Tesla Vision

[Hookalakupua]

Not sure if this has any bearing on the CT but I figured I’d let those of you with much more insight decide.
Yesterday, I received notice that my MY was pushed back from the end of Marco to April/May. Today, I received notice of the following:


Tesla Vision Update:
Replacing Ultrasonic Sensors with Tesla Vision
Safety is at the core of our design and engineering decisions. In 2021, we began our transition to Tesla Vision by removing radar from Model 3 and Model Y, followed by Model S and Model X in 2022. Today, in most regions around the globe, these vehicles now rely on Tesla Vision, our camera-based Autopilot system.
Since launch, we have continued to make incremental improvements in both feature parity and safety. Compared to radar-equipped vehicles, Model 3 and Model Y with Tesla Vision have either maintained or improved their active safety ratings in the US and Europe, and perform better in pedestrian automatic emergency braking (AEB) intervention.
Today, we are taking the next step in Tesla Vision by removing ultrasonic sensors (USS) from Model 3 and Model Y. We will continue this rollout with Model 3 and Model Y, globally, over the next few months, followed by Model S and Model X throughout 2023.
Along with the removal of USS, we have simultaneously launched our vision-based occupancy network – currently used in Full Self-Driving (FSD) Beta – to replace the inputs generated by USS. With today’s software, this approach gives Autopilot high-definition spatial positioning, longer range visibility and ability to identify and differentiate between objects. As with many Tesla features, our occupancy network will continue to improve rapidly over time.
For a short period of time during this transition, Tesla Vision vehicles that are not equipped with USS will be delivered with some features temporarily limited or inactive, including:

• Park Assist: alerts you of surrounding objects when the vehicle is traveling <5 mph.
• Autopark: automatically maneuvers into parallel or perpendicular parking spaces.
• Summon: manually moves your vehicle forward or in reverse via the Tesla app.
• Smart Summon: navigates your vehicle to your location or location of your choice via the Tesla app.

In the near future, once these features achieve performance parity to today’s vehicles, they will be restored via a series of over-the-air software updates. All other available Autopilot, Enhanced Autopilot and Full Self-Driving capability features will be active at delivery, depending on order configuration.
Given the incremental improvements already achieved with Tesla Vision, and our roadmap of future Autopilot improvements and abilities, we are confident that this is the best strategy for the future of Autopilot and the safety of our customers.

Please delete if irrelevant.

Sponsored

 

[ÆCIII]

There is one aspect of Tesla Vision that I'm hoping will be enhanced in the future, that is not directly related to the ultrasonic sensors really.

Watching FSD testing videos, one of the most difficult things the car deals with almost constantly, is assessing the creep line and deciding confidence to pull out from an intersecting street or side street. The car very often seems excessively hesitant when pulling out, due to lacking visibility. This happens especially when there are bushes, trees, walls, or obstructions on a corner making it difficult for the car to 'see' around them left or right.

The cameras in at the top of the windshield are in front of the driver, but they face forward and do not provide detailed views of the sides. The B-Pillar cameras are behind the driver, so they will definitely not be able to see around obstructions without creeping even further than the driver would have to, when checking if the way is clear. The B-Pillar cameras will not be able to see around obstructions as soon as a driver can. The fender repeater cameras are in front of the driver, but they face backwards for their own enhanced blind spot coverage.

I believe there should be an additional pair of cameras in the A-Pillar to assist seeing down range of each side when assessing clearance to pull out. They would be forward enough and yet high enough to be beneficial in those situations. Who knows though, maybe Tesla is considering something like that, since the recent tear-downs of the HW4 components had additional camera interfaces. But I do think A-Pillar cameras would give FSD more rapid visual assessment and better confidence when pulling out.

Just a couple thoughts I had on Tesla 'vision'.

Concerning the discontinuation of the ultrasonic sensors, I think the car could use some cameras in those areas too, so it was encouraging to see a camera low in the Cybertruck bumper.

Wonder if they'll just have filler plugs in the fascias where the sensors used to be, or will they start making new fascias without any sensor holes at all (then they'll have to stock both types of replacement fascias).

- ÆCIII

 

[fhteagle]

Agreed that another set of cameras further forward would be a smart addition. I'm not sure A pillar is even far enough forward. Depending on FOV and orientation, all the way on the front corners of the vehicle might be a better choice even.

