Navigate on Autopilot is Going to Cause Accidents

[GSP]

I do 8 and 4 mostly and sometimes 10 and 2, but i've learned that 10 and 2 is no longer suggested/taught due to air bags. All I know is that my normal grip i've used for over 20 years isn't sufficient to avoid EAP nags most of the time.

If weight of both hands balances, then there is no torque on the wheel.

I shift the weight from one hand to the other every minute or so. I also use only one hand, at 3 or 4, then the other at 8 or 9. This is comfortable for driving 600 miles per day and rarely results in a nag.

GSP

 

[Dan Detweiler]

I think people are confusing when it's recommending a lane change when their is obvious cars in the way. I'm fairly certain if you confirm it will wait until there is a gap to change lanes.

While I haven't 100% tried it - I did wait until a car in my way was out of my blind spot before confirming and the AP still showed a red line for a bit before it merged over. The fact the car auto-slowed to allow some traffic coming onto the highway to safely merge was wild as well.

Yes, this is what happens but the problem is that rather than maintain current speed for the traffic to clear the car will decelerate DRASTICALLY to let the traffic pass. Doesn't exactly put you on the Christmas card list of the people that might be behind you. It needs to learn that there are circumstances where accelerating into a lane change is the safest most prudent way to handle the situation. I have no doubt this will come in time. Right now though...it's a pain for sure.

Dan

 

[chracatoa]

I am cautiously optimistic the software will increase in quality over time. Just be patient and PAY ATTENTION !!!!!!!!!!

This has been said for years, in every Tesla forum. Should Tesla's slogan be "don't worry, it'd get better" just like Pepsi should be "is Pepsi okay?"

 

[CoolSilver]

I've noticed a lot of what has been mentioned here. NAP hates the on and off ramps here. EAP alone does too with ping pinging.

NAP will aggressively oversteer changing onto an exit ramp where there was no prior lane. I blame this on the lack of dashed line marker where the lane appears. Each time it follows the white line so when it makes the sharp cut it tries to stay locked to it. It crosses over and if there is a guard rail might clip it if not paying attention.

Also when a lane is added to make it 2 to 3 lanes with the same type of quick addition and a split it gets very confused. It knows it needs to stay left at split it will suggest a lane change for the split. You can take the center lane for left or right. If you are early on responding it will move to left lane as expected into speeding traffic. If you try as the lane appears it is a 50/50 on moving left or fails because the right lane appears. If you let it start start to move to right lane to maintain as it thinks is the correct lane it will shift you two lanes and assumes the lane is clear. It handles the center lane fine at the split. I don't trust NAP to lane change near ramps or new lanes.

To further add to NAP braking like a panicked 16 year old. Every time it comes up on an exit, same line scheme, while in the right lane but not exiting it will hit the brakes. Usually happens when it wants you to move to left. It also randomly does this with no cars around and going at proper speed. If it suggests a lane change and a car hasn't passed in a few minutes it will hit the brakes. My assumption in both cases is that it is checking the blindspot to see if someone appears...but that's an asshat move and pisses me off driving with it as much as if someone behind me.

It is a little better on not taking the exit when it is enabled but inconsistent line markings from DOT make for a wavey ride or worse one that's not safe

 

[Lawsteve]

This has been said for years, in every Tesla forum. Should Tesla's slogan be "don't worry, it'd get better" just like Pepsi should be "is Pepsi okay?"

i guess some of us just have a little more patience than others. You were not forced to buy a Tesla, and are not forced to enable the features that Tesla clearly marks as beta and warns you to pay attention. You also have the ability to read what, as you correctly point out, has been said for years,in every Tesla forum.

Since I purchased my first Mod S P85 in Sept of 2013, there have been many ups and downs as each update hit the airwaves (as well as physical alterations to the car - anyone recall the “tank mode” modification?). The overall trend, in my opinion, has been an increase in performance and function. That trend continued after I upgraded to my Dec 2016 P100D.

I don't like the taste of Pepsi, so I do not buy it. If someone does not like the current state of the software, or build quality, for the car - don't buy it.

It’s all a personal decision.......

 

[voip-ninja]

i guess some of us just have a little more patience than others. You were not forced to buy a Tesla, and are not forced to enable the features that Tesla clearly marks as beta and warns you to pay attention. You also have the ability to read what, as you correctly point out, has been said for years,in every Tesla forum.

Since I purchased my first Mod S P85 in Sept of 2013, there have been many ups and downs as each update hit the airwaves (as well as physical alterations to the car - anyone recall the “tank mode” modification?). The overall trend, in my opinion, has been an increase in performance and function. That trend continued after I upgraded to my Dec 2016 P100D.

I don't like the taste of Pepsi, so I do not buy it. If someone does not like the current state of the software, or build quality, for the car - don't buy it.

