Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Firmware 7.0

This site may earn commission on affiliate links.
I don't have AP so I'm just asking generally. By the time the car swerved into the next lane and the chime went off (before it corrected itself), if another car was there to the right it sure seems like there would have been a collision? But if there were a car in the next lane, would have collision avoidance kept you in the proper lane despite its propensity to wander to the right?
 
I don't have AP so I'm just asking generally. By the time the car swerved into the next lane and the chime went off (before it corrected itself), if another car was there to the right it sure seems like there would have been a collision? But if there were a car in the next lane, would have collision avoidance kept you in the proper lane despite its propensity to wander to the right?

Very good question, I have no idea! And frankly would not be willing to test it myself given how it's been behaving at that specific location.
 
I don't have AP so I'm just asking generally. By the time the car swerved into the next lane and the chime went off (before it corrected itself), if another car was there to the right it sure seems like there would have been a collision? But if there were a car in the next lane, would have collision avoidance kept you in the proper lane despite its propensity to wander to the right?
This experience isn't quite the example you mention, but it is close. Also a disclaimer: I am not yet totally convinced this wasn't just a coincidental autopilot anomaly.

I was in the center lane of a three-lane interstate (I76 just west of Philadelphia) coming up on a left-exit to join I476S. I changed lanes from the center to the left lane just prior to the exit. I wasn't planning to exit, but I happened to be in a position where others were trying to get left to exit. As I was changing lanes, a car in the right lane cut across both the center and left lanes into the exit lane, right in front of me. However, he didn't fully vacate the left lane (where I was headed) immediately; he took his time moving over the last few feet. This happened just as I was between the two lanes. As soon as the car in front of me lingered partially in my lane, my car stopped going left and swerved back to the right. Once the car in front finally finished changing lanes, my car finished the lane change.

It certainly appears the autopilot detected a new obstruction in my lane and swerved to avoid it. The lane I had exited was still unoccupied, so there was no thread of a collision there.

I wonder what it would have done if there was someone in the center lane. I probably would have gotten the "I give up...take over immediately" message.
 
Marc, interesting video. I know that route and interchange very well. Clearly the car drifted to far to the right early in the curve, and actually got partway out of the lane, then overcorrected slightly. The fact that you were going 70 may have contributed to that behavior. While for a human driver who is paying attention that speed is not excessive, I think it may have been too fast for AP. Have you tried the same curve at 60?
Anyway, I engaged Autosteer and took the carpool lane as usual to see what would happen this time.
Even though it still spilled into the lane to the right of mine, it actually rejoined the carpool lane by itself without asking me to touch the steering wheel (although it did beep for a second without warnings on the IC).
 
Last edited by a moderator:
I don't have AP so I'm just asking generally. By the time the car swerved into the next lane and the chime went off (before it corrected itself), if another car was there to the right it sure seems like there would have been a collision? But if there were a car in the next lane, would have collision avoidance kept you in the proper lane despite its propensity to wander to the right?
I wondered the same thing while watching that video. I suspect that if a car had been to Marc's right as his car drifter to the right at the beginning of the curve while on AP, his ultrasonics would have sensed the adjacent car and corrected for it. I'm sure that is what they are designed to do.
Of course Marc had his hands on or very close to his steering wheel so he would have able to prevent a collision if that had been the situation. In fact if that had been the situation he might have taken control before the AP beeped because he would have perceived the car drifting to the right, and knowing that there was a car to the right he would have taken control right away.
 
I don't have AP so I'm just asking generally. By the time the car swerved into the next lane and the chime went off (before it corrected itself), if another car was there to the right it sure seems like there would have been a collision? But if there were a car in the next lane, would have collision avoidance kept you in the proper lane despite its propensity to wander to the right?

I would guess that the weighting on the collision avoidance is much greater than on lane keeping. Guess being the key word here. But if that's the case, it should theoretically override most of the other inputs.
 
Is it having trouble staying in the HOV (leftmost) lane because that
curve is just too much for the steering to follow at that speed?


Just a couple of questions:

Would it work better at 65 mph instead of 70?
With the video showing an improvement, do you feel that Autopilot is learning over time?

Thanks for sharing. Every video helps others understand the limits of Autopilot with the current software release.


