Current Events > US investigates Autopilot after 11 Teslas crashed into emergency vehicles

(+) Yeah! (+)

https://arstechnica.com/cars/2021/08/us-investigates-autopilot-after-11-teslas-crashed-into-emergency-vehicles/

Regulator worried Autopilot can't spot parked vehicles or keep driver engaged

They should've started the investigation sooner.

---

I fully believe that self driving cars are still a ways off, but...

Like many other ADAS systems, Autopilot requires a driver to keep their hands on the wheel, though such systems can be easily fooled by draping a weight over one of the steering wheels spokes.

These crashes are still a person's fault.

---

It's autopilot's fault that it can so easily be circumvented.

---

pogo_rabid posted...

It's autopilot's fault that it can so easily be circumvented.

The drivers are supposed to have their hands on the wheel at all times. This is like blaming the house for not having better locks when somebody breaks in and robs it because the owner left the doors wide open.

---

If you have to make driving motions on the wheel anyway why even have Autopilot in the first place?
---

Duncanwii posted...

If you have to make driving motions on the wheel anyway why even have Autopilot in the first place?

Because it's not real autopilot. Self driving cars aren't approved for road use. You're driving a beta test you opted into where you still have to drive but tesla gathers data.

---

ban "emergency" vehicles.

---

If all these crashes happened in a similar way then an investigation is warranted, but I'm not sure if 11 crashes over 3 years among 765k Teslas on the US roads is enough to speak of a pattern.

---

Think these autodriving crash investigations may end up being a driver thing again. Like in the end, was the driver doing everything to be in control of the vehicle? If theres any lapse on that, its gonna be the drivers fault because technically the driver checklists for the autodrive features werent being met right? Is that a valid conclusion to how these tesla autodrive investigations go?

---

maybe tesla shouldn't market it as fucking autopilot? hell, they even sell a "higher end service" they call full self driving...

---

Need more select all squares with a fire truck captchas to be solved to help train them.

---

This issue screams politicians and media trying to scare people over something just because it's new.

https://www.thewanderingrv.com/car-accident-statistics/

16,438 car crashes per day in the U.S.
1.25 million people killed in crashes every year globally.

And those are with humans behind the wheel, so it's a complete joke to focus on autopilot cars as a problem. They're the future. Humans suck at driving.

---

Turtlemayor333 posted...

This issue screams politicians and media trying to scare people over something just because it's new.

https://www.thewanderingrv.com/car-accident-statistics/

16,438 car crashes per day in the U.S.
1.25 million people killed in crashes every year globally.

And those are with humans behind the wheel, so it's a complete joke to focus on autopilot cars as a problem. They're the future. Humans suck at driving.

That's just part of a broader problem around ai and autonomy. It gets held to a significantly higher standard than the humans that it might replace, although there are some good reasons for that.

A lot of it comes down to explainability. If a driver crashes you can engage in dialogue to figure out what happened and why. With a lot of AI that's just not possible, and if you can't tell what went wrong you can't address the problem or assign fault which is why they're expected to be almost perfect.

---

obligatory that Anchorman 2 scene:
https://www.youtube.com/watch?v=IfY49zx7RU0

It boggles my mind that Tesla still hasnt crumbled. How many things have been promised years ago, that still havent seen the light of day?

Solar panel roofs? Tesla truck? Tesla Semi? Hyper loop? Boring company/tunnels project? Robo Taxis?

---

Are emergency vehicles really helping anyone? No.

Autonomous driving and teams of Boston Dynamics robots can easily take care of these circumstances.

Ban emergency vehicles and ban any manual human driving and give autonomous vehicles full reign of the road!

---

AngelsNAirwav3s posted...

It boggles my mind that Tesla still hasnt crumbled. How many things have been promised years ago, that still havent seen the light of day?

Solar panel roofs? Tesla truck? Tesla Semi? Hyper loop? Boring company/tunnels project? Robo Taxis?

I don't like Tesla.

Eventually all the carmakers will be implementing some form of this tech though.