 

[firsttruck]

Agreed that another set of cameras further forward would be a smart addition. I'm not sure A pillar is even far enough forward. Depending on FOV and orientation, all the way on the front corners of the vehicle might be a better choice even.

I agree. A-pillar is not far enough forward. The goal should be to see significantly better than human driver unless it becomes cost prohibitive.

There should be left/right cameras at headlights or at front bumper.

 

[flowerlandfilms]

Higher definition cameras, which Hardware 4 has, and wider angle lenses, which it might have, would do just as well as additional cameras I think.

 

[firsttruck]

Higher definition cameras, which Hardware 4 has, and wider angle lenses, which it might have, would do just as well as additional cameras I think.

I disagree. Higher resolution and wider angle cameras can do little when there are objects blocking the view. These cameras do not have Superman see through vision to detect what is on the other-side of wall, solid fence, trees, bushes, parked cars, vehicles in adjacent lane, etc.

 

[flowerlandfilms]

I disagree. Higher resolution and wider angle cameras can do little when there are objects blocking the view. These cameras do not have Superman see through vision to detect what is on the other-side of wall, solid fence, trees, bushes, parked cars, vehicles in adjacent lane, etc.

They have exactly the same ability to "see through" objects as additional cameras would.
But in higher resolution.
And a wider field of view.

 

[firsttruck]

They have exactly the same ability to "see through" objects as additional cameras would.
But in higher resolution.
And a wider field of view.

????

When I and many others talk about additional cameras we are talking about putting the new cameras in much more forward locations where the camera's view of far away objects would not be blocked by objects much closer to the vehicle. These objects can block the distant view of B-pillar camera yet even the human driver can see in the distance because the human drivers eyes even being inches more forward than the camera location in the B-pillar. Some drivers in some vehicles can lean forward to get vision to the left that is 12" to 18" forward of B-pillar camera.

A camera with higher resolution and wider field of view but in the same B-pillar location will still have it's view blocked as the current cameras. A camera with higher resolution and wider field of view in the same B-pillar location will mostly just have a better quality image of the object blocking the view and wider view around the object BUT not what is behind the object. The wider the object is and the closer the object is to the car the more of the distant view is blocked and higher resolution and wider field of view will not make a significant difference.

 

[flowerlandfilms]

What a camera sees is dictated by it's lens shape.
You can give it a lens that will allow it to see behind itself if you want to.
Fortunately you don't have to because multiple cameras all around the vehicle with multiple fields of view can stitch what they see together into a view that sees in all directions at once.
Something humans can not do.
Upon doing so the position of the cameras is largely arbitrary.
But increasing the resolution of the 360 view becomes increasingly helpful.

 

[firsttruck]

What a camera sees is dictated by it's lens shape.

as are lens size, cost, resistance to damage from vibrations and other restraints/factors.

Where on a vehicle, small lens is real important also being resistant to damage from vibrations, not being sensitive to wide temperature variations, and low cost for complete system.

You can give it a lens that will allow it to see behind itself if you want to.
Fortunately you don't have to because multiple cameras all around the vehicle with multiple fields of view can stitch what they see together into a view that sees in all directions at once.
Something humans can not do.
Upon doing so the position of the cameras is largely arbitrary.
But increasing the resolution of the 360 view becomes increasingly helpful.

270-360 degree field of view around camera is completely different from a small camera lens seeing behind a large object that is distant from the camera. A camera lens that can see behind a distant object is extremely large AND the field of view is reduced and distances are much more limited.

Also Tesla FSD is using the camera image to calculate the distant, direction & speed of objects so the see-behind camera's different types of distortions, fuzzyness, noise, and reduced light levels might make accurate movement calculations impossible.

All of this complexity being done to eliminate putting an additional $10-$20 camera in the bumper???


---------------------------------

Hypercentric optics: A camera lens that can see behind objects
Telecentric and hypercentric optics are very different from our eyes or normal camera lenses. They have "negative" perspective or no perspective, respectively, leading to very unusual images. In this video I show how to use a common fresnel lens in the creation of your own telecentric optical system.
Apr 28, 2020
Applied Science

-------



---------------------------------


Building a Next-Level Camera
Jul 30, 2022
DIY Perks

---------



---------------------------------

Sponsored