Tesla to replace parking sensors with vision

Tesla to replace parking sensors with vision

Author
Discussion

egomeister

6,701 posts

263 months

Thursday 6th October 2022
quotequote all
Funkstar De Luxe said:
It's not about cost saving - well not in parts, anyway. It's about trying to reduce the complexity of systems and the complexity of the electrical platform.

Cars are becoming incredibly complex - not only in terms of features but in terms of configurability and modularity. If you have two modules that are essentially replicating features (even if it requires more software) it's the best option to do so.
I don't buy this. So many of Teslas decisions are about cost saving and manufacturing simplicity irrespective of whether it is the better engineering/customer solution. The big touch screens mean will be a huge cost down on a typical interior and likewise the simplifications seen on the 3 over the S. Similarly, on the driving assistance side of things they do everything they can to delete stuff like radar or the parking sensors, whereas others utilise tech like LIDAR to get better data in.

Some of the solutions I've seen on Teslas are very neat, and if they pull off vision only self driving it will be a gamechanger but it feels to me like its about getting the profit per unit up and pushing the design/software teams to deliver a viable solution within those constraints. I think they can get to a minimum viable product situation with vision only, but it will put a cap on what they are ultimately able to achieve compared to those who use a suite of sensors.

CheesecakeRunner

Original Poster:

3,807 posts

91 months

Thursday 6th October 2022
quotequote all
Funkstar De Luxe said:
It's not about cost saving - well not in parts, anyway. It's about trying to reduce the complexity of systems and the complexity of the electrical platform.
Thing is, they’re replacing it with incredibly complex software, and it’s a fallacy that software is cheaper to create than hardware.

Feels to me like it’s a philosophical decision by Tesla, rather than a cost one. “Human work fine with binocular vision, therefore machine can to”. Ignoring the idea that perhaps it would be better to use more senses than a human has, for improved performance in tricky conditions.

RobbyJ

1,570 posts

222 months

Thursday 6th October 2022
quotequote all
CheesecakeRunner said:
Thing is, they’re replacing it with incredibly complex software, and it’s a fallacy that software is cheaper to create than hardware.

Feels to me like it’s a philosophical decision by Tesla, rather than a cost one. “Human work fine with binocular vision, therefore machine can to”. Ignoring the idea that perhaps it would be better to use more senses than a human has, for improved performance in tricky conditions.
Exactly, a human can't easily see though fog or a downpour or for that matter at night with Tesla headlights! Not that I'm suggesting flying into pea soup fog counting on radar but every little helps. Radar for me very much has a place.

Edited by RobbyJ on Thursday 6th October 19:01

DJP31

232 posts

104 months

Thursday 6th October 2022
quotequote all
RobbyJ said:
They are seemingly taking a lot of backward steps for things that I personally care about. If for example I bought a new Model S when they start coming to the UK I'd very much miss:

A full steering wheel (if the stupid yoke makes it to the UK)
An indicator stalk
A gear selector stalk
Radar
Parking sensors
Cruise/autopilot stalk

They aren't seemingly making things better, they are reducing cost in really key areas that you interact with constantly. This at a time when many other manufacturers are coming out with some very very compelling alternatives. I really like my Model S but if the stuff in my list above wasn't there, especially the steering wheel I'd very much consider my options when it's time to change.
You and me both, but I'm struggling to find a different car that delivers the whole package as well as the Model S does.

RobbyJ

1,570 posts

222 months

Thursday 6th October 2022
quotequote all
DJP31 said:
You and me both, but I'm struggling to find a different car that delivers the whole package as well as the Model S does.
Same, there's nothing in the UK that ticks all the boxes the S does, not an SUV, unbeatable luggage space, 3 good seats in the back, very fast, handles pretty well for a barge.

Unless something else that ticks those boxes comes along my next car will probably be an S too. Before someone else pipes up no I don't want to go back to ICE, and even my RS6 didn't compare for back seat and luggage space.

gangzoom

6,303 posts

215 months

Saturday 8th October 2022
quotequote all
DJP31 said:
You and me both, but I'm struggling to find a different car that delivers the whole package as well as the Model S does.
Is there a 'need' to change? Our 2017 X just passes its 3rd MOT with no issues. It works better now than new, with higher DC peak charging rates, and CCS access.

The only question for longterm ownership is the traction battery. Pretty clear now ALL 2015 cars will need a new battery sooner rather than later due to some inherent design faults, the pack design of post 2016 cars is better but longterm data still isn't there, apart from 90 packs degrade like crazy till about 85% health.

