-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
First hand report of autopilot driving in electrek
- Thread starter LetsGoFast
- Start date
Just reading the article now ... comments:
Why no heads-up display? If there are icons, blue lines and other VITALLY IMPORTANT indicators for driving on the driver assistance tools, then they ought to be integrated as best as possible directly where the driver must look to properly control the vehicle, which is out the front windshield, mainly.
The autosteer icon looks like a bull trying to bump me. They should make it look a lot more like a steering wheel than they did -- put some hands on it.
The autospeed indicator looks more like a steering wheel. Sigh.
The speed limit indicator is on the wrong side of the road. Speed limits are posted on the right, not the left. (This should be stuck on the left for only those countries where they drive on the left side of the road.)
I can't believe there are so many UI errors. Who does UI these days? Idiots?
Aha, this is a year ago: surely, they have non-idiots doing the UI who have fixed those errors.
Uhoh: they say this image was stolen from recent betas. IMO, if it works exactly as described with this SUCKY graphics UI, that's far better than it working crappy with a really excellent UI. Graphics is relatively cheap and easy to fix.
So far the rest sounds pretty darn good.
The reviewer had an AWFUL idea: minimum speed to activate. NO! I want it on stop and go freeway traffic, which is 99.9% of the reason I want this type of thing. I surely hope (1) it works well in that condition and (2) it isn't prevented from working in the idea that rural Interstate 280 at 4PM is a city street. If it works well in stop and go freeway traffic, then this becomes an enabler, allowing me to work and live places that normally I absolutely wouldn't consider.
There's certainly a lot of good news in that review, and a lot of good room for improvement.
Why no heads-up display? If there are icons, blue lines and other VITALLY IMPORTANT indicators for driving on the driver assistance tools, then they ought to be integrated as best as possible directly where the driver must look to properly control the vehicle, which is out the front windshield, mainly.
The autosteer icon looks like a bull trying to bump me. They should make it look a lot more like a steering wheel than they did -- put some hands on it.
The autospeed indicator looks more like a steering wheel. Sigh.
The speed limit indicator is on the wrong side of the road. Speed limits are posted on the right, not the left. (This should be stuck on the left for only those countries where they drive on the left side of the road.)
I can't believe there are so many UI errors. Who does UI these days? Idiots?
Aha, this is a year ago: surely, they have non-idiots doing the UI who have fixed those errors.
Uhoh: they say this image was stolen from recent betas. IMO, if it works exactly as described with this SUCKY graphics UI, that's far better than it working crappy with a really excellent UI. Graphics is relatively cheap and easy to fix.
So far the rest sounds pretty darn good.
The reviewer had an AWFUL idea: minimum speed to activate. NO! I want it on stop and go freeway traffic, which is 99.9% of the reason I want this type of thing. I surely hope (1) it works well in that condition and (2) it isn't prevented from working in the idea that rural Interstate 280 at 4PM is a city street. If it works well in stop and go freeway traffic, then this becomes an enabler, allowing me to work and live places that normally I absolutely wouldn't consider.
There's certainly a lot of good news in that review, and a lot of good room for improvement.
Last edited:
FredLambert
Member
A speed floor would have to be implemented with geofences. That way the Autosteer and TACC can be activated at any speed on highways, but disabled at low speed outside of highways, unless the road is known to be AP friendly.
Andyw2100
Well-Known Member
Pretty sure all that happened was that the collision avoidance warning sounded and TACC slowed your car down.
According to the release notes, the current implementation of the Collision Avoidance Assist system will only brake to reduce the impact of "an unavoidable frontal collision" that is going to happen. We don't yet have a true collision avoidance system.
Here's a link to a picture of those notes:
Firmware 6.2 - Page 2
I'm not pointing this out to give you a hard time. I just think it's important that people understand what our cars are and are not capable of, and credit the right systems for what's going on.
An interesting metric lacking from this article : number of alerts / hour. Because once the novelty wears thin, drivers might find those alerts frustrating if they happen too often.
What is clear is that computers are still "pretty dumb", so autopilot should modestly be called "dumb pilot" until CPUs with hardwired neural networks are used by Tesla. Now they are only being tested in research labs, so I am not blaming Tesla. Maybe I am blaming Elon for advertising something that has not been proven to work as well as the human brain: the Google team cheats by mapping every inch of the road in advance, at a far higher cost than Google Maps, and we yet have to see the Google cars perform well in unfavorable weather conditions : fog, rain, snow, ice, storm,...
