OT: Tesla Autopilot

I posted this in a Tesla forum, a collection of my thoughts on autopilot. I thought others here might be interested. It was in response to some blami ng driver inattentiveness alone for a fatal accident when the car plowed in to a crash attenuator while on autopilot.

-------- I've driven my Tesla a number of highway miles on autopilot and I have a fe w things to say. When I am trying to exert enough pressure on the steering wheel to prevent the alerts but little enough to not take control back, I have to spend some portion of my focus on that task, periodically checking to see if I am pushing hard enough on the steering wheel. Most of my audib le alerts happen when I am distracted from that task by safety related even ts on the road in front of me. The mere fact that I receive an audible ale rt when I am trying to focus on the road is itself a distraction from drivi ng.

Then there is the issue of having to not only be wary of events happening o n the road due to other cars, but now I have to be constantly vigilant agai nst the car deciding it wants to do something dangerous. This is not a pro blem to wave off with a shrug.

A Tesla on autopilot is not a normal car. It is a car that lulls you into a false sense of security only to demand your full attention with a moments warning.

I've also noticed an issue I attribute to muscle memory. When I drive, my hands have a feel for driving the car. When I need to turn into a curve th e movement has little to do with my brain and a lot to do with muscle memor y. This is different depending on whether you are actually steering the ca r or you are trying to exert some amount of pressure on the wheel while jus t following the car controlling the wheel. When my eyes are on the road an d my brain is actively monitoring the situation in front of me, it is *not* always monitoring the handling of the steering wheel. So there have been times when I think I am driving and I start steering around the curve only to get the beeps that the car has given me control back. Likewise there ar e times when I think the car is driving and it is thinking I am driving. T his results in my drifting further across my lane but never my drifting out of my lane. Either way it surprises me but I can't say it is distracting enough to be dangerous. It is creepy though.

I guess the point I'm getting to is that I don't think registering pressure by the driver on the steering wheel is a useful way to determine if hands are on the wheel or if the driver is being attentive. In other words, this is very inaccurate. Assuming any given accident happens because the car r ecord shows the driver does not have hands on the wheel is a fallacy. The car has no way of knowing if the driver's hands are on the wheel or if the driver is being attentive.

If the car is good at determining when the driver is being attentive, why i s this only active when the car is on autopilot? Wouldn't it be a good ide a to have this active at *all* times? When is it good to have an inattenti ve driver????

Rick C.

Tesla referral code -

formatting link

Reply to
gnuarm.deletethisbit
Loading thread data ...

n of my thoughts on autopilot. I thought others here might be interested. It was in response to some blamin g driver inattentiveness alone for a fatal accident wh en the car plowed into a crash attenuator while on aut opilot.

er of highway miles on autopilot and I have a few thin gs to say. When I am trying to exert enough pressure on the steering wheel to prevent the alerts but little enough to not take control back, I have to spend some portion of my focus on that task, periodically checki ng to see if I am pushing hard enough on the steering wheel. Most of my audible alerts happen when I am dis tracted from that task by safety related events on the road in front of me. The mere fact that I receive an audible alert when I am trying to focus on the road i s itself a distraction from driving.

is the issue of having to not only be wary of events happening on the road due to other cars, but now I hav e to be constantly vigilant against the car deciding i t wants to do something dangerous. This is not a prob lem to wave off with a shrug.

lot is not a normal car. It is a car that lulls you i nto a false sense of security only to demand your full attention with a moments warning.

ticed an issue I attribute to muscle memory. When I d rive, my hands have a feel for driving the car. When I need to turn into a curve the movement has little to do with my brain and a lot to do with muscle memory. This is different depending on whether you are actual ly steering the car or you are trying to exert some am ount of pressure on the wheel while just following the car controlling the wheel. When my eyes are on the r oad and my brain is actively monitoring the situation in front of me, it is *not* always monitoring the hand ling of the steering wheel. So there have been times when I think I am driving and I start steering around the curve only to get the beeps that the car has given me control back. Likewise there are times when I thi nk the car is driving and it is thinking I am driving. This results in my drifting further across my lane b ut never my drifting out of my lane. Either way it su rprises me but I can't say it is distracting enough to be dangerous. It is creepy though.