It’s all a personal decision.......

As enthusiasts who read these forums we have a better idea of what to expect than the typical consumer. Tesla is very soon going to be delivering much lower versions of these cars to an increasing end-user base of casual customers who chose the Model 3 over a Camry or Accord for largely financial reasons as well as wanting the (arguably) highest tech car on the road.

The fact that Elon Musk said with a straight face a year or so ago that full self driving was probably only about a year or two out doesn't help the argument that people should not buy the product if they aren't ready to deal with these challenges.

Autopilot, after years of development still struggles with very basic driving tasks. Exiting and merging on the freeway are analogous to takeoff and landing in aircraft, in that they are the most dangerous maneuvers one performs in highway driving and I'm troubled with what I see from AP in this version. Absolutely people are going to get into accidents with it even if they are paying attention.... even if that accident consists of getting rear-ended by another motorist when AP decides it should jam on the brakes aggressively in order to make a lane change.

I'm not impressed with it... it has some utility but not worth it to me for the kind of driving I do and I kind of regret buying it. About the only thing it's "really realy" good at doing is sitting in stop and go traffic and freeing me from the monotony of piloting the car in packed traffic.... but that is something that even $2K and $3K traffic aware cruise control systems from other makers can do quite well.

As to this continued kool-aid drinking from people around here that thinks that the current sensor suite, software and GPU on this car will be capable of full self driving, even on highways in perfect conditions... well.....

 

[chracatoa]

I don't like the taste of Pepsi, so I do not buy it. If someone does not like the current state of the software, or build quality, for the car - don't buy it.

It’s all a personal decision.......

Agree. But I love the car as is (w/o EAP).

I may buy EAP to have summon or TACC for stop-and-go traffic. Still trying and deciding. I definitely don't think it's worth 5500, but if you think about any car options when you buy a new car, they're always expensive. They're most of the time hardware upgrades, but there's no reason why advanced software wouldn't be in the same category. I have two weeks to decide. My wife already told me that if I spend another 5500, it'll officially be my mid-life crisis.

I just hope that in 2 years we stop with this don't worry, it'll get better, and instead replace it with "it's awesome now, but you'll be blown away by the next version."

By the way, I must confess that sometimes I drink Pepsi when they don't have Coke. And maybe even kool-aid

 

[Dan Detweiler]

As enthusiasts who read these forums we have a better idea of what to expect than the typical consumer. Tesla is very soon going to be delivering much lower versions of these cars to an increasing end-user base of casual customers who chose the Model 3 over a Camry or Accord for largely financial reasons as well as wanting the (arguably) highest tech car on the road.

The fact that Elon Musk said with a straight face a year or so ago that full self driving was probably only about a year or two out doesn't help the argument that people should not buy the product if they aren't ready to deal with these challenges.

Autopilot, after years of development still struggles with very basic driving tasks. Exiting and merging on the freeway are analogous to takeoff and landing in aircraft, in that they are the most dangerous maneuvers one performs in highway driving and I'm troubled with what I see from AP in this version. Absolutely people are going to get into accidents with it even if they are paying attention.... even if that accident consists of getting rear-ended by another motorist when AP decides it should jam on the brakes aggressively in order to make a lane change.

I'm not impressed with it... it has some utility but not worth it to me for the kind of driving I do and I kind of regret buying it. About the only thing it's "really realy" good at doing is sitting in stop and go traffic and freeing me from the monotony of piloting the car in packed traffic.... but that is something that even $2K and $3K traffic aware cruise control systems from other makers can do quite well.

As to this continued kool-aid drinking from people around here that thinks that the current sensor suite, software and GPU on this car will be capable of full self driving, even on highways in perfect conditions... well.....

Don't like it...don't use it. Simple.

Signed...

Kool-Aid drinker.

 

[Jigglypuff]

This is all terrifying.

 

[confusatron]

I just completed a road trip from Seattle to Portland and back this weekend using Nav on Autopilot and thought I'd post about it here.

I'll start with saying that for the vast majority of the trip it worked great. It kept me at speed, in my lane, and intelligently suggested lane changes based on traffic speeds. While driving south on I-5, it also did a superb job of routing me onto I-205, just north of Vancouver, WA. It asked me to change lanes into the right lane, then took its self onto the exit and merged onto I-205 exactly as I would have done it, at the proper speed.

That's the good, now lets talk about the bad. The first was stop and go traffic while heading south in Tacoma. Initially it was fine. It was a bit more aggressive about speeding up as the car in front of me sped up than I would have been. I was talking to a fellow M3 owner about that today, he suggested I create a EAP driver profile with chill acceleration, which I think is a great suggestion. About mid-way through Tacoma, it started slamming the breaks. It felt like the anti-lock breaking was kicking in. There was no need to slam the breaks. Every time, there was plenty of clearance in front of the car for a gradual slow down. So that was the end of EAP through Tacoma.

