Featured Tesla recalled over "full self-driving"

Why are the pro Tesla people so sure (and vocal about it) that people must be Luddites, uneducated or misinformed if they criticize the methodology which is used by Telsa to create products?

Why are so many of those same Tesla customers trying so hard to discredit Toyota products every single day? They generate so much FUD that makes it very difficult to find valid information on Toyota developments or maintain a discussion about new Toyota products. It's so sad to see the forums degrade this way.

 

You mean why do people that have the car and the software try to correct poor assumptions. Obviously you want to stay misinformed, but I am trying to correct it for others on the thread and perhaps even you if you check out the system. You can ask what the system does well or poorly, but many of your posts pretend a driver should not be responsible or that teslas are hitting things way more things and causing more accidents than reported, and maybe the NHTSA is covering it up. Read their reports at least.

NHTSA noted that incidents of teslas hitting vehicles parked in a lane in a highway while at night and going over 40 mph. They also noted that a driver should be able to correct for this. Tesla is reporting more data than any other car company to the safety agencies. These incidents happen around 6 times a day and the grand majority are by people driving other cars than teslas.

I did not like Toyota trying to slow plug-in adoption. They appear to have changed based on new announcements. My partner would have bought a rav4 prime if the dealer wasn't asking such a high mark up. I posted about a known bug that they fixed in my prius. That was not fud, and I said bugs happen and Toyota corrected it.

 

With friends like this, who needs enemies?



Does it not have the situational awareness to realize that it's stopped at a railroad crossing? Or that trains use railroads, not trucks? Bizarre.

 

You are literally making the same mistake as all the idiots doing stupid hands free stunts in a Tesla; conflating the systems marketing names with capabilities they don't have.

Here is a quick run down of what Tesla's active ADAS are. They all have the passive(blind spot warning, autobraking, collision avoidance, etc.)

• Autopilot; it's dynamic cruise control and lane tracing. That's it.
• Enhanced Autopilot; this is the capabilities the new Prius is getting with TSS-P 3.0. Plus, it follow routes set in the navigation system, and slows down when lane changing into an off-ramp.
• FSD can stop at a light or sign, which lets the above be used on surface streets instead of just highways and country roads.

Now, are you going to count every time you adjusted the following distance of the DCC, or decided to turn it off, as accidents, cause they are interventions in regards to a Level 2 system like Toyota and Tesla have.

Any of these systems, from any manufacturer, could be reducing crashes.

For those honestly interested,
Autopilot and Full Self-Driving Capability | Tesla Support
Autopilot

*checks thread title*
What Toyota product should be discussing here? Or is it just Toyota recalls?

What do you think a Toyota would see? The system's job is to; identify an object, determine if the object is on a collision course with the car, and then keep any collision from happening. According to my Outback, every car, SUV, or truck in front of it is some nondescript sedan. As long as it doesn't plow into what ever it is, all is good.

Does any other manufacturer display what the ADAS sees as detailed as a Tesla?

 

No the system can not tell the difference between a railroad, a bike path, or a new road. Only established roads are in the system. When I got the system in 2018 it really couldn't tell much. Hopefully when the new computer rolls out it will be able to identify such things. About once a month I see a car driving in the bike lane by my house, not yet a tesla, but it could happen. Fortunately I haven't seen anyone get hit. There is are now stand up barriers for the bike lane that is to the side of a two way street with lights and stop signs. The bidirectional bike lane is just wide enough for cars to go into, instead of the north bound lane. I don't think I have ever let autopilot stop for a train, I normally do that myself so I can slow down earlier and use the regen brakes more.

 

policy to live by; always try & deliver one's words, "soft & sweet" so that when one has to EAT them, they go down much much easier.

from google searching for, "definition of a truck .... one of the 1st two definitions of 'truck' . . . . .
*(enjoy)
" ... an undercarriage with four to six wheels pivoted beneath the end of a railroad car ...."

even more telling .... watching how some be havin' to pick the nits
.

 

Exactly.

 

Okay...

I don't feel like I need to, because "brakes" are already regulated to a much stricter standard which has benefited from decades of participation from regulators, industry engineers, manufacturers, and repairers. FMVSS 105 is mostly harmonized with an international standard for "brakes" attended to by the U.N.

That chapter describes all sorts of standards and definitions and test procedures.

I have decided to put some trust in these institutions and their work.

I want to see at least that level of regulatory participation and oversight in place for something as advanced as what many people think of when they hear the words "full self driving."

Now I won't lay (all) blame on Tesla for this- I really think a few different regulatory agencies should have *ahem* hit the brakes on this a while back and dug further into the details.

Either way, it's time for FSD to catch the f up with "brakes" in terms of regulatory oversight and documentation.

