Autopilot Test Fail?

[McRat]

Whatever dude --- Doug has given the Model 3 a HUGE pass and high ranking --- he's too afraid of you Tesla fanbois ---- seriously I'm a huge fan of Doug and he let SO MANY THINGS GO with his Model 3 review --- seriously compare the Model 3 review to his Honda Civic Type-R or ANY other review he does

That's the second review he did where he was clueless that I started watching. I forget the other one. I think it was a performance car he didn't know how to operate? Dunno. It could have been a 10 speed bicycle even. He'd probably say it had 7 gears, two in the front and 5 in the back.

If you like your car reviews to be wrong about the features, then he's your go-to guy. He can't even read a brochure much less an owner's manual.

 

[MXWing]

....and how many of those miles were you not holding the steering wheel and paying full attention to your surroundings, ready to take over at a moment's notice?

Zero but you won’t like the details.

Only needed one relaxed hand -resting- on the wheel.
Both feet flat -resting- on the ground.

I pay full attention when driving in a Tesla, when not driving a Tesla (rare on a road trip).

I also pay full attention as a passenger. For the benefit of the driver and my own situational awareness radar that never turns off anyway.

Nice thing about Tesla Autopilot is I am free to spend even more vision and mental cycles to process to look further down the road, further behind me and anticipate the decisions and emotions of drivers around me. I can do this because I don’t have to exert cycles on controlling speed, distance or lane position.

Am I doing this whole auto pilot thingy correctly?

 

[Chickenlittle]

https://jalopnik.com/this-test-shows-why-tesla-autopilot-crashes-keep-happen-1826810902

Interesting scenario. In this kind of situation a driver might be able to swerve where auto pilot just brakes, but if there are other cars in the next lane I think a crash is inevitable. Essentially the Car in front is driving irresponsibly or not paying attention and end up causing the accident?

I'm 50/50 on this one. At least the autopilot went into brake mode immediately, where distracted driver might hit with more speed. But, a real driver might also glance in the mirror and be able to avoid the accident.

thoughts?

YOU are the driver. Glance in mirror and take over. Isn’t this the number 1 teaching point of years of discussions and manual

 

[voip-ninja]

That's the second review he did where he was clueless that I started watching. I forget the other one. I think it was a performance car he didn't know how to operate? Dunno. It could have been a 10 speed bicycle even. He'd probably say it had 7 gears, two in the front and 5 in the back.

If you like your car reviews to be wrong about the features, then he's your go-to guy. He can't even read a brochure much less an owner's manual.

He’s an idiot but he’s a sincere and enthusiastic idiot which is why he has 1.2 million YouTube followers and has been able to turn goofy car reviews into a career and quit his day job.

When he did his Model 3 review of a very early production car it was obvious he was going to give it a wide pass on stuff that any real car reviewer would have bagged on.

 

[MXWing]

Do you want to see a neutral evaluation of AutoPilot on a Model 3?

The interesting point is at 11:15. Jesus.

JESUS? Nice selective hit pieceing on Tesla.

There's a follow up video that shows AP kicking ass on the same stretch of road the AP has had 3-4 MAJOR improvements AFTER the already improved video I am linking below.

Man, I should publish the video of no cruise control on my Model X in Dec 2016 with today's date and pretend its relevant.

 

[Electroman]

JESUS? Nice selective hit pieceing on Tesla.

There's a follow up video that shows AP kicking ass on the same stretch of road the AP has had 3-4 MAJOR improvements AFTER the already improved video I am linking below.

Man, I should publish the video of no cruise control on my Model X in Dec 2016 with today's date and pretend its relevant.

This is the one I was looking for. The guy says there is a link at the end of the first video, but there was no link.

 

[PhaseWhite]

Not sure they mention it but it would help to know the exact speed it was traveling and the firmware they were using. As we know OTA Updates have been happing which could have affected the results. Especially after 2018.10.4. Not evaluation is helpful without knowing the fw version.

Agreed, none of the testers, including Edmunds can be bothered to list the FW version and their results become dated weeks later. They are stuck in legacy mode where the car delivered is frozen in time. As other's have pointed out this test doesn't give any other important details like the following distance setting. It appears to be set to minimum (1) or (2) since at speed Tesla's following distance has gotten even more conservative lately. All they've done is proven they can get it to crash. How many cuts did they take where it didn't crash? Since the guy leads into the cut with "this could happen" obviously they did that exact cut before and know it will crash, but again, I bet if we saw all the takes they had to do it would paint a very different picture.

 

[PhaseWhite]

No fire engines either!! Unfair to body shops and hospitals!!!

What makes you think super-cruise won't hit a fire engine protruding into it's lane on the side of an interstate? It uses camera and radar just like Tesla's system. The only Lidar help is from other vehicles that create HD maps it uses. It only takes a few seconds at highway speeds for things to change.

 

[McRat]

What makes you think super-cruise won't hit a fire engine protruding into it's lane on the side of an interstate? It uses camera and radar just like Tesla's system. The only Lidar help is from other vehicles that create HD maps it uses. It only takes a few seconds at highway speeds for things to change.

