Tesla's autopilot, on the road

[Oils4AsphaultOnly]

Does full self driving require the user to keep a hand on the wheel? The videos suggest not, but I don't know definitively.

Ultimately, no, since FSD isn't a driving aid. During the beta tests, not sure, since I'm not in the test pool.

[Oils4AsphaultOnly]

Pro Pilot forces the driver to keep paying attention, and keep hands on the wheel. Since people are sleeping with Tesla's AP on, I don't think it behaves the same way.

People who sleep with AP are using a defeat device. otherwise they would've been forced to pay attention as well.

[SageBrush]

Does full self driving require the user to keep a hand on the wheel?

During the beta tests,

Yes, and more. It was Tesla's intent to only allow careful and attentive drivers into the Beta test. One can presume that the filter was not entirely successful but the Beta driver pool is far from random from a safe driver perspective. I've wondered more than once how that filter was implemented. The nerd in me says that the car graded its driver.

Think about that for a moment. We may have already crossed the line from the car assisting the driver to the driver assisting the car.

[coleafrado]

Does full self driving require the user to keep a hand on the wheel?

During the beta tests,

Yes, and more. It was Tesla's intent to only allow careful and attentive drivers into the Beta test. One can presume that the filter was not entirely successful but the Beta driver pool is far from random from a safe driver perspective. I've wondered more than once how that filter was implemented. The nerd in me says that the car graded its driver.

Think about that for a moment. We may have already crossed the line from the car assisting the driver to the driver assisting the car.

doubt it.

[jlv]

the NTSB concluded that the way the Tesla Autopilot system monitored and responded to the driver's interaction was not an effective method of ensuring driver engagement.22 As a result, the NTSB recommended that Tesla and five other manufacturers of vehicles equipped with SAE Level 2 driving automation systems take the
following action:

H-17-42

Develop applications to more effectively sense the driver's level of engagement and alert the driver when engagement is lacking while automated vehicle control systems
are in use.

And in response, Tesla changed the way AP monitors the driver and required more active feedback that the driver was engaged with the car. If you take your hand off the steering wheel (don't put torque on it or enable with any buttons) for a few seconds, it will tell you to take control. This is changed from a prior period of minutes.

You write as if AP was a static thing and the version that was shipped in the car is the version that is still in use (like what other automakers do). That's just not so.

I just needed to make an emergency roundtrip to CT yesterday; 100 miles each way, 90 of it highway. Each way I was on AutoPilot for about 95 miles, including while it took highway interchanges between I90/I84 and then I84/I91. I easily got over a dozen alerts to "touch the steering wheel" even though I had my hands on it the whole time (and was paying attention).

Reading posts from people who have no clue what it is actually like to use AutoPilot is like the joke about blind men describing an elephant by feeling only one part of it. This topic title is "Tesla's autopilot, on the road", but there are so many strong opinions without knowing what it is like "on the road".

[LeftieBiker]

I posted what I did because, to the best of my recollection, in all of the news stories I've seen about Teslas being involved in AP-related crashes, not one of them mentioned a defeat device being used. If the newest version requires driver attention in an effective way, that's great news. I still think that these systems should be tested in the way that Mercedes tests them, not the way that Tesla does. Americans still have a bit of a Wild West mentality about safety...

[SageBrush]

I easily got over a dozen alerts to "touch the steering wheel" even though I had my hands on it the whole time (and was paying attention).

This is one of things I most disliked about the Tesla AP -- being constantly nagged even though my hands were always on the wheel in the same way I drive off AP. I would have to move the car off its track before the AP was happy with me. Hopefully that has improved a lot.

Reading posts from people who have no clue what it is actually like to use AutoPilot is like the joke about blind men describing an elephant by feeling only one part of it. This topic title is "Tesla's autopilot, on the road", but there are so many strong opinions without knowing what it is like "on the road".

This is MNL. Expect posts about Tesla AP to be at about the same level as posts about the Tesla stock price. As in, clueless, or worse.