e point I'm getting to is that I don't think registeri ng pressure by the driver on the steering wheel is a u seful way to determine if hands are on the wheel or if the driver is being attentive. In other words, this is very inaccurate. Assuming any given accident happe ns because the car record shows the driver does not ha ve hands on the wheel is a fallacy. The car has no wa y of knowing if the driver's hands are on the wheel or if the driver is being attentive.

s good at determining when the driver is being attenti ve, why is this only active when the car is on autopil ot? Wouldn't it be a good idea to have this active at *all* times? When is it good to have an inattentive driver????

code -

formatting link

That would se em to be as much work as self driving, early days I gu ess. I suppose it will come down to whether you trus t the autonomous driving or not, I'm inclined to thi nk they will get it right particularly when the cars talk to each other.

Reply to
Rheilly Phoull

n of my thoughts on autopilot. I thought others here might be interested. It was in response to some blamin g driver inattentiveness alone for a fatal accident wh en the car plowed into a crash attenuator while on aut opilot.

er of highway miles on autopilot and I have a few thin gs to say. When I am trying to exert enough pressure on the steering wheel to prevent the alerts but little enough to not take control back, I have to spend some portion of my focus on that task, periodically checki ng to see if I am pushing hard enough on the steering wheel. Most of my audible alerts happen when I am dis tracted from that task by safety related events on the road in front of me. The mere fact that I receive an audible alert when I am trying to focus on the road i s itself a distraction from driving.

is the issue of having to not only be wary of events happening on the road due to other cars, but now I hav e to be constantly vigilant against the car deciding i t wants to do something dangerous. This is not a prob lem to wave off with a shrug.

lot is not a normal car. It is a car that lulls you i nto a false sense of security only to demand your full attention with a moments warning.

ticed an issue I attribute to muscle memory. When I d rive, my hands have a feel for driving the car. When I need to turn into a curve the movement has little to do with my brain and a lot to do with muscle memory. This is different depending on whether you are actual ly steering the car or you are trying to exert some am ount of pressure on the wheel while just following the car controlling the wheel. When my eyes are on the r oad and my brain is actively monitoring the situation in front of me, it is *not* always monitoring the hand ling of the steering wheel. So there have been times when I think I am driving and I start steering around the curve only to get the beeps that the car has given me control back. Likewise there are times when I thi nk the car is driving and it is thinking I am driving. This results in my drifting further across my lane b ut never my drifting out of my lane. Either way it su rprises me but I can't say it is distracting enough to be dangerous. It is creepy though.

e point I'm getting to is that I don't think registeri ng pressure by the driver on the steering wheel is a u seful way to determine if hands are on the wheel or if the driver is being attentive. In other words, this is very inaccurate. Assuming any given accident happe ns because the car record shows the driver does not ha ve hands on the wheel is a fallacy. The car has no wa y of knowing if the driver's hands are on the wheel or if the driver is being attentive.

s good at determining when the driver is being attenti ve, why is this only active when the car is on autopil ot? Wouldn't it be a good idea to have this active at *all* times? When is it good to have an inattentive driver????

code -

formatting link

Haven't exper ienced the autopilot stuff yet but I believe I agree. I think autonomous vehicles can be made pretty relia ble and likely better at getting me somewhere safer than any human cab driver (gosh, doesn't seem too ha rd) but I don't really want some half-human, half-auto pilot system where the car is, by design, at any mom ent going to say "I give up!" and throw it back into my hands.

Either make a fully autonomous car or le t me drive myself but I don't want half and half. Ha lf-and-half "autopilot" works out better in a passen ger jet aircraft and likely does increase passenger sa fety significantly, a ground passenger vehicle is no t a 737-800.

The only autonomous-like feature I can see myself wanting in my next vehicle is emergency auto-brake/adaptive cruise control, in that one spec ific situation where you're following behind someone i n a lane on a crowded highway it can react faster th an I can. One always tries to leave appropriate spac e but not always possible with people often cutting in front with little room to spare.

Reply to
bitrex

The autopilot-controlled car driver should handle the driving in the same way as an airplane pilot flying on autopilot. The device is a great help indeed, taking care of the details, but the driver must be awake and ready to take over if needed.