I guess we will see. There are tens of thousands of cars out there, perhaps 100,000, with the GM's AEB, FCW, FAB systems. There haven't been a lot of posts by enthusiasts of malfunctions and no reported deaths due to system failure. Not to mention the GM/SC system uses head position.

One of GM's claim to fame is some of the world's best automotive embedded systems engineering.

 

[d21mike]

Most if not all radar based adaptive cruise control systems can't see stopped objects. So the car isn't going to see a stopped car in front of a moving car.

Since firmware 2018.10.4 my Tesla can definitely stop for a stopped car. And I think it is even getting better with each release I install. And, no it had not seen the stopped car moving before stopping for it. Example: While exiting the freeway and going around a bend it sees a stopped car at the light and stops.

So how does it do it now? This is something it could not reliably do before 2018.10.4. Does it use cameras now where it was using radar in the past?

However, I can only confirm this working when traveling < 50 mph.

 

[SomeJoe7777]

Since firmware 2018.10.4 my Tesla can definitely stop for a stopped car.

Sometimes. Even most of the time. But not ALL the time. That 1% of the time when it fails to stop for the stopped vehicle is what lulls people into a false sense of security. Even if you see it stop for a stopped car yourself 50 times, there may be an instance where it won't.

My advice: Do NOT rely on it to stop for a stopped car. It's great that it's getting better, but it's not ready to be 100% trusted yet.

 

[d21mike]

Sometimes. Even most of the time. But not ALL the time. That 1% of the time when it fails to stop for the stopped vehicle is what lulls people into a false sense of security. Even if you see it stop for a stopped car yourself 50 times, there may be an instance where it won't.

My advice: Do NOT rely on it to stop for a stopped car. It's great that it's getting better, but it's not ready to be 100% trusted yet.

I keep at hand on the wheel at all times and ready to brake in these and other situations. I was just responding to the point that some cars (I assumed including Tesla) can not stop for a stopped car. I was just correcting that. Using your example. It might be possible that it has failed once (just not sure of the details) and that is why i do not trust it 100% of the time. But prior to 2018.10.4 it would fail almost every time. When I say almost every time I mean that it may not fail if going pretty slow like < 25-30 mph.

 

[MXWing]

Agreed, none of the testers, including Edmunds can be bothered to list the FW version and their results become dated weeks later. They are stuck in legacy mode where the car delivered is frozen in time. As other's have pointed out this test doesn't give any other important details like the following distance setting. It appears to be set to minimum (1) or (2) since at speed Tesla's following distance has gotten even more conservative lately. All they've done is proven they can get it to crash. How many cuts did they take where it didn't crash? Since the guy leads into the cut with "this could happen" obviously they did that exact cut before and know it will crash, but again, I bet if we saw all the takes they had to do it would paint a very different picture.

The follow distance was unnaturally bad. Don't really see that in actual practice. Unless somehow you were *intentionally* trying to create a desired outcome. hmmmm

I guess we will see. There are tens of thousands of cars out there, perhaps 100,000, with the GM's AEB, FCW, FAB systems. There haven't been a lot of posts by enthusiasts of malfunctions and no reported deaths due to system failure. Not to mention the GM/SC system uses head position.

One of GM's claim to fame is some of the world's best automotive embedded systems engineering.

No one cares about GM vehicle autonomy so you don't hear about it. Just because you didn't hear a tree fall in a forest, doesn't mean it never happens.

Since firmware 2018.10.4 my Tesla can definitely stop for a stopped car. And I think it is even getting better with each release I install. And, no it had not seen the stopped car moving before stopping for it. Example: While exiting the freeway and going around a bend it sees a stopped car at the light and stops.

So how does it do it now? This is something it could not reliably do before 2018.10.4. Does it use cameras now where it was using radar in the past?

However, I can only confirm this working when traveling < 50 mph.

Yes I notice it gets "better" with every iteration. I think the Radar is getting better at picking up 'minute' movements and detecting objects at further ranges so it knows when to stop.

From my understanding, the final evolution of the system is the cameras will know the object in front of you is a stopped car because of license plates, tail lights and other indications that would simulate human vision.

Even if it stops 99,999/100,000 times for a stopped car, a driver should sense when the 1/100,000 time it didnt. Based on hey, I am going at X speed with Y stopping distance - I better brake.

It is always the drivers fault and you MUST PROTECT YOURSELF AT ALL TIMES. As much as I hate insurance companies, they will back me up on this one.

 

[McRat]

No one cares about GM vehicle autonomy so you don't hear about it. Just because you didn't hear a tree fall in a forest, doesn't mean it never happens.
.

Just because you didn't hear a nuclear weapon detonated in your neighborhood doesn't mean it did just happen. Go look!

Tesla is small fry when it comes to lawsuits. If a lawyer heard there was a sharp corner on a GM hood release that gave people scratches, it would be "CALL SAUL!!!", Infomercials, a feature length movie, the plot of TV shows, campaign ads, and a Senate inquest. There is precedent to back this up.