What is clear is that computers are still "pretty dumb", so autopilot should modestly be called "dumb pilot" until CPUs with hardwired neural networks are used by Tesla. Now they are only being tested in research labs, so I am not blaming Tesla. Maybe I am blaming Elon for advertising something that has not been proven to work as well as the human brain: the Google team cheats by mapping every inch of the road in advance, at a far higher cost than Google Maps, and we yet have to see the Google cars perform well in unfavorable weather conditions : fog, rain, snow, ice, storm,...
FredLambert
Member
that metric would be useless. It's too dependent on road conditions. it could theoretically not send you any alert for hours, but start it in city streets and you get one every 20 seconds...
tstafford
Active Member
Glad you pointed that out - I didn't know of that distinction. In retrospect I should have taken control.
You probably should have. The system is designed to minimize the impact of an unavoidable collision, not stop it.
lolachampcar
Well-Known Member
I suspect the car can come to a complete stop but, liability wise, it would not be smart to say so as then you've obligated yourself to saving people from everything (which, of course, is not possible).
From the article, it is obvious there should be a different metric for day and night. To compare Tesla cars to other brands, owners should precisely compare autopilot features on the exact same road at the same time.
I believe this is exactly true. The car does attempt to come to a complete stop without a collision and probably can accomplish that under many circumstances, but they don't want to promise that so they have written the description the way they did. I know that there have been reports of cars coming to a full stop with a small amount of room to spare, but I certainly wouldn't want to test it out.
Andyw2100
Well-Known Member
I wouldn't necessarily say you should have taken over. As long as you were ready to if you thought the car was not going to be able to prevent a collision, and you still could prevent the collision, that was probably fine.
My reason for pointing out the distinction was that the more frequently even well-informed owners--people like you who are participating in the forums--misunderstand the nuances of these related features, the more likely it is that the misinformation and misunderstandings will spread. That could quickly become dangerous. The more people understand exactly what their cars' safety systems are doing, the better.
You're probably right. But unless I am mistaken, the system we're talking about will work even when TACC is not functioning, or when TACC has not been tracking a target car. By definition, if TACC has a target vehicle, it should be able to not rear-end that vehicle on its own, without involving the collision avoidance system. The way I understand it, the collision avoidance system is meant to operate over and above TACC--to help mitigate the impact of a frontal collision (or as you suggest, lolachampcar--possibly even eliminate it)--when there had not been any vehicle being tracked.
- - - Updated - - -
As I wrote above, you and Lolachampcar are probably correct.
But I question how many of the reports you are talking about were really collision avoidance system braking incidents, and how many were simply the collision avoidance system warning, with TACC braking.
Skotty
2014 S P85 | 2023 F-150L
This system seems realistic and think people will generally be happy with it if they aren't expecting too much. There may be a lot of scenarios where it will ask you to take over, and in some cases, it may be prudent to take over even when it's not asking you to. Even with just highway travel, there are a lot of scenarios to encounter and the auto-pilots ability to handle them will likely be limited.
Here's a few scenarios I personally encounter that I would carefully monitor the auto-pilot on:
1) New road surfaces with reflectors but no lines.
2) Exits and forks where the left and right lines do not stay parallel.
3) Road construction cones diverting traffic.
4) Road construction lane shifts with poorly done lines, or where the old lines were not well removed.
5) Stopped traffic in roadway (e.g. from road construction or accident) where you need to slow down gradually in advance to not only avoid running into something yourself, but to ensure you don't get rear ended by the doofus behind you.
6) Emergency vehicles on the side of the road, where you should either slow down or change lanes to provide them space.
7) Deciding how to handle road debris, including tractor trailer tires, boxes, carcasses, etc.
8) Slowing down when people or animals are near the roadway.
9) Being cautious around erractic drivers and avoiding driving in blind spots.
10) Maintaining appropriate distance from wide loads and red flag loads where someone has junk poking several feet out of the back of their truck or car.
Those are just a few off the top of my head.