The next problem came when I got into Oregon. I was on I-205 S and had to merge onto I-84 E. Nav got me into the correct lane and took the exit. But as it curved around to the I-84 onramp, the car slowed way down to about 30mph. Traffic around me wanted to do about 50mph. That's slow enough to be dangerous. So I had to disable it.

Heading back north the next day, Nav was mostly fine. It got me from I-84 to I-205, then on to I-5. The only problem I had on day 2 was a bad one. At one point, Nav suggested a lane change to get around slower traffic. I confirmed it. Traffic was very light at that point and the car had plenty of room to make the lane change. About 2/3rds of the way through the change, the car violently steered back into the original lane. There was absolutely no reason for this. There were 0 obstacles in the way. I can't figure out what the car must have thought it had seen. The only thing I can think of is at that point there was an onramp merging onto the freeway, but that was another lane away and there were no cars on it.

Those were all of the big problems we had. There were a few times when we were cruising down the road and for no reason the car decided it needed to shed several miles an hour. That was annoying, but it didn't drop the speed so drastically as to be dangerous.

My take away from this trip is that EAP worked pretty well for the most part, with some issues. But I don't think Nav on AP is ready for prime time. I'm not sure I'll use it again anytime soon. If I do, I'll have my hands at 10 and 2 and my foot on the pedals, ready to take over in an instant when Nav makes a maneuver.

 

[AnesDragon]

I am more alert using EAP especially the NAV on AP then I am just driving on my own. NAV on AP is a dicey experience so I have to be able to immediately take over. The only useful aspect of NAV on AP is that it may remind you when you need to exit and that is about it. I only continue to use it because I am trying to see how useful it is and to see what behavior I can modify to make it more useful, but that is about it. It is definitely safer for me to not use it at this point but, hey, I am always willing to be convinced as it improves over time.

 

[voip-ninja]

I have to wonder what the SCRUM meetings are like with the lead developers and engineers for the autopilot system at Tesla. I mean, they know this crap only works marginally and does dangerous things. They are comparing what "decisions" their software is making and comparing it to what actual human drivers do and see that they are woefully behind the 8 ball.

I think I better understand why so many people involved in AP program have actually quit. This task could prove to be insurmountable with current tech in these cars. I also understand, 100% (even if Elon doesn't) why most automakers are insisting on the use of LIDAR in cars that will operate in an autonomous or semi-autonomous mode.

//interesting aside...

I saw my primary care doc (who is also a friend of mine) for a routine check last week. He was asking me all about my newish Tesla. He mentioned that one of his patients is a guy who is recently retired but was doing work for Tesla not too long ago test driving cars. The guy used to be a race car driver and through a bunch of circumstances ended up being one of Tesla's drivers.

My doc buddy asked him if he ever met Elon. The guy said "yeah, I know him really well, I have him in my favorites list on my phone, want to see"?

My doc buddy mentioned that I had said some things about being unimpressed with AP performance and this guy apparently said "yeah, I wouldn't be using that".

Too funny.

 

[StellarRat]

I mean, they know this crap only works marginally and does dangerous things.

I'm going to disagree. It works way better than marginally if you use it according to the directions.

 

[voip-ninja]

I'm going to disagree. It works way better than marginally if you use it according to the directions.

Yeah sorry but having to intervene and slam on the brakes or take over steering in a split-second on multiple occasions to avoid it doing something stupid and potentially slamming into other traffic or creating a dangerous situation on the road means it is no better than marginal to me.

You and others can do whatever the hell you want, just remember that if AP causes a wreck or kills you, it's your own fault because it's BETA... at least according to Tesla and their lawyers.

 

[chracatoa]

I'm going to disagree. It works way better than marginally if you use it according to the directions.

You can forgive anything if the directions are be ready to take over at any moment because AP may put you in danger.
Even TACC (the best part of this technology) may be horribly wrong like in this video.

 

[StellarRat]

You can forgive anything if the directions are be ready to take over at any moment because AP may put you in danger.
Even TACC (the best part of this technology) may be horribly wrong like in this video.

I went about 120 miles on EAP on I-5 yesterday and I didn't have to intervene once. Also, you don't show all the videos where EAP prevented an accident. Let's be fair here.

 

[atonaltheory]

I don't trust Navigate On Autopilot. Be very careful, particularly around interchanges and exits. Be even more careful if you are in the HOV lane. HOV interchanges where two different HOV lanes head toward different highways have confused my car causing it to make unexpected and sometimes erratic maneuvers. Sometimes it works flawlessly, and other times it does not. Be careful. Don't get me wrong I love the car and the possibilities but Beta means Beta.