How much of Autopilot is Tesla and how much is straight Mobileye?

Tesla created the learning network that constantly updates the AP software. MobilEye has one too, but they are not the same.

It is my opinion that Tesla Motors has implemented the Mobileye system of sensors, cameras and the associated software libraries. Tesla and Mobileye have probably worked closely with Mobile's pre sales support consultants to port and implement the Mobileye bits as a Tesla solution on the Tesla hardware platform of the S and X (perhaps Nvidia is also involved in this porting effort). Additionally, from what I have read, Tesla Motors has created a proprietary back end database and learning algorithms that is not shared with Mobile's similar version of the database.

I did view a video of an early Mercedes or BMW beta of Lane Keeping. The UI was very similar to the Tesla driver's pod display with the exception of the automobile graphic. There was also a laptop that was showing camera views of the sensors' acquisition of the center and right hand lines and objects in nearby lanes. Additionally a running script log of events was scrolling. Darn it now but I have lost the link to that video. Sorry.

Firewired said:

I know that Tesla uses Mobileye sensors, but I had thought that Autopilot was Tesla's own proprietary software for the sensors. The video in this link looks very similar to Autopilot to me. So I was wondering is Autopilot essentially just the rollout of off the shelf Mobileye package with some tweaking, or is it truly unique software that Tesla is writing to utilize the sensors?

London welcomes Mobileye road safety technology | ISRAEL21c

Click to expand...

Mobileye has the special chips and video interpretation software. It appears Tesla is a layer on that.

Elon mentions HPP (holistic path prediction) as if it was Teslas but that is Mobileye. See this:
Tesla Autopilot Vs Competition

Good thread:

Juicy tidbits on Autopilot, deep learning, Tesla's advantage from Nov 3 Mobileye Call

Good thread: Juicy tidbits on Autopilot, deep learning, Tesla's advantage from Nov 3 Mobileye Call

Tesla uses the Mobileye eyeQ3 chip which decodes the video camera scene. It uses a static "deep learning" network (aka sophisticated neural network) to interpret the visual scene into cars, pedestrians, curbs, lane markings and other things. Tesla takes this information and then layers on top of it their own proprietary high resolution GPS map.

Near as we can tell (best info has been from the Elon closed conference call with reporters about Autopilot), Tesla gathers GPS paths from the entire Tesla fleet, does some processing on the paths and thus creates very accurate lane maps, such that they know the center point of each lane on a multi lane freeway.

So the Tesla lane keeping system uses both Mobileye visual scene decoding and Tesla high resolution GPS maps blended in some unknown way. We do know that the map information seems to update fairly rapidly unlike Mobileye which is static, doesn't change.

So far in the car industry, Tesla is the only manufacturer that has developed a system that allows the cars to learn from driver input, and learn fairly rapidly. This is a huge advantage, and will allow Tesla to remain the Autopilot leader forever unless other car manufacturers (or Mobileye) decide to do this too. But since no one else has even talked about having a car as connected as Tesla, I'm not holding my breath on the rest of the industry catching up anytime soon.

Mobileye does the vision processing. They normally work with an integrator, like Bosch, who does all the drivetrain and steering control logic. The integrator will then sell a complete package to car manufacturers, like Volvo and Mercedes, who can essentially drop in a plug and play system and be done.

In the case of Tesla, they act as their own integrator, doing all of their own drivetrain and steering control logic. In addition, Tesla augments the Mobileye output with their own lane-based GPS data as well as fleet learning.

Tesla is also in a unique position as an automaker to have a direct line of communication to Mobileye, as opposed to all others who have to go through their integrator for any changes.

EarlyAdopter said:

In the case of Tesla, they act as their own integrator, doing all of their own drivetrain and steering control logic. In addition, Tesla augments the Mobileye output with their own lane-based GPS data as well as fleet learning.

Click to expand...

Does this suggest that whenever autopilot cars lose the GPS signal, they instantly become less capable?

is the mobileye stuff much different than opengl/openCVf? I was looking into some of the graphics processing modules for doing traffic counts and it all seems very similar to what one of the the mobileye videos that someone posted shows. It is pretty interesting and seems relatively easy to image analysis, I'd imagine processing speed is the limiter

If you'd like some in depth information about what mobileye is doing (as opposed to what tesla is doing) watch this presentation from the CEO of mobileye where he explains how their technology works. He starts speaking around 4 minutes into the video. Brains, Minds Machines Seminar Series: Computer Vision that is changing our lives - YouTube

Forty Creek said:

Does this suggest that whenever autopilot cars lose the GPS signal, they instantly become less capable?

Click to expand...

I would characterize it as it just becomes as capable as other cars' lane holding.

It's pretty hard to loose GPS lock with modern GPS chipsets and externally mounted antennas, like on a car. Tunnels maybe.