Marc, interesting video. I know that route and interchange very well. Clearly the car drifted to far to the right early in the curve, and actually got partway out of the lane, then overcorrected slightly. The fact that you were going 70 may have contributed to that behavior. While for a human driver who is paying attention that speed is not excessive, I think it may have been too fast for AP. Have you tried the same curve at 60?


Ha, 3 similar questions about speed! You are all most likely correct that taking the curve at a lower speed (60 or even 65 mph) would not be an issue.

I haven't tried that yet, and I'll be at my company's office in San Jose all next week, but I'll give it a shot the following week. I'm guessing it will behave better when going slower.

At the same time, I thought it was interesting to see that AP has already improved for an identical situation (same location, same weather/lighting conditions, same speed). Skynet is indeed learning! :)
 
I would guess that the weighting on the collision avoidance is much greater than on lane keeping. Guess being the key word here. But if that's the case, it should theoretically override most of the other inputs.

I think we need to be careful with respect to the capabilities we believe the car has. I'm not sure it has any side collision avoidance in the way you are describing it.

The car now has, as of version 7, some sort of side collision warning, but there is no indication that it will take evasive action based on that.

We also know that if we have initiated auto lane change, the car will not change lanes into a lane if it senses another vehicle there.

But the situation above--where the car is, for whatever reason, veering into another lane while Auto Steer Beta was active--doesn't really cover either of the situations above. In other words, I am not convinced that in the situation MarcG encountered the Model S had any capability of avoiding a car in the lane it veered into. I could be mistaken.
 
I think we need to be careful with respect to the capabilities we believe the car has. I'm not sure it has any side collision avoidance in the way you are describing it.

The car now has, as of version 7, some sort of side collision warning, but there is no indication that it will take evasive action based on that.

We also know that if we have initiated auto lane change, the car will not change lanes into a lane if it senses another vehicle there.

But the situation above--where the car is, for whatever reason, veering into another lane while Auto Steer Beta was active--doesn't really cover either of the situations above. In other words, I am not convinced that in the situation MarcG encountered the Model S had any capability of avoiding a car in the lane it veered into. I could be mistaken.

Didn't Elon go on at length about the car preventing you from steering into objects to the side and barriers on the conference call?
 
I think we need to be careful with respect to the capabilities we believe the car has. I'm not sure it has any side collision avoidance in the way you are describing it.

The car now has, as of version 7, some sort of side collision warning, but there is no indication that it will take evasive action based on that.

I don't know, I think "Digital control of motors, brakes, and steering helps avoid collisions from the front and sides" (Your Autopilot has arrived | Tesla Motors) is at minimum, strongly suggestive of active side collision avoidance. I agree I'm not aware of any reports of it engaging in the wild. (Nor of it failing, though.)
 
Didn't Elon go on at length about the car preventing you from steering into objects to the side and barriers on the conference call?

I didn't listen to the call. (I couldn't deal with the audio quality of the one source I found.) If so, and if it is supposed to be active now, then perhaps the technology is there.

- - - Updated - - -

I don't know, I think "Digital control of motors, brakes, and steering helps avoid collisions from the front and sides" (Your Autopilot has arrived | Tesla Motors) is at minimum, strongly suggestive of active side collision avoidance. I agree I'm not aware of any reports of it engaging in the wild. (Nor of it failing, though.)

I wouldn't put too much stock in the blog post. That's just marketing. We know marketing over-sold the "Auto Lane Change" functionality in that write-up.

From the same post: "Autopilot allows Model S to steer within a lane, change lanes with the simple tap of a turn signal, and manage speed by using active, traffic-aware cruise control."

The blog post doesn't say anything about checking to make sure the lane is clear before tapping the turn signal, but that's critically important. (The release notes do make that clear.) So based on that, I just wouldn't read too much into anything in that blog post.
 
I wouldn't put too much stock in the blog post. That's just marketing.

Yeah, being marketing you have to parse it closely, but I can't think of any examples of them outright lying in their written communication. In any case, I provided that more as reinforcement of the comments Stonymonster mentioned, which of course weren't in writing.

It would be quite interesting to have further detail about what functionality is there, ideally with a demo. That's something only Tesla (or a really foolhardy owner) can provide, though. And in any case, of course it would be foolish to rely on the automatic systems as a replacement for good driving.

- - - Updated - - -

Next step.. We need some testers and test subjects. :)

See, just as I'm typing something about "a really foolhardy owner", someone else is too!