Luckily the 'new' 90kWh pack from Tesla is essentially a 100D pack with module removed, so it's essentially the best battery pack Tesla makes right now. Prices started out at $25k+ 18 months ago, but US owners are now been quoted $17-18k, so prices are falling. Tesla have also appear to have 'run out' of 'remanufactured' packs. Hopefully by 2025 when our X is 8 years old prices will be closer to $15k, at which point these cars really could last 'a lifetime'.

Given the aluminium chassis, and all other suspension/drive train is essentially Merc parts - air suspension apparently even shares Merc part numbers, longterm upkeep seems OK.

Edited by gangzoom on Saturday 8th October 04:11

CheesecakeRunner

Original Poster:

3,807 posts

91 months

Saturday 8th October 2022
quotequote all
gangzoom said:
DJP31 said:
You and me both, but I'm struggling to find a different car that delivers the whole package as well as the Model S does.
Is there a 'need' to change?
Can’t speak for DJP31 but my Model 3 is a lease and is up next December so I do have a ‘need’ to change. I imagine I’m not the only one.

Also, this change to vision only is going to be retro active to cars that have radar, including your X. And then the radar will be disabled. So unless you stop doing software updates, you’re getting it too.

Register1

2,142 posts

94 months

Saturday 15th October 2022
quotequote all
z4RRSchris said:
says cars after october 22, so mine should be delivered in december so maybe ill get a radar car?
Our M3 SR+

is due between November 20 - December 6

Hoping all is present and correct by then.

annodomini2

6,862 posts

251 months

Saturday 15th October 2022
quotequote all
Register1 said:
z4RRSchris said:
says cars after october 22, so mine should be delivered in december so maybe ill get a radar car?
Our M3 SR+

is due between November 20 - December 6

Hoping all is present and correct by then.
This month they plan to transition. So it depends if yours is already on the ship.

annodomini2

6,862 posts

251 months

Saturday 15th October 2022
quotequote all
egomeister said:
Funkstar De Luxe said:
It's not about cost saving - well not in parts, anyway. It's about trying to reduce the complexity of systems and the complexity of the electrical platform.

Cars are becoming incredibly complex - not only in terms of features but in terms of configurability and modularity. If you have two modules that are essentially replicating features (even if it requires more software) it's the best option to do so.
I don't buy this. So many of Teslas decisions are about cost saving and manufacturing simplicity irrespective of whether it is the better engineering/customer solution. The big touch screens mean will be a huge cost down on a typical interior and likewise the simplifications seen on the 3 over the S. Similarly, on the driving assistance side of things they do everything they can to delete stuff like radar or the parking sensors, whereas others utilise tech like LIDAR to get better data in.

Some of the solutions I've seen on Teslas are very neat, and if they pull off vision only self driving it will be a gamechanger but it feels to me like its about getting the profit per unit up and pushing the design/software teams to deliver a viable solution within those constraints. I think they can get to a minimum viable product situation with vision only, but it will put a cap on what they are ultimately able to achieve compared to those who use a suite of sensors.
It's about cost saving, Munroe and Associates reckon it'll save about $150 per car, which is a lot in Automotive terms.

They will apply it to existing vehicles as well. They've done it with the Adaptive Cruise Control, mine has radar, but it no longer uses it.

RichardM5

1,739 posts

136 months

Saturday 19th November 2022
quotequote all
The big issue I have with just using Vision is that, particularly at this time of year, it does not work very well!

The cameras in Tesla's don't have infra red, so when it's dark they don't work (multiple cameras blocked message etc). When the sun is low in the sky, they don't work because they are blinded. When it's foggy they don't work well. When the camera is dirty after a long drive in the wet, they don't work because they are obscured. Often when reversing there is significant delay with the displayed video, many times if I had relied on the camera view when reversing I'd have hit something.

This just seems to be a really bad idea, just like the auto wipers, which in my Model 3 are utter garbage and just don't work as well as those in my 20 year old BMW.

Hans_Gruber

275 posts

171 months

Sunday 20th November 2022
quotequote all
The signal from radars can also be blocked by dirt, snow and ice. Radars have a very limited field of view to achieve range, where a camera ( with recognition software) has a much wider field of view.

Radar has reached its limits. Recognition software is the way to progress. It should never replace the mark one eyeball and no one should ever drive faster than a reasonable distance you can see anyway. In the same way our eyes have a wide field of view and we can differentiate a child running out from the side of the road to a stationary lamppost, recognition software can make up the shortcomings of an inattentive driver.