[GRA]

the NTSB concluded that the way the Tesla Autopilot system monitored and responded to the driver's interaction was not an effective method of ensuring driver engagement.22 As a result, the NTSB recommended that Tesla and five other manufacturers of vehicles equipped with SAE Level 2 driving automation systems take the
following action:

H-17-42

Develop applications to more effectively sense the driver's level of engagement and alert the driver when engagement is lacking while automated vehicle control systems
are in use.

And in response, Tesla changed the way AP monitors the driver and required more active feedback that the driver was engaged with the car. If you take your hand off the steering wheel (don't put torque on it or enable with any buttons) for a few seconds, it will tell you to take control. This is changed from a prior period of minutes.

You write as if AP was a static thing and the version that was shipped in the car is the version that is still in use (like what other automakers do). That's just not so.

Yet Tesla's method remains less effective than also having a driver-monitoring camera, because it's entirely possible to keep a hand on the wheel while not looking at the road. For that matter, it's possible to sleep while keeping a hand on the wheel.

These days the drivers who most frequently try to kill me while I'm walking or cycling are driving with one hand on the wheel. The other is in their lap holding or typing on their phone, and their eyes are directed at their laps 50-90% of the time. And this is for cars where the driver is totally responsible for driving the car.

Now give drivers a system that tells them it can handle the driving most of the time, but they must keep watching the road and paying attention so that they can resume control in an instant, yet which doesn't monitor that they're doing so. What could possibly go wrong with such a system, especially when that system is known to be incapable of handling some commonly-encountered situations? Re the Delray Beach crash:

In the Delray Beach crash, the driver turned on the car’s adaptive cruise control system, which keeps it a set distance from vehicles ahead of it, 12.3 seconds before impact, the NTSB found. Autosteer, which keeps the car centered in its lane, was turned on 2.4 seconds later. No pressure was detected on the steering wheel in the 7.7 seconds before the crash, the report said.

Tesla told the NTSB that the driver wasn’t warned about not having his hands on the wheel “because the approximate 8-second duration was too short to trigger a warning under the circumstances,” the report said.

The NTSB’s report said Autopilot wasn’t designed to work in areas with cross traffic, yet Tesla allows drivers to use it under those circumstances. Tesla told the NTSB that forward collision warning and automatic emergency braking systems on the Model 3 in the Delray Beach crash weren’t designed to activate for crossing traffic or to prevent crashes at high speeds. . . .

“The Delray Beach investigation marks the third fatal vehicle crash we have investigated where a driver’s over-reliance on Tesla’s Autopilot and the operational design of Tesla’s Autopilot have led to tragic consequences,” NTSB Chairman Robert Sumwalt said in a statement.

https://amp-insurancejournal-com.cdn.am ... 562009.htm

I just needed to make an emergency roundtrip to CT yesterday; 100 miles each way, 90 of it highway. Each way I was on AutoPilot for about 95 miles, including while it took highway interchanges between I90/I84 and then I84/I91. I easily got over a dozen alerts to "touch the steering wheel" even though I had my hands on it the whole time (and was paying attention).

Reading posts from people who have no clue what it is actually like to use AutoPilot is like the joke about blind men describing an elephant by feeling only one part of it. This topic title is "Tesla's autopilot, on the road", but there are so many strong opinions without knowing what it is like "on the road".

Again, are the NTSB, CR and others, who've tested the systems side by side, clueless? How about the German court?

Tesla Autopilot and Full Self-Driving claims are judged ‘misleading’ by German court

https://electrek.co/2020/07/14/tesla-au ... man-court/

[cwerdna]

Someone posted this at work. I haven't had time to watch all of it yet but they had a drone following.
https://www.youtube.com/watch?v=iKlpCG367AE

At about 1:25, they almost rear-end a parked car while turning left. Near the end, it gets confused at a left turn and stops. There are other near misses and issues, not surprisingly.

He has some passengers along speaking Russian.

[Andy11]

Pro Pilot forces the driver to keep paying attention, and keep hands on the wheel. Since people are sleeping with Tesla's AP on, I don't think it behaves the same way.

People who sleep with AP are using a defeat device. otherwise they would've been forced to pay attention as well.

Sometimes AP works better than driver himself. It has better reaction for sure, we would have less car incidents if we have only AP cars on the road