---

This is a problem with advanced cruise control systems in general, not just Tesla. Hopefully they can figure out a solution for this, but I don't think it's likely.

---

Turtlemayor333 posted...

I don't like Tesla.

Eventually all the carmakers will be implementing some form of this tech though.

What tech? His batteries aren't even that good compared to what is on the market, let alone vs. a regular gas car.

---

CableZL posted...

Hopefully they can figure out a solution for this, but I don't think it's likely.

Autonomous cars are going to be a mess until all of them are, and a universal signalling protocol is implemented.

Questionmarktarius posted...

Autonomous cars are going to be a mess until all of them are, and a universal signalling protocol is implemented.

I'm inclined to agree with this. The problem in this regard with Teslas and other vehicles running into emergency vehicles is that advanced cruise control systems are programmed to ignore stationary objects when traveling at highway speeds. This is so that they ignore things like trees, overpasses, road signs, barriers, etc.

Tesla has rolled out a new purely vision-based autopilot, but I'm not sure if that will be able to help. Maybe, but maybe not. We'll have to see what they can develop with that. I can't wrap my head around how removing radar and only relying on cameras will benefit the system, but they seem to think the new system is better. We'll see.

---

AngelsNAirwav3s posted...

What tech? His batteries aren't even that good compared to what is on the market, let alone vs. a regular gas car.

What EV on the market has better batteries?

---

Questionmarktarius posted...

Autonomous cars are going to be a mess until all of them are, and a universal signalling protocol is implemented.

Tesla built their own underground tunnel system completely isolated from normal roads in Vegas, and they STILL couldn't get autonomous driving to work. Vegas blew like $120 million on this useless tunnel project lol

---

CableZL posted...

What EV on the market has better batteries?

https://fortune.com/2021/08/16/evs-how-long-does-one-charge-last-long-distance-ev-tesla-mercedes-eqs-lucid/

Mercedes is about to be 20% better.

---

CableZL posted...

Tesla has rolled out a new purely vision-based autopilot, but I'm not sure if that will be able to help. Maybe, but maybe not.

Probably not.

What we have now:


What we actually need:

AngelsNAirwav3s posted...

https://fortune.com/2021/08/16/evs-how-long-does-one-charge-last-long-distance-ev-tesla-mercedes-eqs-lucid/

Mercedes is about to be 20% better.

"About to be 20% better" = not on the market yet

Also

https://www.caranddriver.com/mercedes-benz/eqs

Mercedes offers the sizable EV sedan with either a 329-hp, rear-drive powertrain (EQS450+) or a 516-hp, all-wheel-drive setup (EQS580 4Matic). Both are powered by a large 107.8-kWh battery that should provide more than 300 miles of driving range. While calling the 2022 EQS an electrifying luxury car is an easy pun, it's also entirely accurate.

"Should provide more than 300 miles of driving range." - Tesla already has this. Any idea what the exact specs are? The article you posted has a pay wall. Mercedes doesn't seem to list the expected range on their web site, at least not that I've found so far.

---

AngelsNAirwav3s posted...

Tesla built their own underground tunnel system completely isolated from normal roads in Vegas, and they STILL couldn't get autonomous driving to work. Vegas blew like $120 million on this useless tunnel project lol

From what I'm reading, it's not that they "couldn't get it to work." It's that they're not allowing autonomous driving in the tunnels yet. There could be things they're trying to work out before allowing autonomous driving in the tunnels, but it's unclear from what I've seen.

You're speaking as if it's a "try it one time and it if it doesn't work, it won't ever work" kind of thing. The system is constantly in development, so just because something doesn't work one time doesn't mean they can't make it work.

---

So your definition what makes an EV a good vehicle is purely just the range? By that logic, a 1994 geo metro is one of the greatest ICE cars of all time because it gets almost 50 mpg.

---

https://www.motortrend.com/news/2022-mercedes-benz-eqs-vs-tesla-model-s-specs/

Granted we are talking about cars here that cost more than $100,000

---

CableZL posted...

