Model 3 crash on autopilot

Model 3 crash on autopilot

Author
Discussion

MarkwG

4,859 posts

190 months

Wednesday 11th December 2019
quotequote all
wired said:
Just about everybody working on a fully self-driving system—the kind that doesn't depend on lazy, inattentive humans for support—plans to use lidar, along with radar and cameras.

Except for Elon Musk. The Tesla CEO insists he can make his cars fully autonomous—no supervision necessary—with just radars and cameras. He hasn't proven his claim just yet, and no one knows if he ever will.
..and the bit that worries me - it's all very well being an Alpha male, reaching for the stars, but there's a good reason for doing it properly rather than charging on regardless. One might think it's a cynical ploy to take money off stupid people with no thought of the consequences...

RobDickinson

31,343 posts

255 months

Wednesday 11th December 2019
quotequote all
MarkwG said:
..and the bit that worries me - it's all very well being an Alpha male, reaching for the stars, but there's a good reason for doing it properly rather than charging on regardless. One might think it's a cynical ploy to take money off stupid people with no thought of the consequences...
They have a very good argument against Lidar, who knows if it will work out for them.

What the car is seeing now (on HW3) is very impressive.

Car warned me yesterday about a ~4yo kid on the edge of a pavement among usual clutter ( road was bending left and up so if I didnt steer I would have run him over). Spots road cones 50-100m out etc.

mcdjl

5,451 posts

196 months

Wednesday 11th December 2019
quotequote all
RobDickinson said:
Wow. I find that scary. Volvo thinks the best thing to do is accelerate at the static object in the way. Mind you at least that's active cruise control which sounds a lot less control than autopilot, even if it isn't. I'd really like to think those systems had the resolution to tell the difference between a static thing in the way and not in the way.

RobDickinson

31,343 posts

255 months

Wednesday 11th December 2019
quotequote all
Its over a year old so AP might be doing something else now but its still not perfect as it relies on radar a lot.

In the last ~month we've had updates that improve lane changes, show road cones, much smoother driving and features to slow the car down if its going substantially faster than adjacent lanes, making the car move over in the lane when passing a truck etc etc

Edited by RobDickinson on Wednesday 11th December 22:26

CanAm

9,239 posts

273 months

Wednesday 11th December 2019
quotequote all
g3org3y said:
So what happened in this instance then? As far as I understand it, the obstruction was directly in front of the Tesla. Why didn't the auto braking system activate?
From the www.wired.com article:-
"The car's manual does warn that the system is ill-equipped to handle this exact sort of situation: “Traffic-Aware Cruise Control cannot detect all objects and may not brake/decelerate for stationary vehicles, especially in situations when you are driving over 50 mph (80 km/h) and a vehicle you are following moves out of your driving path and a stationary vehicle or object is in front of you instead.”

So a human driver swerves to avoid a stationary object, but autopilot doesn't!!

MarkwG

4,859 posts

190 months

Thursday 12th December 2019
quotequote all
RobDickinson said:
They have a very good argument against Lidar, who knows if it will work out for them.

What the car is seeing now (on HW3) is very impressive.

Car warned me yesterday about a ~4yo kid on the edge of a pavement among usual clutter ( road was bending left and up so if I didnt steer I would have run him over). Spots road cones 50-100m out etc.
Impressive it may be, but autopilot is not yet fit for public use, & the somewhat cavalier way Mr Musk & co demonstrate by letting their customers live test it with minimal instruction & guidance doesn't fill me with confidence. The software industry attitude of beta testing isn't appropriate when innocent third parties could be injured or killed by bugs in the system.

RobDickinson

31,343 posts

255 months

Thursday 12th December 2019
quotequote all
It's 8 times safer yet not fit for use. Lol.

MarkwG

4,859 posts

190 months

Thursday 12th December 2019
quotequote all
RobDickinson said:
It's 8 times safer yet not fit for use. Lol.
If you're quoting this, for example - https://eu.usatoday.com/story/money/cars/2019/09/1... - that's how it performs IN an accident, not how it avoids them. There's no data to compare on how well humans avoid accidents, because the accident doesn't happen. Therefore any statistics based on that are flawed from the start.

RobDickinson

31,343 posts

255 months

Thursday 12th December 2019
quotequote all
I posted the data on the previous page.

But yes teslas are safer than other cars if they do crash

kambites

67,593 posts

222 months

Thursday 12th December 2019
quotequote all
MarkwG said:
If you're quoting this, for example - https://eu.usatoday.com/story/money/cars/2019/09/1... - that's how it performs IN an accident, not how it avoids them. There's no data to compare on how well humans avoid accidents, because the accident doesn't happen. Therefore any statistics based on that are flawed from the start.
One can, however, look at number of accidents per mile driven (or various other metrics) for a car with a large suite of driver aids vs one without, and cars with driver aids such as autopilot universally come out as less likely to be involved in accidents when you look at figures in that way.

