Radar deleted on all new Model Y and Model 3. Is FSD worth it?

[DayTrippin]

I don't know if I'd call it a snap decision per se. If I were a betting man, I'd say the future state roadmap adoption rate was accelerated by a logistical issue, most likely the availability of radar units. This probably saves a lot of money too, making each car more profitable as well.

 

[avs007]

I have 25+ years experience in the aerospace industry (engineer/scientist). Removing radar from the vehicles is insane. This eliminates little cost but massively reduces reliability.

Having redundant sources of range data, especially in all-weather, is FAR more reliable. And we've already seen our Teslas' driver assistance features behave sketchily, with camera blockages being reported in rain all the time.

If the software stops using the radar in my current radar-equipped Model Y, I'm selling it and getting a Volvo XC40 Recharge all-electric or a Mustang Mach-E.

Can't stress enough how bad a move this was on Tesla's part.

If you are going to sell your car, I'd probably avoid the Mach-E... Go to 3:20 in this video... Ford's Radar based collision mitigation is a POS. In IIHS's test, the Ford system couldn't detect the pedestrian, and ran it over... From the looks, it didn't even attempt to slow down.

 

[Yoliber]

Interesting, it seems like the Ford detected well ahead of time but didn't apply the brakes at all. Thankfully this doesn't apply to newer Fords like the Mach-E that got superior rating. Apparently Tesla reported to the NHSTA the radarless models don't even have FCW or AEB.

 

[DanDi58]

If you are going to sell your car, I'd probably avoid the Mach-E... Go to 3:20 in this video... Ford's Radar based collision mitigation is a POS. In IIHS's test, the Ford system couldn't detect the pedestrian, and ran it over... From the looks, it didn't even attempt to slow down.

Well those tests are a year and a half old. I imagine they've all improved since then. And there tests of a M3 in China that show it plowing through pedestrians as well. So no one is perfect yet.

 

[DanDi58]

Good video here covering the radar removal and future of TeslaVision.

 

[Yoliber]

Do we know for sure that radar is causing phantom braking? My lexus has radar cruise control and doesn't phantom brake. I heard AP1 doesn't phantom brake as much. Also green mentioned the Vision only FCW build kept giving false positives in sections his vehicle would typically phantom brake.

 

[Fourdoor]

I don't understand why people don't get that a fully camera-based system is what Elon has said they feel will work best - for years he has said this. So eliminating the radar isn't a cost saving or constraint issue - it's because it doesn't fit in with the strategy! He has talked about this for years. Why is it so hard to understand? Geez Louise....

More info: Elon Musk updates Tesla Full Self-Driving Beta v9 timeline and FSD Subscription once again - Electrek

I understand that this is Elon's philosophy. I just disagree with him.

My problem with this is the idea that "since humans drive with eye's, why can't a car do just as well?" The answer is "with enough proccessing power and data from unwitting beta testers it can"... but humans are shitty drivers in poor visibility conditions, and cars that have vision based autonomous features will also be shitty drivers in poor visibility conditions.

To put it simply, I want a car that is BETTER at driving than me in poor visibility conditions... not one that is just as good as me.

Keith

 

[Tam]

I don't understand why people don't get that a fully camera-based system is what Elon has said they feel will work best - for years he has said this. So eliminating the radar isn't a cost saving or constraint issue - it's because it doesn't fit in with the strategy! He has talked about this for years. Why is it so hard to understand? Geez Louise....

More info: Elon Musk updates Tesla Full Self-Driving Beta v9 timeline and FSD Subscription once again - Electrek

There's a less disruptive way to implement his experiment without losing the NTHSA ratings for standard safety features.

Other companies would first test out internally, submit the result to NTHSA to make sure the ratings are not downgraded, then announce to the whole world the success as agreed by the NTHSA, then delete the radar.

Skipping these steps suggest that Tesla's in a hurry after YEARS of planning to figure out the least disruptive way for implementation.

 

[avs007]

Do we know for sure that radar is causing phantom braking? My lexus has radar cruise control and doesn't phantom brake. I heard AP1 doesn't phantom brake as much. Also green mentioned the Vision only FCW build kept giving false positives in sections his vehicle would typically phantom brake.

