last step before full autonomous driving is realized,” the company says)
• Huawei calls it NCA (Navigation Cruise Assist)
• Haomo.AI, an AI startup, calls it NOH (Navigation on HPilot)
• Baidu calls it Apollo City Driving Max

... Some might only cover the major beltways in a city, while others go into smaller streets; some use LiDAR (a laser-based sensor) to help improve accuracy, while others only use cameras. And there’s no standard on how safe the tech needs to be before it is sold to consumers."

One thing is clear, no one calls it "self driving" let alone "fully self driving".

 

[MarcG]

One thing is clear, no one calls it "self driving" let alone "fully self driving".

There can be only one (Elon)