We have special training for pilots to learn hwo to fly using the autopilot. Maybe there should be similar training for Tesla drivers.

--

-TV
Reply to
Tauno Voipio

America isn't Germany, driver's licenses are trivial to get and once someone has one you essentially have to negligently kill someone to get it revoked for any significant length of time in most areas.

Nobody's takin no friggin' class, nobody spends 50, 60, 70k on a car to be forced to take some driver's ed class that shit is for poor people.

Reply to
bitrex

there are suburban drivers in my area who probably wreck out a couple of their own cars every year plus the car of whomever they hit, they still driving around just fine. It doesn't take a particularly high-speed impact to total out a modern vehicle.

Reply to
bitrex

t. I thought others here might be interested. It was in response to some b laming driver inattentiveness alone for a fatal accident when the car plowe d into a crash attenuator while on autopilot.

a few things to say. When I am trying to exert enough pressure on the stee ring wheel to prevent the alerts but little enough to not take control back , I have to spend some portion of my focus on that task, periodically check ing to see if I am pushing hard enough on the steering wheel. Most of my a udible alerts happen when I am distracted from that task by safety related events on the road in front of me. The mere fact that I receive an audible alert when I am trying to focus on the road is itself a distraction from d riving.

ng on the road due to other cars, but now I have to be constantly vigilant against the car deciding it wants to do something dangerous. This is not a problem to wave off with a shrug.

nto a false sense of security only to demand your full attention with a mom ents warning.

my hands have a feel for driving the car. When I need to turn into a curv e the movement has little to do with my brain and a lot to do with muscle m emory. This is different depending on whether you are actually steering th e car or you are trying to exert some amount of pressure on the wheel while just following the car controlling the wheel. When my eyes are on the roa d and my brain is actively monitoring the situation in front of me, it is * not* always monitoring the handling of the steering wheel. So there have b een times when I think I am driving and I start steering around the curve o nly to get the beeps that the car has given me control back. Likewise ther e are times when I think the car is driving and it is thinking I am driving . This results in my drifting further across my lane but never my drifting out of my lane. Either way it surprises me but I can't say it is distract ing enough to be dangerous. It is creepy though.

sure by the driver on the steering wheel is a useful way to determine if ha nds are on the wheel or if the driver is being attentive. In other words, this is very inaccurate. Assuming any given accident happens because the c ar record shows the driver does not have hands on the wheel is a fallacy. The car has no way of knowing if the driver's hands are on the wheel or if the driver is being attentive.

hy is this only active when the car is on autopilot? Wouldn't it be a good idea to have this active at *all* times? When is it good to have an inatt entive driver????

In Tesla's defense, autopilot is "beta" and you are given warnings about be ing prepared to take control at a moments notice. Also, they have pointed out numerous times that "auto pilot" is much like that in an airplane. It isn't going to manage every aspect of flying, just steering and altitude co ntrol. It is not an invitation for the pilot to take a nap.

"half-human, half-autopilot" is a total mis-characterization.

No one is stopping you from driving yourself. Please continue. As I point ed out to someone complaining about the new beta navigator assisted autopil ot, they release beta versions only in part to get feedback from the users. It is more about collecting driving experience data from many miles of us er testing. That is the important part.

I'm never sure what modes are activated or not, but as far as I can tell I have the emergency braking enabled and the few times I've tested it I can't see where it is working at all. I'm only willing to get so close before I hit the brakes.

My point in making the original post was not about the car's performance in these beta test modes. My point was about the way they sense driver atten tiveness. I think requiring the driver to exert force against the wheel is a very poor method for the various reasons I gave in the original post.

Rick C.

Tesla referral code +

formatting link

Reply to
gnuarm.deletethisbit

ot.

sure

s you

turn

ou

en

ss my

ses

is

ing any

it be

to

That's simply not true. You may require several speeding tickets to lose y our license, but killing someone is definitely not required. Every state I 've had a license in has a point system where you got points for each and e very moving violation.

There was one driver who apparently was watching a DVD when a truck pulled in front of him and the car's cameras did not see it because the trailer wa s white against the sky. Proper training that emphasized how you can't rel y on the autopilot may well have saved his life. They really don't give yo u much education on the car. I would have liked having a *lot* more info o n charging the battery. Turns out there is a lot more to it than just plug ging into Superchargers every time you want.

