FSD rewrite will go out on Oct 20 to limited beta

[powertoold]

Tesla doesn't define "HD maps". There are industry-accepted definitions of HD maps, and there hasn't been any evidence that the FSD beta ever used HD maps.

Some examples below:

End-to-end HD Mapping for Self-Driving Cars from NVIDIA Automotive

HD map | TomTom

Journal article description below:

Autonomous Driving in the iCity—HD Maps as a Key Challenge of the Automotive Industry - ScienceDirect

The necessity of HD maps for autonomous driving becomes obvious when trying to anticipate appropriate behavior in traffic beyond the sensors of a robotic car. To date, there is no sensor available that can localize and determine a car in reference to its surroundings. Although a few research projects are working on a solution based on quantum positioning, they are still in early stages, and the first prototypes are larger than a trunk of a station wagon. It will take a few more years before this quantum physics technology is ready for use [7]; therefore, autonomous driving relies on dynamic HD maps. Localization is ensured by the mapping of HD map data with landmarks such as buildings as reference positions. The position of a car in relation to various landmarks provides an exact localization. The next move is the integration of HD mapping in series of cars [8].

The sensor data collected by thousands of cars allows the establishment of dynamic HD maps with high accuracy and real-time information. With this kind of self-learning cloud service, the latest updates can be sent to all registered cars over a 4G cellular connection. The more cars participate, the more precise the HD map cloud solution will become.

 

[alstoralset]

Tesla's internal map resolution is actually fairly low, but they up-sample and interpolate the data for improved aesthetics.

Here's a view of the raw data:

 

[verygreen]

If you miss it ... it's the large "No HD Maps" in red at the bottom of that image...

I know.

We already discussed this. Tesla does not consider maps they use to be HD maps. I even referenced this in my original tweet on the topic.

 

[heltok]

As this thread now is about our favorit topic HD map, I remembered that I used to be a part of the group defining the ADASISv3 standard. Unfortuneatly Tesla was not a member of this group:
Members | new adasis website

Just to catch up a bit, I found this video, maybe you HD vs SD maps experts will like it:
How HD maps and ADASIS v3 can extend the vision of automated vehicles | Automotive World

Seems like not much has changed since I was involved ~4 years ago. So I guess Tesla was right to not be a part of it and instead do their own solution.

 

[Bladerskb]

Tesla doesn't define "HD maps". There are industry-accepted definitions of HD maps, and there hasn't been any evidence that the FSD beta ever used HD maps.

Some examples below:

End-to-end HD Mapping for Self-Driving Cars from NVIDIA Automotive

HD map | TomTom

Journal article description below:

Autonomous Driving in the iCity—HD Maps as a Key Challenge of the Automotive Industry - ScienceDirect

The necessity of HD maps for autonomous driving becomes obvious when trying to anticipate appropriate behavior in traffic beyond the sensors of a robotic car. To date, there is no sensor available that can localize and determine a car in reference to its surroundings. Although a few research projects are working on a solution based on quantum positioning, they are still in early stages, and the first prototypes are larger than a trunk of a station wagon. It will take a few more years before this quantum physics technology is ready for use [7]; therefore, autonomous driving relies on dynamic HD maps. Localization is ensured by the mapping of HD map data with landmarks such as buildings as reference positions. The position of a car in relation to various landmarks provides an exact localization. The next move is the integration of HD mapping in series of cars [8].

The sensor data collected by thousands of cars allows the establishment of dynamic HD maps with high accuracy and real-time information. With this kind of self-learning cloud service, the latest updates can be sent to all registered cars over a 4G cellular connection. The more cars participate, the more precise the HD map cloud solution will become.

Hd maps consists of geometeric and/or semantic maps. While Tesla HD map doesn't have a geometric layer. Their map does have a semantic layer.

Rethinking Maps for Self-Driving

 

[powertoold]

Hd maps consists of geometeric and/or semantic maps. While Tesla HD map doesn't have a geometric layer. Their map does have a semantic layer.

Rethinking Maps for Self-Driving

Tesla fsd beta doesn't use HD maps, period.

From your article:

These HD maps need to represent the world at an unprecedented centimeter resolution, which is one to two orders of magnitude greater than the roughly meter level resolution that web map services offer today.

 

[helvio]

Once and for all, let's end this topic. The definition of HD maps is "HIGH". Basic grammar, people!

 

[DanCar]

The higher the definition the better.

 

[mikes_fsd]

Once and for all, let's end this topic. The definition of HD maps is "HIGH". Basic grammar, people!

Just one more time before 2020 runs out!

 

[yerEVan]

Tesla's internal map resolution is actually fairly low, but they up-sample and interpolate the data for improved aesthetics.

Here's a view of the raw data:

View attachment 618693

Looks like my apple ][ monitor.

 

[Bladerskb]

Tesla fsd beta doesn't use HD maps, period.

From your article:

These HD maps need to represent the world at an unprecedented centimeter resolution, which is one to two orders of magnitude greater than the roughly meter level resolution that web map services offer today.

Again it seems like you don't undersand.
HD map is just a term that covers multiple layers.
The only layer that requires CM accuracy is the geometric layer.
The semantic layer isn't cm accurate nor does it need to be.

 

[Silicon Desert]

BEV view. .... Battery Electric Vehicle view? Too many TLAs here. (three letter acronyms )

yes, agree, so maybe I can add a 12 char one.
EWTHAWGTGFSD

Elon When The Hell Are We Going To Get Full Self Driving

 

[Silicon Desert]

It's like those stupid tags on pillows and mattresses that say "Do not remove until penalty of law" or something to that effect. Those warnings have nothing to do with the consumer but they are there for a very discrete reason.

Oh my God.... now you tell me. I must have removed hundreds of then over the decades. I'm going to prison for life

 

[powertoold]

HD map is just a term that covers multiple layers.

Except it's not.

Any map with layers != HD map. Please stop with this nonsense.

 

[EVNow]

Again it seems like you don't undersand.
HD map is just a term that covers multiple layers.
The only layer that requires CM accuracy is the geometric layer.
The semantic layer isn't cm accurate nor does it need to be.

Here comes the HD Map brigade.

The basic question is scalability. Tesla usage of maps is highly scalable, Waymo's is not. End of story.

 

[Cheburashka]

Well it's been 2 months since the "beta" was released.

While I'm happy for the 5 people that actually have it, it'd be nice to see if Tesla can use what they learned during the beta to make progress for the common folk like us.

If we don't see something big next week, I'll be disappointed.

 

[n.one.one]

The magic behind FSD visualizations - 15:08 - AIDRIVR

 

[drtimhill]

Once and for all, let's end this topic. The definition of HD maps is "HIGH". Basic grammar, people!

But how far is up?

 

[powertoold]

Looks like FSD has toll booth prediction (from green's video):

https://twitter.com/greentheonly/status/1340821815124496385

 

[thewishmaster]

Looks like FSD has toll booth prediction (from green's video):

View attachment 619631

https://twitter.com/greentheonly/status/1340821815124496385

I’ve seen tollbooth-related messages and reactions (eg car slowing down) for a long time. Not sure if it would stop if I needed to pay cash though - I always use the electronic payment lanes.