Autonomous Car Progress

[mikes_fsd]

So with with just camera vision, Tesla has achieved what Google's FSD could do 10 years ago!

No, that is just what an ex Google employee who has a fetish for lidar says!

 

[2101Guy]

Tesla PSD 2020.40.8.11 57 minute video in North Carolina

Commentators have good comments like: This is like science fiction

i dont recall the millenium falcon almost running into the back of a parked car

 

[mikes_fsd]

i dont recall the millenium falcon almost running into the back of a parked car

You just have to know where to hit it!

 

[diplomat33]

No, that is just what an ex Google employee who has a fetish for lidar says!

He is simply comparing what Tesla's FSD can do on city streets with what he saw Google's FSD do with his own eyes when he worked there in 2010.

But you are deflecting from an important fact:

Autonomous cars like Waymo can drive off map but suffer a degradation in safety

Regardless of how you feel about HD maps, that is an important fact to know about autonomous cars that use HD maps.

 

[mikes_fsd]

Regardless of how you feel about HD maps, that is an important fact to know about autonomous cars that use HD maps.

Yes, anyone who uses HD maps relying on a lidar, cannot go off the map without killing somebody.
I got it!
/Sarcasm

On a serious note:
The next three to six months will determine the fate of HD maps.
You can take that to the bank!

 

[ZeApelido]

"But these don’t just occur at the corners, and finding most of them is a very long and difficult project. It took Waymo’s top-notch team a decade to do it. Nobody else has really done it."

In a small suburb in the Phoenix desert

 

[DanCar]

... The next three to six months will determine the fate of HD maps.
You can take that to the bank!

What makes you say that? I doubt Waymo will be in the people shipping business once Tesla's robotaxi service is wide. I'll estimate 4+ years for that. I think Waymo will be successful somewhere there, such as freight shipping. I also think Waymo will shine bright in years prior to Tesla starting its Robotaxi service. I expect Waymo to announce another city in service in the coming months.

 

[DanCar]

The lowest bar for Tesla to do FSD is highway traffic jam assist on limited access freeway. Lets say for example Tesla chooses 15 mph. Below this speed, Tesla is responsible. Above this speed Tesla wants you to be responsible. Seems like Tesla could implement this at level 4. If user doesn't take over, Tesla pulls over and parks. Tesla could enable this now in areas that aren't known for potholes, like Chandler, AZ. What is the risk for accident? Tesla has the data for number of disengagements. Would be great if we could see that data and more importantly how are disengagements changing over time.

@mspisars Why did you mark this funny?

 

[mikes_fsd]

What makes you say that? I doubt Waymo will be in the people shipping business once Tesla's robotaxi service is wide.

It has nothing to do with what business they will be in.
When you have gone a decade, with a set of beliefs that kept you from commercializing and expanding as you had originally planned.
Then in 4 years (announced in 10/2016) an "upstart" proves your "set of beliefs" to be without foundation.
There will be questions and long meetings.

The 3 to 6 months will determine the fate because if Tesla FSD shows rapid improvement (quantum leaps) in that time frame that means the approach is sound at the foundation.
There is no company that will want the additional overhead of the lidar hardware and the map creation and upkeep.
They all say it themselves. It's not like we want to have to generate these maps.
They couldn't engineer themselves out of it!

 

[mikes_fsd]

Lets say for example Tesla chooses 15 mph. Below this speed, Tesla is responsible. Above this speed Tesla wants you to be responsible. Seems like Tesla could implement this at level 4. If user doesn't take over, Tesla pulls over and parks. Tesla could enable this now in areas that aren't known for potholes, like Chandler, AZ. What is the risk for accident?

Dang it, I thought I had replied.
I found it funny because, I don't think that gels well with the "general solution" that Tesla is going for.
That would sound more like, we solved for "granny driving".
I do not see them handicapping and fragmenting their solution to such extent - especially after the years of being able to turn on AP (meant for highway only) anywhere that the system could read lane markings.

 

[DanCar]

... There is no company that will want the additional overhead of the lidar hardware and the map creation and upkeep. ...

Not that I agree to disagree with anyone but when talking to Waymo, they have told me that lidar cost is no big deal and 3D centimeter accurate maps are no big deal. A hardware engineer at Waymo, one lives in my neighborhood, told me that the bill of materials for Lidar costs less than $100 and Google makes their own. (This is from memory from conversation 4+ years ago, so I may have gotten something wrong.) Compare this to the costs of cameras that are under $2 for the sensor. Lens and other things increase the cost.