If you’re unconvinced, take a look at some of the latest ncap independent tests, when protecting vulnerable pedestrians, of teslas verses cars that rely on radar.

annodomini2

6,862 posts

251 months

Sunday 20th November 2022
quotequote all
They're all flawed in certain ways, it's the combination that's important.

Focus on optical gives you no backup, humans predominantly use sight, but it's not our only sense.

If they want it to be safe you need redundant and multi sourced.

gangzoom

6,303 posts

215 months

Monday 21st November 2022
quotequote all
annodomini2 said:
They're all flawed in certain ways, it's the combination that's important.

Focus on optical gives you no backup, humans predominantly use sight, but it's not our only sense.
I didn't realise the DVLA allows you to drive if you are blind? What other sense do you have that allows you drive if you cannot see?

CheesecakeRunner

Original Poster:

3,807 posts

91 months

Tuesday 22nd November 2022
quotequote all
That’s not the point. Why not make it safer by providing the car more more abilities to ‘see’ than a human can?

Hans_Gruber

275 posts

171 months

Tuesday 22nd November 2022
quotequote all
Because having 2 systems, in the split second a decision has to be made, causes confusion. Computers, like humans, have reaction times.

Radar is incredibly basic and the field of view very limited. Assuming the camera and recognition software works, it is far advanced. Other manufacturers only rely on radar, that doesn’t mean it is better!

If we are referencing human senses - ever notice how the very best engineering solutions take their inspiration from nature? Submarines copy whales, air turbine blades copy whale fins, racing boats copy shark skin - there are many, many others. Im not David Attenborough but to my knowledge only one animal* (above water) uses radar to see. They use eyes! By switching to cameras with recognition software, as this evolves, will easily
out perform radar.

(Bats don’t even use the same basic radar cars use, they use echolocation, not radio waves that are subject to interference. They also have very limited vision for the same reason why radar has been disabled on your Tesla)



Edited by Hans_Gruber on Tuesday 22 November 07:30


Edited by Hans_Gruber on Tuesday 22 November 07:36

CheesecakeRunner

Original Poster:

3,807 posts

91 months

Tuesday 22nd November 2022
quotequote all
A well built system wouldn’t be deciding between two system inputs. All data would be coming in, aggregated, and the system making a decision based on all the data it has. This is how machine learning, that Tesla are a proponent of, works.

egomeister

6,701 posts

263 months

Tuesday 22nd November 2022
quotequote all
Hans_Gruber said:
(Bats don’t even use the same basic radar cars use, they use echolocation, not radio waves that are subject to interference. They also have very limited vision for the same reason why radar has been disabled on your Tesla)
On the plus side, it does allow them able to essentially fly blind.

And if we are talking senses, it's reductive to just refer to sight. Take the example of crossing the road - at minimum we use sight and hearing (although clearly sight is dominant). We even teach it to kids - Stop, Look, Listen!


Hans_Gruber

275 posts

171 months

Tuesday 22nd November 2022
quotequote all
That is a very good point but i’d have to quantify the negative effect on wind resistance by having 2 huge ears on the side of the car smile

(Only joking with you I know you’re making a serious point)

budgie smuggler

5,385 posts

159 months

Tuesday 22nd November 2022
quotequote all
Hans_Gruber said:
Because having 2 systems, in the split second a decision has to be made, causes confusion. Computers, like humans, have reaction times.

Radar is incredibly basic and the field of view very limited. Assuming the camera and recognition software works, it is far advanced. Other manufacturers only rely on radar, that doesn’t mean it is better!

If we are referencing human senses - ever notice how the very best engineering solutions take their inspiration from nature? Submarines copy whales, air turbine blades copy whale fins, racing boats copy shark skin - there are many, many others. Im not David Attenborough but to my knowledge only one animal* (above water) uses radar to see. They use eyes! By switching to cameras with recognition software, as this evolves, will easily
out perform radar.

(Bats don’t even use the same basic radar cars use, they use echolocation, not radio waves that are subject to interference. They also have very limited vision for the same reason why radar has been disabled on your Tesla)
I thought Tesla's (now removed) parking sensors were ultrasonic (i.e. echo-location) rather than RADAR?

I don't doubt optical processing can in theory be better than ultrasonic sensors but the thing is, ultrasonics are so simple to implement. All you need to do is time the receiving of a reflected signal and use that delay to change the tone: done.

Whereas with an optical system, you need binocular vision to process depth, the ability to see in pitch black/bright sunlight/dazzling reflections from shop windows/puddles, condensation on the lens, fog etc etc.

PS. IIRC shrews also use echolocation. smile