The thing with Autopilot is that because of the huge amount of public interest in Tesla, accidents such as this tend to get reported in the press. The thousands of fatal accidents a day caused by poor driving on the part of humans do not.


To a large degree this is why we're going to have to have an attitude of "the driver is ultimately in charge" long past the point where computers can do just about everything better than a human driver (although even that point is some way off in practice).

Edited by kambites on Thursday 12th December 08:59

Dave Hedgehog

14,569 posts

205 months

Thursday 12th December 2019
quotequote all
CanAm said:
So a human driver swerves to avoid a stationary object, but autopilot doesn't!!
on the current level of software or older and especially on older hardware its not situationally aware enough, AP is very dumb it follows a 'basic' set of code in comparison to a human driving, its not powered by a fully self aware AI'd terminator

even the best systems are still bad compared to a human driver

musk's logic about not using lidar is sound in so far as we can drive perfectly fine without any radar or laser range finding equipment, where it falls down is teaching a computer to replicate how a human works is turning out to be vastly harder than initially thought

kambites

67,593 posts

222 months

Thursday 12th December 2019
quotequote all
Dave Hedgehog said:
even the best systems are still bad compared to a human driver
Compared to which human driver? There's quite a big variety!

kiseca

9,339 posts

220 months

Thursday 12th December 2019
quotequote all
MarkwG said:
RobDickinson said:
It's 8 times safer yet not fit for use. Lol.
If you're quoting this, for example - https://eu.usatoday.com/story/money/cars/2019/09/1... - that's how it performs IN an accident, not how it avoids them. There's no data to compare on how well humans avoid accidents, because the accident doesn't happen. Therefore any statistics based on that are flawed from the start.
There is. Number of miles travelled per recorded accident. If Teslas on autopilot have enough miles to make a reliable average, and they travel more miles per crash than average, then the system would be proving to be safer overall than an unassisted human.


MarkwG

4,859 posts

190 months

Thursday 12th December 2019
quotequote all
RobDickinson said:
I posted the data on the previous page.

But yes teslas are safer than other cars if they do crash
No link that I can find - you state it, & link to the Wired article & a youtube video, is that what you mean? If I've missed it, perhaps you could point me to it?

RobDickinson

31,343 posts

255 months

Thursday 12th December 2019
quotequote all
I posted a graph compiled from tesla safety report data.

MarkwG

4,859 posts

190 months

Thursday 12th December 2019
quotequote all
kiseca said:
There is. Number of miles travelled per recorded accident. If Teslas on autopilot have enough miles to make a reliable average, and they travel more miles per crash than average, then the system would be proving to be safer overall than an unassisted human.
Ok, so I'll try explaining differently: if a human driver sees a hazard & avoids it, then there will be no record of the event - there has been no collision to record. If he doesn't see, & doesn't avoid, there will be. The first scenario will be happening all the time. Some of automated systems above acknowledge they cannot handle that first scenario. Therefore, any data that shows they are "safer" that doesn't take that into account, leave me with an uncomfortable feeling about it's voracity. Some manufacturers are reluctant to launch the technology until they've resolved that difficulty, others are happy to. I've no axe to grind for or against automation,however statistics that don't paint the whole picture & are used to push a particular view point bother me.

Dave Hedgehog

14,569 posts

205 months

Thursday 12th December 2019
quotequote all
kambites said:
Dave Hedgehog said:
even the best systems are still bad compared to a human driver
Compared to which human driver? There's quite a big variety!
taking an average

MarkwG

4,859 posts

190 months

Thursday 12th December 2019
quotequote all
RobDickinson said:
I posted a graph compiled from tesla safety report data.
Doesn't show for me, I'm afraid, might be the company filters I suppose - just shows an icon about free software - computers eh, can't trust 'em wink

jamoor

14,506 posts

216 months

Thursday 12th December 2019
quotequote all
I went from Sheffield to Manchester but took the m62 on purpose as autopilot works flawlessly on the motorway and the extra 30 miles cost 1.20 in electricity.

There was rain, patchy lines, heavy traffic and it was dark too. The car didn’t miss a single thing, the only issue now is that you can get bored while driving as you’re watching a computer drive rather and doing it yourself.

Lt. Coulomb

202 posts

55 months

Thursday 12th December 2019
quotequote all
RobDickinson said:
Car warned me yesterday about a ~4yo kid
Tread carefully, ”pedo guy”...