I'm inclined to agree with this. The problem in this regard with Teslas and other vehicles running into emergency vehicles is that advanced cruise control systems are programmed to ignore stationary objects when traveling at highway speeds. This is so that they ignore things like trees, overpasses, road signs, barriers, etc.

Tesla has rolled out a new purely vision-based autopilot, but I'm not sure if that will be able to help. Maybe, but maybe not. We'll have to see what they can develop with that. I can't wrap my head around how removing radar and only relying on cameras will benefit the system, but they seem to think the new system is better. We'll see.

I don't think any "pure" system is going to fix the problem. Each has their own limitations and what's really needed is a hybrid system with a variety of signals and good data fusion. The fact that they're ignoring stationary objects in the first place is a problem, but it's an easy fix to a much harder problem to solve in intelligently filtering out motion.

The problem with a hybrid system is it means more sensors, and their business model doesn't really work if they're trying to sell an "affordable" car with $300,000 in sensors strapped to it that are difficult to maintain and replace.

---

Teslas are the pitbulls of the road.

s0nicfan posted...

I fully believe that self driving cars are still a ways off, but...

These crashes are still a person's fault.

Reminds me of a crash that happened with a CN Rail train. Regulations require that the conductor be at their station in the engine at all times to respond to things. That includes holding the gas down to go. Common practise at the time was to put a weight on the gas pedal and let the vehicle drive itself while you napped or did whatever. It led to a serious derailment of hazardous materials.

---

pogo_rabid posted...

So your definition what makes an EV a good vehicle is purely just the range? By that logic, a 1994 geo metro is one of the greatest ICE cars of all time because it gets almost 50 mpg.

We're specifically discussing batteries. The article he posted says the Mercedes EQS will have 20% better range than what Tesla offers, but I can't seem to find anything confirming that.

From his article:

The German automaker's new EQS 450+ is a real feat of engineering, one that will give drivers 20% more driving range on a full charge than Tesla's Model S.

It's lazy writing, IMO, because Tesla has different versions of the Model S with different amounts of range.

Model S Long Range:

• EPA rated 405 miles

Model S Plaid:

• EPA rated 396 miles

So, in order to get 20% better than either, you'd need to get around 475 miles of rated range in comparison to Plaid and 486 in comparison to the Long Range version. I'll be interested to see what Mercedes actually does with the EQS.

---

s0nicfan posted...

I don't think any "pure" system is going to fix the problem. Each has their own limitations and what's really needed is a hybrid system with a variety of signals and good data fusion. The fact that they're ignoring stationary objects in the first place is a problem, but it's an easy fix to a much harder problem to solve in intelligently filtering out motion.

The problem with a hybrid system is it means more sensors, and their business model doesn't really work if they're trying to sell an "affordable" car with $300,000 in sensors strapped to it that are difficult to maintain and replace.

I'm not so sure the removal of the "ignore stationary objects" programming is an easy fix, since every advanced cruise control system also has that programming. I think it would have been fixed by now if it was an easy fix.

---

CableZL posted...

So, in order to get 20% better than either, you'd need to get around 475 miles of rated range in comparison to Plaid and 486 in comparison to the Long Range version. I'll be interested to see what Mercedes actually does with the EQS.

Does it even say if they have the same battery capacity? If it gets 20% better range but it has a 20% larger battery then, well, duh.

---

AngelsNAirwav3s posted...

https://www.motortrend.com/news/2022-mercedes-benz-eqs-vs-tesla-model-s-specs/

Granted we are talking about cars here that cost more than $100,000

According to the WLTP test cycle, the Mercedes-Benz EQS offers up to 478 miles of range, which may be adjusted to under 400 miles by America's EPA cycle.

So they haven't had an official EPA rating yet, which is likely why Mercedes doesn't list the range. I'm not familiar with the differences in EPA testing vs the WLTP test cycle. Especially not enough to understand why Europe would give it 478 miles and why one would be able to predict that the EPA might adjust it to under 400.

