Bli medlem i Norsk elbilforening og støtt driften av Elbilforum. Som medlem får du i tillegg startpakke, medlemsfordeler og gode tips om elbil og lading. Du blir med i et fellesskap som jobber for mindre utslipp fra veitrafikken. Medlemskap koster 565 kroner per år. elbil.no/medlemskap

Dundret rett inn i veisperringene: – Skremmende

Startet av Motorwatt, lørdag 13. november 2021, klokken 09:45

« forrige - neste »

Ellemelle

Ja hvis du kun mente at dette er status quo, så er vi enige.
Jeg tolket det slik at du mente at kamera aldri ville kunne detektere dette, og at radar eller lidar derfor er nødvendig.

Det er nok mye enklere å få EU-godkjent et collision avoidance system basert på kamerateknologi enn å få EU-godkjent FSD, så du skal ikke se bort i fra at de finner en løsning på dette ganske snart.

Lynet

Har flere ganger opplevd å nå igjen stillestående kø, både med AP1 på modell S og full FSD hardware på modell X og bilen bremser pent og pyntelig bak bilen som står bakerst. Så mine biler har uten tvil detektert stillestående biler i min fil.
Trondheim:
Tesla X100D 2019 modell.

Motorwatt

Tesla S75D - 09/2018
__________________________________________________________________________________________________________

Jeronan

#63
Sitat fra: Lynet på tirsdag 16. november 2021, klokken 18:39
Har flere ganger opplevd å nå igjen stillestående kø, både med AP1 på modell S og full FSD hardware på modell X og bilen bremser pent og pyntelig bak bilen som står bakerst. Så mine biler har uten tvil detektert stillestående biler i min fil.

Then the car in front of you was detected and locked onto when it was still rolling. So from there it will follow stop and go traffic, otherwise ACC would never work.

I have experienced plenty examples where I approached a totally stopped traffic jam and AP did not slow down.
So please do not blindly trust in AP to always slow down in time when approaching a traffic jam / stopped car.
Model Y - Sept 2021
Model S 75D - Sold
BMW i3 2017 - Feb 2017

Lynet

Trondheim:
Tesla X100D 2019 modell.

daktari

Sitat fra: jlan på tirsdag 16. november 2021, klokken 14:33
Sitat fra: OlaM på tirsdag 16. november 2021, klokken 12:20
Det er ikke helt riktig at VO kun ser bilen rett foran, men doppler radar er mye bedre på akkurat det. Pga frekvensendringen til doppler-effekten kan den si noe om den tingen som reflekterte signalet holder lik fart, drar unna eller nærmer seg.
Nettopp. Og når objektet står stille i forhold til omgivelsene, så er det vanskelig å skille ut for en dopplerradar.

Tror poenget var at VO kun kan identifisere det som er i siktlinjen til kameraet. En bil foran vil dekke til bilen foran der igjen. Med radar skulle de jo sende signalene under bilen.
Det første kan man se på FSD-BETA videoer fra f.eks lyskryss.
VO måler både fart og retning kun med kameraene, det kan være endel å gå på presisjonen foreløpig.
Mine videoer av Autopilot 2 Mine I-Pace videoer
2023 ID.4 GTX 2021 ID.4   2022 BMW iX40   2021 Model 3 LR   2019 Audi e-tron 55   2019 I-Pace HSE   2017 S90D