 

I am in complete agreement with you about regulatory oversight.
My question about the brakes was because I wanted to ask you if ‘brakes’ should be removed until brake failures never happen?

I get that regulatory oversight should be stringent. But what if it is?

What if the presence of FSD(beta) means there are 5% of the deaths as there are without? Is it still a bad thing because of that rare death that still occurs?

 

Nope. Results like that are why I think it holds promise, yet to be validated by experts and regulators.

My point of view, my opinion on this overall is that more of this should have happened hand-in-hand with regulators and tested in closed circumstances. Later allowed out in public. Maybe that would have been by now anyway- we'll never know.

I'm a huge fan of the basic levels of driver assistance I get now- radar cruise, lane nudges and carplay, in a nutshell. I'm also very interested in completely automated mobility- SAE level 5 description. But I feel an aversion to that gulf between.

EDIT: adding:

I don't think it's a good comparison because the two systems are very different in terms of sophistication. When brakes fail, they fail to stop the car. A driver can be reasonably ready to use alternate methods for slowing and stopping.

What happens when FSD fails? more like... *. Wildcard. There are a broader range of categories of errors for you to watch out for, rather than a fairly basic pedal effort vs. deceleration with brakes.

I think a look at that level of complexity presented to the driver should factor into decisions about development in a closed environment vs. testing in the wild.

 

The last time I checked, Tesla only releases that information when forced to by the laws or regulations. Read more here: https://electrek.co/2022/01/12/tesla-forced-report-full-self-driving-beta-data-dmv-after-videos-showing-dangerous-use/

The news report about the current recall shared messages from Musk in which he objected to using the term "Recall" when the NHTSA instructed Tesla to notify customers.

 

But that's not the case, is it. A good straw man, but it's not the issue. The current FSD Beta not only requires constant driver supervision, it also requires frequent disengagement to prevent injury or property damage. While Tesla won't release the data, not even to the DMV, there is a group of Tesla owners doing the beta test and posting their statistics.

volunteer tesla data FSD Is a web site that has some neat statistics that don't seem to agree with what tesla testers on this forum want you to believe.

TL-DR: The fewer deaths question is just diversion. There is no documented and authenticated proof that any lives have been saved. The only factual data that we have comes from a group that is sharing their data via a Google web site. It shows that in some states the driver has to over-ride the FSD software very frequently in city driving. With a sample of ~ 70,000 miles, the drivers had to take control on an average of every 7 miles.

 

When the brake power assist system fails from any common sort of single fault, including loss of power, those federal regulations require that the brakes still work in old-fashioned manual mode. The regs specify a maximum stopping distance (fairly long), with a maximum pedal force (lots more force than normal braking), from a specified highway speed.

... especially considering that the driver is already engaged with the situation in order to discover that the brakes failed.

 

I often wonder how any Autonomous Vehicle could be programmed to deal with the “Trolley Problem”

 

Thank you, I misunderstood you.
When you said:

You weren’t looking at the events/failures, you were uncomfortable with the lack of regulations? Is that correct?

From what I understand, Europe is ahead of us in that arena. I agree there should be regulations, but I am also quite happy with the amount of in-house testing done on FSD. I don’t think people understand that Tesla does do both in-house testing as well as billions of miles of ‘ghost’ mode testing (which puts nobody at risk).

Anyway, thank you for the clarification

 

Sure thing! I don't make a big deal of the crash reports in the news because I work in media. I intuitively understand how easy it is to make scary news about anything the producer feels like that morning. Not thrilled about how it is used by some in the field, but that is another conversation altogether.

I'm frustrated by the overall lack of pace in development for autonomous driving- that's the old "I was promised flying cars by now!" talking.

I'm just a regular guy on the sidelines. I am not an automotive safety expert. But I cannot hold back the sense that the actual experts are either checked out or have been deliberately bypassed in an attempt to iterate faster; cast uninvolved in these efforts. There's certainly evidence of Tesla's CEO showing... distate... for regulators, and that makes me uneasy about any fruits his company's efforts may bear. After all, code gets recycled.

This all makes me sad because it seems like all the effort being thrown at the overall self-driving problem currently will at best become a template for a truly proper effort later. I'll probably get over it soon since it isn't my money.

 

Well yeah tesla does not respond well to California regulators, probably the animosity is one reason they chose to move the headquarters and expanded production outside the state. Tesla responded to that saying they don't test their level 4 system on public roads in the state. They do test their level 4 system on private roads and in other states, but all the California incidents were on the level 2 system. California objected to the name full self driving beta, which I agree with. Until they get the software and hardware to level 4, I don't think they should use that name. California did request that they put language in to accept that the driver is responsible and these are not self driving cars. That has popped up in various forms on the full self driving beta 3 times since I installed it in October of last year. This is like the warning label not to drink bleach, people still do it despite the label, and if you aren't an idiot you will not drink bleach.