---

CableZL posted...

According to the WLTP test cycle, the Mercedes-Benz EQS offers up to 478 miles of range, which may be adjusted to under 400 miles by America's EPA cycle.

So they haven't had an official EPA rating yet, which is likely why Mercedes doesn't list the range. I'm not familiar with the differences in EPA testing vs the WLTP test cycle. Especially not enough to understand why Europe would give it 478 miles and why one would be able to predict that the EPA might adjust it to under 400.

My point is, they aren't using patented Tesla technology are they? Tesla is extremely over priced, I am getting some serious Theranos vibes from how many projects they have announced that have gotten world wide news coverage that will never see the light of day.

---

CableZL posted...

I'm not so sure the removal of the "ignore stationary objects" programming is an easy fix, since every advanced cruise control system also has that programming. I think it would have been fixed by now if it was an easy fix.

It's explicitly an extremely hard fix. In case it was unclear, "ignore stationary objects" IS the "easy fix" to the much harder problem of "contextually interpret stationary objects in real time to determine the level of risk to the vehicle"

---

AngelsNAirwav3s posted...

My point is, they aren't using patented Tesla technology are they? Tesla is extremely over priced, I am getting some serious Theranos vibes from how many projects they have announced that have gotten world wide news coverage that will never see the light of day.

Your point was that Teslas don't have the best batteries on the market and you used a car that isn't currently on the market as an example. A car that hasn't received an EPA rating yet for its range at that. Hopefully Mercedes can push their battery technology to that level, but they haven't done it just yet.

Also, lots of companies have multiple projects running that don't see the light of day. Tesla isn't unique in that regard.

---

s0nicfan posted...

It's explicitly an extremely hard fix. In case it was unclear, "ignore stationary objects" IS the "easy fix" to the much harder problem of "contextually interpret stationary objects in real time to determine the level of risk to the vehicle"

Oh, I see. My mistake.

---

pogo_rabid posted...

So your definition what makes an EV a good vehicle is purely just the range? By that logic, a 1994 geo metro is one of the greatest ICE cars of all time because it gets almost 50 mpg.

Well Doug DeMuro just called the Model S Plaid the best car ever made. You post his vids so I assume you respect his opinion.

---

You act as if what he says is the word of god. Is it a bad car? Not at all, but the best car ever? Laughable.

---

pogo_rabid posted...

You act as if what he says is the word of god. Is it a bad car? Not at all, but the best car ever? Laughable.

As stated, I brought him up because you frequently post his videos. That is not acting like what he says is the word of god.

---

AngelsNAirwav3s posted...

My point is, they aren't using patented Tesla technology are they? Tesla is extremely over priced, I am getting some serious Theranos vibes from how many projects they have announced that have gotten world wide news coverage that will never see the light of day.

The Boring Company is separate from Tesla. If you want to say Elon is a clown, agreed.

The Model Y alone having enough demand to be among the best selling cars changed some priorities at Tesla. In a battery and chip constrained world, theyre delaying resources from battery hungry trucks and into Ys that now have a 6 month waiting list.

It is the legacy automakers that are Theranos. Their only affordable offering is in total recall because their crap pouch batteries arent safe.

The others are still low vol and/or 100k+ vaporware stuffed with extra battery to compensate for how far behind their efficiencies are. Never mind their software structure thats stuck in 2005.

Anyone can announce a paper tiger prototype. Manufacturing the EV at scale is the challenge. And Teslas the only one doing it.


---

Remember when the Toyota unintended acceleration was determined to be a stupid driver fault?

Hornezz posted...

If all these crashes happened in a similar way then an investigation is warranted, but I'm not sure if 11 crashes over 3 years among 765k Teslas on the US roads is enough to speak of a pattern.

There been a few times where autopilot definitely caused the crash and even counting those times autopilot is still safer than driving manually.

---

Trumpo posted...

Remember when the Toyota unintended acceleration was determined to be a stupid driver fault?

I thought it was a floor mat issue?

---