Tesla feels beset upon by the media. They could not survive a single GM legal accusation, especially the false ones.

 

[Knightshade]

I guess we will see. There are tens of thousands of cars out there, perhaps 100,000, with the GM's AEB, FCW, FAB systems. There haven't been a lot of posts by enthusiasts of malfunctions and no reported deaths due to system failure. Not to mention the GM/SC system uses head position.

The post you are replying to was about supercruise.

Which isn't on tens of thousands, or 100,000, cars.

It's on the Caddy CT6. Only. And only on any CT6s made since September 6, 2017.

So a few thousand, not tens of thousands.

They're only a tiny fraction of the # of Tesla AP cars on the road, and have driven an even tinier fraction the # of miles (and given how few miles of total road in the US it actually works on, it's likely to remain a tiny fraction of # of miles driven in Teslas for years at least...might change once GM gets the system into more cars but that's 2020 earliest)

 

[McRat]

The post you are replying to was about supercruise.

Which isn't on tens of thousands, or 100,000, cars.

It's on the Caddy CT6. Only. And only on any CT6s made since September 6, 2017.

So a few thousand, not tens of thousands.

They're only a tiny fraction of the # of Tesla AP cars on the road, and have driven an even tinier fraction the # of miles (and given how few miles of total road in the US it actually works on, it's likely to remain a tiny fraction of # of miles driven in Teslas for years at least...might change once GM gets the system into more cars but that's 2020 earliest)

Super Cruise technically is only the steering system.

The braking systems and alert systems are found in many GM products.

It's not the steering wheel that runs into the back of parked cars.

It's an automatic warning/braking system that neither warns nor brakes in time to have a significant effect.

 

[MXWing]

Just because you didn't hear a nuclear weapon detonated in your neighborhood doesn't mean it did just happen. Go look!

Tesla is small fry when it comes to lawsuits. If a lawyer heard there was a sharp corner on a GM hood release that gave people scratches, it would be "CALL SAUL!!!", Infomercials, a feature length movie, the plot of TV shows, campaign ads, and a Senate inquest. There is precedent to back this up.

Tesla feels beset upon by the media. They could not survive a single GM legal accusation, especially the false ones.

GM faces lawsuit after crash between motorcyclist and self-driving Chevy Bolt

Slightly more than a sharp corner that gave someone a scratch.

I barely heard about this and I spend an overwhelming amount of time being brushed up on EV happenings than the general public.

'Everyone' knows about Tesla death machines.

I rest my case.

 

[xav-]

Do you want to see a neutral evaluation of AutoPilot on a Model 3?

The interesting point is at 11:15. Jesus.

Lol.. as if a US auto press journalist can be “neutral” when it comes to GM..

From my recollection this is a pre March 2018 autopilot review. There has been major changes to the autopilot since then. Ask anybody who has driven AP20 now and then. It is day and night.

 

[McRat]

GM faces lawsuit after crash between motorcyclist and self-driving Chevy Bolt

Slightly more than a sharp corner that gave someone a scratch.

I barely heard about this and I spend an overwhelming amount of time being brushed up on EV happenings than the general public.

'Everyone' knows about Tesla death machines.

I rest my case.

Perfect case study and was reported in most major outlets.

MC was riding at 17 mph weaving through traffic.
Bolt was going 12 mph, when it could not complete a lane change safely.
MC runs into the side of the Bolt.
SFPD and digital recordings both point the blame at the scooter rider.
MC rider is now on disability with a lawyer aiming for big bucks.

Searching "Cruise Automation Bolt hits motorcycle SFPD" shows 270,000 hits on Google and dozens if not hundreds of media outlets.

One even has a title that indicates GM is to blame, then continues with an essay about how AVs must behave better, and how GM had nearly 2 dozen incidents and Waymo and Google had 3 in 2017.

Waymo/Google is the same team. They moved most all their testing operations out of California to Arizona that year. 2 out of the 3 incidents were Waymo errors.

GM ran in SF all year, no at fault accidents, which were 22 incidents.

The article was obviously aimed at colorizing the topic. Not lying, just selective truth telling. Often these are press releases by the legal team.

GM is deep pockets. The crash will cost GM, but mostly in publicity. It won't help AV tech much either.

 

[McRat]

Lol.. as if a US auto press journalist can be “neutral” when it comes to GM..

From my recollection this is a pre March 2018 autopilot review. There has been major changes to the autopilot since then. Ask anybody who has driven AP20 now and then. It is day and night.

Most the US press has always hated GM. Remember they sold the Deadly Corvair, Exploding Pickups, Faulty Killer Keys, and Destroyed the Entire EV Industry to Stop Electric Cars. GM remains the symbol in the US for how evil a corporation can be.

But Edmunds is pretty neutral. Read their stuff. Like most, they tend to favor imports.

Like I said. The cars Edmunds tested were As Sold cars. They did not try to find a bad one, and they even noted there has been an update since the first test, which they do include a link to. The fact remains that Tesla sold that AP system as being safe. Obviously 10 seconds looking down would have been lethal when that truck was coming.