Rick C.

Tesla referral code --

formatting link

Reply to
gnuarm.deletethisbit

n

ssure

shrug.

ls

I o

in is

een

oss

ming any

f

When

Gotta call male bovine coprolite on that one (old bull shit). I don't know of anywhere in the US where you can afford to insure your driving after a few accidents and the law doesn't take kindly unless they are just parking lot fenderbenders.

Rick C.

Tesla referral code -+

formatting link

Reply to
gnuarm.deletethisbit

In MA if you have enough wrecks to the point that you're considered so high risk that the open market won't insure you at any price you get put in a state insurance pool for high-risk drivers so you can keep on truckin'. The reasoning being that it's better to provide some always-available option for high-risk drivers who haven't outright killed anyone than nothing, if it's nothing they'll just drive anyway.

You end up pay a very high rate like maybe $700/month, more than a low-income driver could afford, the pool is for wealthy bad drivers, job creators, on-the-regular BMW-crashers.

The person who totaled out my first Volt was in that pool and lived in my town so I vaguely knew who they were, husband get her a new car before the weekend was out no problem back on the road.

Reply to
bitrex

The penalties are usually absurdly lenient like four or five wrecks/moving violation conviction in two years and you lose your license for a month, big whoop. I don't consider a month to be "a significant length of time" YMMV

If you have three DUIs it's on paper pretty bad news in most states but those are also successfully appealed all the time, too.

Education is counterproductive to selling replacement cars!

Reply to
bitrex

I think he was probably educated OK, that it was a beta version, it's just that Tesla's autonomous system is probably very good already. He was surely cautious at first, but then over time and with "experience" he saw that it managed tricky situations just fine without any significant input from him.

He thought he understood its capabilities enough to relax, gained faith in it, it had worked great 100 times before. Began to anthropomorphize the machine and thought of it as his buddy maybe who would always "have his back."

Turns out it wasn't, it worked great 99.9% percent of the time until the day it didn't and "betrayed" him. It was never his buddy it was just a computer running some algorithms. It's a sadly not uncommon tale wrt technology-human relations

Reply to
bitrex

BTW this is my hypothesis because if I'm not wrong there's actually been a fair amount in the media written about the life and times of that particular driver; he was former military/Navy and seemingly rather well-educated and bright judging by how his associates tell of him. He wasn't just some schmo who jumped in a new Tesla and started watching DVDs on the highway first day.

Reply to
bitrex

I suppose the way you test it is to set a cardboard barrier on the hiway and drive into it at full speed. I need to try that with our two cars.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Isn't the second test to be driving behind another (friendly) car at the same speed, then the other car pulls out into another lane to go round the cardboard. For a video of what happens under /ideal/ conditions, see

formatting link

Or what happens when it isn't following a car.

formatting link

The key point is that /partial/ automation has long been known to be problematic. Apparently if the autopilot gives up and hands control back to the driver, it takes up to 17s(!) for the driver to see, assess and react.

formatting link

Reply to
Tom Gardner

of

l

know of anywhere in the US where you can afford to insure your driving afte r a few accidents and the law doesn't take kindly unless they are just park ing lot fenderbenders.

Yeah, but the costs are commensurate and at thousands of dollars a year mos t just give in and stop driving.

What I find funny is I've seen people who end up on the state insurance poo l where the rates are still dependent on the vehicle they drive and they st ill won't give up on the 500 horse muscle car that costs them half the cost of the car each year to insure it!

Yeah, a rich bitch hit me in a parking lot and jumped out of her car and ye lled, "You g...d... West Virginia hick!" LOL Her insurance company paid t he damage on her Mercedes... I didn't really have any. The car was leased by their company.

Rick C.

Tesla referral code +-

formatting link

Reply to
gnuarm.deletethisbit

led in front of him and the car's cameras did not see it because the traile r was white against the sky. Proper training that emphasized how you can't rely on the autopilot may well have saved his life. They really don't giv e you much education on the car. I would have liked having a *lot* more in fo on charging the battery. Turns out there is a lot more to it than just plugging into Superchargers every time you want.