Here's a few scenarios I personally encounter that I would carefully monitor the auto-pilot on:
1) New road surfaces with reflectors but no lines.
2) Exits and forks where the left and right lines do not stay parallel.
3) Road construction cones diverting traffic.
4) Road construction lane shifts with poorly done lines, or where the old lines were not well removed.
5) Stopped traffic in roadway (e.g. from road construction or accident) where you need to slow down gradually in advance to not only avoid running into something yourself, but to ensure you don't get rear ended by the doofus behind you.
6) Emergency vehicles on the side of the road, where you should either slow down or change lanes to provide them space.
7) Deciding how to handle road debris, including tractor trailer tires, boxes, carcasses, etc.
8) Slowing down when people or animals are near the roadway.
9) Being cautious around erractic drivers and avoiding driving in blind spots.
10) Maintaining appropriate distance from wide loads and red flag loads where someone has junk poking several feet out of the back of their truck or car.
Those are just a few off the top of my head.
The main problem lies with the average driver according to this article: "Automakers, Google take different roads to automated cars"
http://in.reuters.com/article/2015/09/04/autos-selfdriving-gurus-idINKCN0R40CZ20150904
"Alerting a driver to retake control during an emergency is one of the biggest safety challenges for manufacturers of partially automated cars, industry officials and scientists said.
Depending on the level of automation and intensity of alert, some drivers took an average of 17 seconds to respond to a takeover request and regain control of the vehicle, in a study just released by the National Highway Traffic Safety Administration and supported by Google and several leading automakers and suppliers. In that time, a car traveling at 60 miles per hour would travel more than a quarter of a mile.
Time to respond and regain control was reduced to just a few seconds when visual and audible warnings were accompanied by non-visual alerts such as a nudge from a mechanism in the seat.
But "there were alerts that were missed" by some study participants, NHTSA said. When drivers shift their attention to other tasks in a self-driving vehicle, such as sending an e-mail, "their readiness to respond to driving-related prompts and alerts can be delayed."
http://in.reuters.com/article/2015/09/04/autos-selfdriving-gurus-idINKCN0R40CZ20150904
"Alerting a driver to retake control during an emergency is one of the biggest safety challenges for manufacturers of partially automated cars, industry officials and scientists said.
Depending on the level of automation and intensity of alert, some drivers took an average of 17 seconds to respond to a takeover request and regain control of the vehicle, in a study just released by the National Highway Traffic Safety Administration and supported by Google and several leading automakers and suppliers. In that time, a car traveling at 60 miles per hour would travel more than a quarter of a mile.
Time to respond and regain control was reduced to just a few seconds when visual and audible warnings were accompanied by non-visual alerts such as a nudge from a mechanism in the seat.
But "there were alerts that were missed" by some study participants, NHTSA said. When drivers shift their attention to other tasks in a self-driving vehicle, such as sending an e-mail, "their readiness to respond to driving-related prompts and alerts can be delayed."
Petra
Member
This is a problem in the aviation industry as well but, from a regulatory standpoint right now, getting a fully autonomous system certified by the FAA is basically impossible even if it is better/safer (many airline crashes have been due to pilots not understanding the information the automated systems were giving them, not reacting properly when systems handed over control to them, or a combination thereof). Ultimately, the solution for both is taking the elevator approach to automation.
There was actually a decent episode of NPR's Planet Money which touched on autonomous cars and the above mentioned elevator approach: Episode 642: The Big Red Button : Planet Money : NPR
sillydriver
Member
Convincing drivers to think of this as a 'cruise control with steering' rather than an auto-pilot or self-driving system may be the best way for Tesla to convey how it is appropriately used.
And that is my biggest fear about the autopilot system: that there will inevitably be idiots who take it as license to stop paying attention to the vehicle hurtling down a freeway or, even worse, a surface street and start emailing or texting. In the hands of responsible drivers this system sounds like a huge boon, but all it takes is a few morons to ruin it for everyone else.
The human brain is wired to save energy. Once it has read "autopilot", it will use this feature to save as much energy as possible, hence the danger. The lower the IQ, the more the brain will resent expanding energy if saving it is an option. Well, it should also be combined with psychoanalysis : a brain fixated at the oral phase will tend to be lazier, while a brain fixated at the anal phase may be more active.
