Separate names with a comma.
Discussion in 'Model S' started by richardl007, Oct 18, 2015.
Does anyone find auto lane change slightly to aggressive (turns to quick) ?
Several times I thought it was actually too slow. Some times it just doesn't change, with no indication as to why (no one is in the lane, and the sensor do not indicate it thinks there is anyone there either)
I'm not talking about about how long it takes but rather the smoothness. I find it turns to hard.
I think it's just right.
I think it is a good compromise between a very aggressive lane change and an overly tentative lane change. I can imagine that Tesla spent quite a while tuning that algorithm, but of course it won't satisfy everyone. I expect that in the future there will be user selectable options: "Leisurely", "Moderate", "Aggressive".
I think they did a pretty good job of putting it right in the middle of what most people do.
Some people are more aggressive, and some people are less aggressive.
Maybe they could key off of how the turn signal is used -- holding but not fully engaging the stalk means change lanes now. Fully engaging means "whenever you get around to it".
While I do find the lane changing a bit more aggressive than I'd like I wish it would start moving a bit quicker when stopped behind another vehicle at a red light and the light turns green.
I get this a lot. More when changing into the right lane than changing into the left lane. No discernable logic or reason but often it will not auto lane change even though it is completely clear. Other times, it will lane change with a car in the adjacent lane only a half a length behind me.
We've had several cases where the car changed lanes while simultaneously issuing a warning to put hands on the wheel because the car was departing the lane! It's happened several times, but is not reproducible.
I'm not sure of this. However, I believe here in Ye Ole Republic of Massachusetts, there is a particular amount of time you are suppose to complete a lane change. May not be true, but, it's something I heard. To that, I believe there have been studies showing a quick(er) lane change is safer in the long run. FWIW
Click below to read a well written critique of the lane change feature safety - er.... lack thereof.
Tesla's Autopilot Lane Change: Likely To Be Unsafe At Any Speed - Tesla Motors (NASDAQ:TSLA) | Seeking Alpha
Gah! WARNING: people should not follow that link. He's a Tesla bear/short and he's deliberately confuting the lane change feature with autonomous lane changing.
Autopilot lane changing is simply a car-controlled maneuver initiated by the driver. It's an instruction to the car to move to the next lane instead of following the current lane. I think what some people like the OP might find is that auto lane change changes lanes at a consistent rate, while they would vary depending on conditions and traffic. But, like an automated gear change, the automated steering should give a more consistent and smoother entry and exit from the maneuver than would most drivers, which helps make changing lanes safer. It's a further reduction in physical effort that will reduce errors.
Those are all good points. But I think he makes a big leap when he implies that just because there are better solutions that the current Tesla solution is unsafe. I agree it's basically just blind spot coverage and that longer range coverage would be much better. But I've had blind spot warning in many of my cars for years and it's saved my butt many times from crashing into cars in my blind spot but I cannot think of even a single instance where the presence or absence of longer range 50 meter detection would have resulted in or prevented an accident. I understand he's saying that Tesla's marketing material will lead to people having a false sense of security or misunderstanding of how far the ultrasonics can actually see, but the exact same argument could be made about any blind spot system (manufacturers warnings, fine print, and all). Possible solutions:
1) tap into the rear view camera and use the front camera image detection alogorithms to detect cars approaching at high speed.
2) display the rear view camera in the dash screen (perhaps just the relevant portion of the rear view to bring attention to the danger zone) whenever you engage the turn signal during autopilot.
3) ultimately it seems autopilot 2.0 will have rear and side radar, but I'm glad that autopilot 1.0 can still do much or most of what autopilot 2.0 can do.
The lane change is pretty nice. Today for the first time it refused to change lanes, highway was straight and no other cars anywhere near our location. I see from other comments that this has happened to others as well. I don't see anything "unsafe" about how it works. Check that the lane is clear and use the blinker. Not any different from pre-autopilot times.
I don't particularly like the way the lane change UI works. I was lead to believe by the marketing hype that all that was required was a tap on the turn signal lever and it would initiate the turn. Of course, it's the driver's responsibility to make sure that the turn is safe at that point. I thought tapping the turn signal was supposed to mean approximately the same thing as actually turning the steering wheel. But no, I have to HOLD the turn signal for an extended period of time before it even starts to turn. I've always thought the way to make a lane change is to: check that it's OK, signal, wait for three clicks, then move over completing the move in about the same time as the three clicks. If you just tap the turn signal in manual mode it does the three clicks. In auto pilot mode it should do the three clicks then the turn WITHOUT having to hold the lever. If the driver wants to abort at any point he merely has to use the steering wheel.
I take it that you don't like to latch the turn signal for the auto lane change to complete the lane change, correct?
As much as I dislike SeekingAlpha trolls, this one has a very good point, and I've made the same point ever since Autopilot was announced. The lack of long-range rear radar makes even the driver-initiated autopilot lane change an accident waiting to happen.
Consider this: The front camera could clearly be programmed with the capability to recognize red lights and stop signs, but Tesla intentionally does not even try, leaving the responsibility 100% to the driver. There is no ambiguity. The autopilot will allow you to barrel straight through a red light without so much as a blip, and the driver knows this, so there will never be a false sense of security that the car will do it for you. But with the lane changing feature, the fact that it _usually_ recognizes danger and refuses to do the maneuver, naturally leads an ordinary driver into a very false sense of security that it will _always_ do so. When it comes to sharing mission-critical responsibility, _usually_ is a very dangerous thing.
TACC, on the other hand, is astonishingly reliable, because it has sufficient hardware to do its job. I'd be surprised if there's been even a single case where a Tesla on TACC has plowed into the car in front of it. It's not easy to construct a situation where it could even do so, short of another driver deliberately attempting to swerve into the lane and crash. But with the autopilot lane-changing, it's trivial to construct real-world scenarios that would lead to unavoidable collisions. I believe it's only a matter of time before this happens. It may still technically be the driver's "fault", but Tesla should be designing its car for imperfect human drivers, not perfect robots. They have to take human nature into account, and in this case it seems like they really haven't.
Of course, I hope that I'm completely wrong, and that years pass without any incidents like this occurring. But for now, I'll be very hesitant to quickly pass another Tesla on the left.
Sir: You are very correct. I only hope that accident due to auto lane change does not happen too soon, due to driver error and over reliance on the far from reliable system.
Right. When driving manually I typically use the single tap three click feature, that's what it's there for. Latching it down for "auto" lane change would mean having to do a second action to unlatch it. I would rather concentrate on the road. OK, it's not that hard. It just would be a lot easier the other way. What were they thinking?