I don't know about Tesla's case, but yes, on other cars, Radar absolutely can cause phantom braking. I have experienced it on two different manufacturers systems, in pretty much identical situations... The first being the street near my daughter's school. There is a curve on the street, with a lightpost near the apex of the curve, that always causes a phantom breaking event if I'm in the right most lane approaching the lightpost...

The other scenario, is if the person if front of me is making a right turn, and they complete the turn, but are slow in accelerating.... The lane ahead of me is clear between all pavement markings, but the rear of the car is still within view... I'll get a "brake check" like event until the other car accelerates.

Does your Lexus have Forward Collision Mitigation, or just Adaptive/Radar Cruise Control? Because in the two scenarios I mentioned above, the brake check that occurs with the car turning right, only happens if I have Driver's Assist enabled... But the curve on the street near my daughters school, occurs regardless if I have drivers assist enabled or not, as in that particular case, I've noticed the seat belt pre-tensioners activate too, so my car thinks I'm about to crash.

 

[Yoliber]

I don't know about Tesla's case, but yes, on other cars, Radar absolutely can cause phantom braking. I have experienced it on two different manufacturers systems, in pretty much identical situations... The first being the street near my daughter's school. There is a curve on the street, with a lightpost near the apex of the curve, that always causes a phantom breaking event if I'm in the right most lane approaching the lightpost...

The other scenario, is if the person if front of me is making a right turn, and they complete the turn, but are slow in accelerating.... The lane ahead of me is clear between all pavement markings, but the rear of the car is still within view... I'll get a "brake check" like event until the other car accelerates.

Does your Lexus have Forward Collision Mitigation, or just Adaptive/Radar Cruise Control? Because in the two scenarios I mentioned above, the brake check that occurs with the car turning right, only happens if I have Driver's Assist enabled... But the curve on the street near my daughters school, occurs regardless if I have drivers assist enabled or not, as in that particular case, I've noticed the seat belt pre-tensioners activate too, so my car thinks I'm about to crash.

It has both. Only false positives I get in this system is from RCTA but the reasons for it is understandable.

 

[avs007]

I understand that this is Elon's philosophy. I just disagree with him.

My problem with this is the idea that "since humans drive with eye's, why can't a car do just as well?" The answer is "with enough proccessing power and data from unwitting beta testers it can"... but humans are shitty drivers in poor visibility conditions, and cars that have vision based autonomous features will also be shitty drivers in poor visibility conditions.

To put it simply, I want a car that is BETTER at driving than me in poor visibility conditions... not one that is just as good as me.

Keith

The reason why cars are shitty drivers in poor visibility is different then why people are shitty drivers in poor visibility. Human response times are slow, and their thought process irrational at times, or completely lacking becuase of distraction, etc. Not all the cameras will be affected by weather, as the main forward sensor module is behind the windshield where the wipers can sweep the area. Dynamic range should not be an issue especially if frame rates are high, as even my doorbell camera has very good dynamic range, even if the sun is in the field of view.. But with that being said, radar is not impervious to blockage either... Whenever there is light snow, the forward radar in all my cars can become blocked/disabled relatively quickly, according to the info center in the dash.

 

[Yoliber]

Personal opinions...

1. FSD was never worth it
2. Radarless FSD is worth more than Radar FSD because if they return to radar then you'll get it covered

 

[mark95476]

Kyle from Out of Spec Motoring found that the Mustang's Co-pilot 360 will run into the back of other cars on the freeway. This is from a month ago so it's recent.

See 21:45

Does the Mustang have radar? If it does, then it's completely worthless ("unbelievably dangerous" per Kyle).

If you are going to sell your car, I'd probably avoid the Mach-E... Go to 3:20 in this video... Ford's Radar based collision mitigation is a POS. In IIHS's test, the Ford system couldn't detect the pedestrian, and ran it over... From the looks, it didn't even attempt to slow down.

Well those tests are a year and a half old. I imagine they've all improved since then. And there tests of a M3 in China that show it plowing through pedestrians as well. So no one is perfect yet.