jlan

Sitat fra: snakkesalig på onsdag 17. november 2021, klokken 12:40
Sitat fra:
https://slate.com/technology/2021/08/teslas-allegedly-hitting-emergency-vehicles-why-it-could-be-happening.html
... It's basically a super-duper, very sophisticated pattern-matching scheme. The problem is that, in the real world, it is given an image where it sees an obstacle that it has never seen before. The patterns really do not match, so it will not detect it as a vehicle. For example, when the first person was killed using the Tesla autopilot in Florida, the truck [hit by the Tesla] was perpendicular to the direction the motion. The training did not have those images at all, so therefore the pattern matcher did not recognize that pattern. There've been many recent incidents where the Tesla vehicles run into a firetruck or police vehicle. The lights are on, so the red, green, blue pixel values look different as well, and therefore the patterns do not match. Lo and behold they declare that there's no obstacle ahead, and the vehicle very promptly dysfunctions and has no idea there's something in front of it. ...
Nettopp mitt poeng, jo mer blinkende lys og ukjente objekter jo vanskeligere blir det for kamerasystemet. :
Sitat fra: jlan på tirsdag 16. november 2021, klokken 14:33
For å unngå kollisjon med en stillestående barriere er det derfor viktig at objektet kan gjenkjennes. Det var derfor jeg foreslo bevegelige dukker i veibanen.
2016-2019: TMX 90D AP1
2019-2021: TM3 LR AWD HF
2021-2021: TMY LR
2021-        : TM3 LR 82kWh HF

Motorwatt

Han sier også at lidar til en stor grad ville hjulpet i slike situasjoner. Men EM har "lovet" oss at det aldri kommer, men kanskje hvis myndighetene forlanger det?
Tesla S75D - 09/2018
__________________________________________________________________________________________________________

Jeronan

#68
Sitat fra: Motorwatt på onsdag 17. november 2021, klokken 13:54
Han sier også at lidar til en stor grad ville hjulpet i slike situasjoner. Men EM har "lovet" oss at det aldri kommer, men kanskje hvis myndighetene forlanger det?

Elon Musk holy believes that he can solve everything with cameras alone, because he compares cameras as our human eyes. While many experts have already said that it doesn't work that way, as no matter how sophisticated our brain is and its ability to map a 3D view of our world through our eyes, many research has shown that even our brain can easily be fooled.

Tesla's AP cameras get easily blinded/obstructed by low sun light and stray light from vehicles and street lights in the dark.
Tesla's AP cameras also have rather poor vision in the dark, complicating object recognition. Hence why a lot of these accidents happen in low light conditions.

The further an object is away, the more difficult it is to determine its distance by Camera vision alone. This problem can be multiplied when driving at high speed.
Add low light / dark conditions and you have a recipe for disaster.
There is simply no camera on the market that has a high enough resolution and has enough sensitivity to see far enough in the dark. It would be too expensive and too big to fit in the windshield. You can then just as well use Lidar and be done with it.

Then we also haven't even talked about weather conditions, like fog, heavy rainfall, snowfall, etc.
Model Y - Sept 2021
Model S 75D - Sold
BMW i3 2017 - Feb 2017

Motorwatt

Cheaper and Thinner LIDAR Sensors Brighten the Future of Autonomous Transport

https://www.lightcounting.com/research-note/july-2021-cheaper-and-thinner-lidar-sensors-brighten-the-future-of-autonomous-transport-164

Sales of LIDAR modules have continued to decline in 2021 since several leading companies canceled or postponed their LIDAR projects last year. Apart from lower volumes, average selling prices for LIDARs dropped sharply in 2020. The target price for automotive lidar units to-day is reportedly in the $500 - $600 range but will likely continue to drop to sub-$500 in 2021. The stakes are high at this second phase. The companies that have technology to manufacture the lower cost LIDARs will emerge as winners, adding billions of dollars to their market valuations.
Tesla S75D - 09/2018
__________________________________________________________________________________________________________

kodax

Sitat fra: Jeronan på onsdag 17. november 2021, klokken 14:39
Sitat fra: Motorwatt på onsdag 17. november 2021, klokken 13:54
Han sier også at lidar til en stor grad ville hjulpet i slike situasjoner. Men EM har "lovet" oss at det aldri kommer, men kanskje hvis myndighetene forlanger det?