Absolutely not correct. The Tesla autopilot is good at coloring between th e lines, but it is all the "tricky situations" where it is pretty much crap . The one fatal wreck would appear to be a case of the car not deciding wh ether to go left or right at a fork instead running up the middle. I have seen it do this. Once I saw it take the exit (which was actually the corre ct thing to do, but how would it know that?) instead of staying on the main road? It often would jog to the right on both exit and entrance ramps whe re the white line diverges and the dotted line isn't there yet.

Just the other day (this is post release 9 which improves on this behavior in general) it twitched to the left for a short de-acceleration turn lane w hich would have been very bad. The car is fond of these little surprises. They do still call it a beta release after all!

So no one in their right mind is going to be "lulled" into a false sense of security with such reminders.

The car definitely doesn't "have your back"... It can't even manage the wi ndshield wipers or high beams automatically all the time. All of the "beta " systems are... well, beta!

Hardly. Normally such safety related systems aren't allowed to be handled automatically unless the systems can be proven to operate correctly. In th e case of autos we don't seem to have clear guidelines on how good they sho uld be. Ultimately when the car is running autonomously in a real sense re sponsibility will be with the companies designing the cars. There's no oth er way to assign responsibility that provides the needed feedback to optimi ze safety.

Rick C.

Tesla referral code ---

formatting link

Reply to
gnuarm.deletethisbit

No need. I don't actually care if the car brakes or not if it is only goin g to brake in the event a collision is immanent. I want it to brake in a w ay that won't take years off my life even if I don't have the accident. Ha rd braking is one thing, emergency braking is another. This thing has *hug e* tires, but then it weighs nearly 6,000 lbs too.

Rick C.

Tesla referral code --+

formatting link

Reply to
gnuarm.deletethisbit

I'm not watching a 30 second ad to see this video. But I already know the Tesla fails and plows into the barrier. But the fault wasn't because the a utonomous car didn't react appropriately, the problem was *following too cl osely*!!! I don't think a human driver could have avoided the obstacle in traffic if they weren't sure an adjacent lane was empty.

I may lose focus while messing with the entertainment display (more than ju st FM radio these days) but otherwise I pay full attention to the road. Th e autopilot has on more than one occasion commanded me to take control for no apparent reason (not related to my attention or lack thereof). I think in one case it simply could not find the lane markings as it passed through an intersection. Not sure in the other cases. I can see what's going on just fine and I am not aware of what the car is not seeing. It's a bit dis turbing, but I've gotten used to it.

Lol, I was giving some friends a ride once. The cruise control uses a chim e to announce it is engaging or disengaging (other than by a scram). The f riend in the front seat wasn't used to a car making such sounds and was fri ghtened each time I hit the brake to approach a light. She thought the car was going to do something erratic!

It is a fun car, but as I keep pointing out to people, the rest of it doesn 't matter if you can't get it charged. My trips to Maryland have been a bi t hard because there is not just no Supercharger in Frederick, there's noth ing faster than a J-1772 available (about four times faster than 120V, 15A which is walking speed). So while posting in the Tesla forums I found ment ion of a new installation of Superchargers on my preferred route! So now, even in winter, I will be able to complete my trip with 120V charging overn ight at my friends house and then a 10 minute top off at the Supercharger o n the way back. Charging from 120V rather than hitting Superchargers all t he time is better for the battery too, so this is a great solution.

Rick C.

Tesla referral code -+-

formatting link

Reply to
gnuarm.deletethisbit

You are too sure of yourself; you are thinking of a different video. There is no barrier in this video. There is no other traffic, except for the car in front that is an integral part of the demo. It is on a test track in good weather.

If it gets it seriously wrong under those ideal conditions, what chance does it (or innocent bystanders) have in normal conditions?

Curiously that is the wrong reference, and neither of us spotted it. I didn't spot it because of my copy-paste error. You didn't spot it because?

A better reference is

formatting link
but that's not the one I previously found, since it mentions 20s, not 17.

That's a different situation: you are already partially concentrating on driving.

You seem to be a fanboi with pre-determined beliefs - and you are going to skip over evidence that conflicts.

I can think of other people in this group with the same attitude to a different topic.

Reply to
Tom Gardner

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.