 

[DanCar]

Dang it, I thought I had replied.
I found it funny because, I don't think that gels well with the "general solution" that Tesla is going for.
That would sound more like, we solved for "granny driving".
I do not see them handicapping and fragmenting their solution to such extent - especially after the years of being able to turn on AP (meant for highway only) anywhere that the system could read lane markings.

Understood. I think most people would consider it a monumental step. Would make Tesla "Full" self driving real. Baby steps are good.

 

[mikes_fsd]

but when talking to Waymo, they have told me that lidar cost is no big deal and 3D centimeter accurate maps are no big deal.

This is why I also have doubts that Waymo will release anything to the public. Or sell as an OEM.

A decade gone, still no hints of being able to scale.
But "lidar and costs are not a problem" said the company that makes no money.

 

[FloridaJohn]

Looks like Mercedes has taken it's ball and gone home:

Mercedes-Benz throws in the towel on self-driving efforts: ‘We can no longer win’

 

[diplomat33]

Looks like Mercedes has taken it's ball and gone home:

Mercedes-Benz throws in the towel on self-driving efforts: ‘We can no longer win’

Smart move IMO. Mercedes can focus on making nice cars. Hopefully, they focus on EVs. And in a few years, Mercedes can just buy a "plug and play" FSD ready to go from several providers.

 

[DanCar]

Smart move IMO. Mercedes can focus on making nice cars. Hopefully, they focus on EVs. And in a few years, Mercedes can just buy a "plug and play" FSD ready to go from several providers.

Several providers: Aurora (Chris Urmson), Waymo?, Mobileye, Tesla?, Ford VW partnership (Argo), etc...

 

[diplomat33]

Several providers: Aurora (Chris Urmson), Waymo?, Mobileye, Tesla?, Ford VW partnership (Argo), etc...

Yes. I think in another few years, there will be several viable FSD systems that an automaker can pick from. You mentioned the main ones.

 

[mikes_fsd]

Several providers: Aurora (Chris Urmson), Waymo?, Mobileye, Tesla?, Ford VW partnership (Argo), etc...

We definitely need more than one!!

 

[Daniel in SD]

60 permit holders in California (with safety driver).

1. AIMOTIVE INC
2. AMBARELLA CORPORATION
3. APEX.AI
4. APPLE INC
5. ARGO AI, LLC
6. ATLAS ROBOTICS, INC
7. AURORA INNOVATION
8. AUTOX TECHNOLOGIES INC
9. BAIDU USA LLC
10. BMW
11. BOSCH
12. BOX BOT INC
13. CHANGAN AUTOMOBILE
14. CONTINENTAL
15. CRUISE LLC
16. CYNGN INC
17. DEEPROUTE.AI
18. DELPHI
19. DiDi RESEARCH AMERICA, LLC
20. EASYMILE
21. FORD
22. GATIK AI INC
23. HELM.AI INC
24. HONDA
25. IMAGRY INC
26. INCEPTIO TECHNOLOGY INC
27. INTEL CORPORATION
28. KAIZR, INC.
29. LEONIS TECHNOLOGIES
30. LYFT, INC
31. MANDO AMERICA CORP
32. MERC BENZ
33. NAVYA INC
34. NIO USA INC.
35. NISSAN
36. NURO, INC
37. NVIDIA CORPORATION
38. PHANTOM AI
39. PLUSAI, INC
40. PONY.AI
41. Qcraft.ai
42. QUALCOMM TECHNOLOGIES, INC
43. RENOVO.AUTO
44. RIDECELL INC
45. SF MOTORS INC
46. SUBARU
47. TELENAV, INC.
48. TESLA
49. THORDRIVE, INC.
50. TOYOTA RESEARCH INSTITUTE
51. UATC, LLC (UBER)
52. UDACITY
53. Udelv, Inc
54. VALEO NORTH AMERICA, INC.
55. VOLKSWAGEN
56. VOYAGE AUTO INC
57. WAYMO LLC
58. WeRide Corp DBA WeRide AI
59. XMOTORS.AI, INC
60. ZOOX INC

5 without a safety driver

1. AUTOX TECHNOLOGIES INC
2. Cruise LLC
3. NURO, INC
4. WAYMO LLC
5. ZOOX, INC

 

[DanCar]

I've stated a few times that Tesla wants to go fast with FSD, versus my impression of Waymo that is safety first (in other words slow). Here is something from Elon that insinuates this from 2018:
https://twitter.com/elonmusk/status/996133894893129728

It certainly needs to be better & we work to improve it every day, but perfect is enemy of good. A system that, on balance, saves lives & reduces injuries should be released.

I'm exaggerating a bit, but Waymo wants perfect rather than good.