Elon Musk holy believes that he can solve everything with cameras alone, because he compares cameras as our human eyes. While many experts have already said that it doesn't work that way, as no matter how sophisticated our brain is and its ability to map a 3D view of our world through our eyes, many research has shown that even our brain can easily be fooled.

Tesla's AP cameras get easily blinded/obstructed by low sun light and stray light from vehicles and street lights in the dark.
Tesla's AP cameras also have rather poor vision in the dark, complicating object recognition. Hence why a lot of these accidents happen in low light conditions.

The further an object is away, the more difficult it is to determine its distance by Camera vision alone. This problem can be multiplied when driving at high speed.
Add low light / dark conditions and you have a recipe for disaster.
There is simply no camera on the market that has a high enough resolution and has enough sensitivity to see far enough in the dark. It would be too expensive and too big to fit in the windshield. You can then just as well use Lidar and be done with it.

Then we also haven't even talked about weather conditions, like fog, heavy rainfall, snowfall, etc.
So do you believe that VO a "scam", that Elon does not understand this and do not listen to his engineers, or that the engineers themselves do not understand this?
The arguments you present does absolutely sounds legit, but here are many "experts" on the internet, listening to other "experts", fueld by competitiors wanting to make people believe their own solution is better, by manufacturers of products not being used, misintepreting information comming from real experts etc.
I am pretty certain the engineers at Tesla knows this way better than we do, that Elon absolutely listen to them, and that VO is not a scam. Wether or not it will be good enough for FSD remains to bee seen, though:)
Tesla Model X 75D 7-seter FSD - Januar 2019
Model Y - Model 3

Motorwatt

Det er vel også EMs skyld at det finnes en stor grad av skepsis?
Tesla S75D - 09/2018
__________________________________________________________________________________________________________

filtvet

Jeg jobber med bla. analyse og AI er en stor scam. Det finnes ikke ekte AI er min mening men kun tall og sannsynligheter. Helt ubrukelig ift den menneskelige hjerne.


Jeronan

#73
Sitat fra: kodax på torsdag 18. november 2021, klokken 10:25
So do you believe that VO a "scam", that Elon does not understand this and do not listen to his engineers, or that the engineers themselves do not understand this?
The arguments you present does absolutely sounds legit, but here are many "experts" on the internet, listening to other "experts", fueld by competitiors wanting to make people believe their own solution is better, by manufacturers of products not being used, misintepreting information comming from real experts etc.
I am pretty certain the engineers at Tesla knows this way better than we do, that Elon absolutely listen to them, and that VO is not a scam. Wether or not it will be good enough for FSD remains to bee seen, though:)

I wouldn't put it in such harsh words, but yeah pretty close it is. Just like FSD and AI are a basically a scam at this point in time.

We are not talking about so called "armchair" or "internet" experts here, but actual real experts and researchers in the field that have explained in great detail the extreme challenges and risks with VO (Visual Only) with today's available technology.

What aggravates me even more, is how all these tech companies are throwing around the term AI.

When we use the term AI, we talk about Artificial Inteligence (self awareness). Hoarding lots of data and running some algoritms against it on fancy server farms is not AI !

They have cheapened and degenerated AI into a cheap marketing ploy to fool the public and investors.
Model Y - Sept 2021
Model S 75D - Sold
BMW i3 2017 - Feb 2017

daktari

Det som hvertfall er fakta er at:
Lidar er blitt mye billigere og mye mindre siden høsten 2016
Radarer har også fått bedre oppløsning og rekkevidde siden 2016 (men Tesla bruker en standard radar fra ca 2013/14)
Mine videoer av Autopilot 2 Mine I-Pace videoer
2023 ID.4 GTX 2021 ID.4   2022 BMW iX40   2021 Model 3 LR   2019 Audi e-tron 55   2019 I-Pace HSE   2017 S90D

© 2024, Norsk elbilforening   |   Personvern, vilkår og informasjonskapsler (cookies)   |   Organisasjonsnummer: 982 352 428 MVA