 

[tkizzy]

It feels like these forums are full of conspicuously anti-Tesla trolls these days. A simple thread went all over the place with hyperbolic fudding around… 4 pages of useless vitriol (With exceptions).

my take - fog seems like an obvious use case for radar, so that’s the initial shortcoming that comes to mind when leaving out radar. BUT I understand some digital imaging can see* through fog etc better than the human eye. Also there are visible, infrared cameras as well as thermal imaging.. Imagine a neural net processing 360 of various spectrum light data overlayed in some novel capacity (adjusted perfectly), or simultaneous processing of the layers… seeing as how the end result is still a video you get rid of potential sensor fusion issues of radar/vision/lidar. IN conclusion, let’s wait to learn more before jumping in making conclusions but I can see a logical path forward here.

 

[andromeda70]

Agreed. Saw a lot of posts and videos where Tesla stoped in dense fog or rain before driver could see or react.
So I wonder how purely relying on cameras will work.

On the other hand, a driver that has FSD turned on in dense fog, should have his/her drivers license revoked. You can’t blame new technology for the lack of personal responsibility.

 

[LoganNolag]

It feels like these forums are full of conspicuously anti-Tesla trolls these days. A simple thread went all over the place with hyperbolic fudding around… 4 pages of useless vitriol (With exceptions).

my take - fog seems like an obvious use case for radar, so that’s the initial shortcoming that comes to mind when leaving out radar. BUT I understand some digital imaging can see* through fog etc better than the human eye. Also there are visible, infrared cameras as well as thermal imaging.. Imagine a neural net processing 360 of various spectrum light data overlayed in some novel capacity (adjusted perfectly), or simultaneous processing of the layers… seeing as how the end result is still a video you get rid of potential sensor fusion issues of radar/vision/lidar. IN conclusion, let’s wait to learn more before jumping in making conclusions but I can see a logical path forward here.

Yeah that’s true for sure. Depending on the sensitivity of the sensors and aperture of the lens a camera can see WAY better than a human eye. My A7RIII with a 50mm f 1.2 or even a 1.8 can see much better in the dark than I can.

 

[DayTrippin]

It feels like these forums are full of conspicuously anti-Tesla trolls these days. A simple thread went all over the place with hyperbolic fudding around… 4 pages of useless vitriol (With exceptions).

my take - fog seems like an obvious use case for radar, so that’s the initial shortcoming that comes to mind when leaving out radar. BUT I understand some digital imaging can see* through fog etc better than the human eye. Also there are visible, infrared cameras as well as thermal imaging.. Imagine a neural net processing 360 of various spectrum light data overlayed in some novel capacity (adjusted perfectly), or simultaneous processing of the layers… seeing as how the end result is still a video you get rid of potential sensor fusion issues of radar/vision/lidar. IN conclusion, let’s wait to learn more before jumping in making conclusions but I can see a logical path forward here.

I wouldn't mind if they dumped the radar so much, or at all, IF they had upgraded the other equipment to make sure we are at parity. What it seems like more and more is it was just a cost-cutting or a logistics (couldn't get the radar units in time) measure to meet the Q2 sales goals.

This doesn't look like a smooth rollout of anything but a knew jerk reaction. The best thing about this entire mess is they had to disclose it to NHTSA or likely Tesla would have tried to bury it under a webpage somewhere and wouldn't get the attention it deserves.

If they make it work as well as what they had, or better, then I won't have any issue. Dropping the automatic emergency braking is a big problem for me. So hopefully by the time my car arrives, they'll have it worked out. Meanwhile the people who got the currently crippled cars, I feel for you. Hopefully Tesla will make it right on something other than Elon Time.

If the system was good enough for prime time now, they wouldn't have informed the NHTSA. I don't mind beta testing something like FSD if I thought it was worth it but basic safety features shouldn't need beta testing.

 

[twc]

Losing the NHTSA rating might increase the insurance rate for these cars. I guess Tesla didn't think about this when they decided to do this. I will have to follow when I get my new rate.

 

[tkizzy]

NHTSA requires new testing for any non-trivial changes that could impact the safety systems. Obviously a radar-deleted version is novel so doing a new rating is business as usual. Just like ap1 v ap2+ you’ll get your feature parity…..and, judging by the progress we’re seeing on FSD, “and then some”

 

[DayTrippin]

Losing the NHTSA rating might increase the insurance rate for these cars. I guess Tesla didn't think about this when they decided to do this. I will have to follow when I get my new rate.

That was my thought too. I mentioned this in another thread. They don't need to make the Tesla tax any higher...