I was referring to the NHTSA request to all car manufacturers in june 2021 to report all collisions from cars while using L1 and L2 ADAS starting within 30 days. Tesla has telematics that report all accidents and gave the NHTSA full data of every crash from 2019. Only Tesla and Honda gave data before june 2021 and the Honda data was not complete. NHTSA notes that the data from automakers is likely incomplete. Perhaps everyone with an L2 system should be recording all incidents like tesla for us to get a more complete picture.

NHTSA is behind the times as are many regulatory agencies. Recall is a bad word but it is the word that they use for things like pushed software as well as getting parts replaced, so it is misleading. I think the NHTSA should clarify what is changing and not use misleading words, but again tesla uses self driving which to most means L4 or L5 on a L2 system and that is even more confusing.

AAA has been testing these L2 systems. They tested some non teslas in 2020 on a private road and found that 66% of the time they hit a parked vehicle in part or completely in a lane and did not slow sufficiently. They tested various attention systems in 2022 and found camera based systems have a much faster reaction time to detect driver inattention than the steering wheel based system. Tesla turns on its camera based system on cars that have it (all model 3/Y, recent X/S) as part of the self driving beta. AAA recommended all L2 ADAS have visual monitoring. Mercedes and BMW have touch sensitive steering wheels so can not be fooled by weight. The other steering torque and visual systems can be fooled by weight or gag glasses with painted eyeballs placed where the camera is looking. IMHO those that use weights or gag glasses to fool these systems and get in an accident should be severely fined and/or sent to prison. Most of these incidents occur in California. I do agree with AAA that using these systems should require optical detection for the quicker response.

All Driver-Assist Systems We Tested Would Work with No Driver

We do have incidents like the 2021 double fatality in Texas where media were quick to blame autopilot. NHTSA finally issued its report this month. Autopilot was not engaged and it was in an area without lane markers so it would not have allowed engagement. The driver had twice the legal limit of alcohol, was speeding, hopped a curb and crashed into a tree. Yes you can crash any car if you act like that.

 

Strange question as I never made any claims about Toyota. I'm just laughing at Tesla...

Was your Outback advertised as having autopilot or self-driving features, comparable to a Tesla? I feel apples may be getting compared to oranges here.

 

I am too, not that I am frustrated that full self driving isn't here even though I paid for it. It was really to get computer hardware updates to have better L2 results. These actually are quite good, but I am frustrated I haven't had a second upgrade and that the cars still hit things parked in the wrong place and have worse phantom braking than they did two years ago (probably to avoid that first problem). I haven't had the misfortune of not paying attention or being in that first problem, but I have had to intervein numerous times on phantom braking. I do not trust summon mode, and self parking does great when it identifies a spot but has a hard time doing that well.

Well we can thank or curse Tesla for providing the NHTSA with a mountain of data on L2 ADAS. The problem could be that Autopilot is included in the price of every new tesla so that people just get it and use it without thinking about the warnings. It works well enough that it is used in a lot more situations than other systems, so a lot more miles in more dangerous situations. Agree that tesla's management has a distaste for regulators, but this is more of a war with the SEC they seem to comply well with NHTSA.

I do believe at least the level 2 will greatly improve. It is a tough problem to have. The system needs to detect stopped cars in the wrong place, cars entering the lane that you need to brake for, while ignoring things that shouldn't slow you down. My guess is my tesla has the sensors to do this but not nearly the computing power. The bigger problem is people believe today's systems can do more than they can, and then many make it sound as if they are more dangerous than they are, leading to distrust. That distrust will lead to a slower adoption when companies get it right.

 

Our Prius has appropriately warned us to "BRAKE!" once, it was faster than me at seeing the problem, but I got on the pedal before it did. I was faster than it on 3 other occasions that made me wonder if it should have triggered, but there's no practical way for me to know. It has thrown false "BRAKE!" warnings with no brake actuation maybe 5 or 6 times. These events were distributed over the almost 5 years we've owned it. I'm okay with that in terms of performance.

Two important distinctions there- the first is that data and operating code are different. I think the data is wonderful, and I hope it can be built on and used for all sorts of good things. The software is where I'm much more concerned.

The second, you make a fair point that Tesla shows a better record of complying with the NHTSA than other regulators.

My comments in this thread are that reflex distrust. But I should be clear that I don't fear driver assist technology done right and used appropriately. I save that for "people [who] believe today's systems can do more than they can," as you put it. I think it would be better if the driver assist feature had a klutzy name, some countermarketing if you will. Just different names and terminology, that's all. It's a really stupid blunder, and CEO Musk should be mocked for having made it. So in that sense I do hope there is a regulatory penalty eventually.