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

Quick question

This site may earn commission on affiliate links.
I know there've been at least two or three videos out there that purport to show a Tesla swerving to avoid rear-ending someone after having been rear-ended itself. They seemed kind of sketchy in their provenance though, and I don't have them to hand to show.
 
  • Helpful
Reactions: jebinc
I know there've been at least two or three videos out there that purport to show a Tesla swerving to avoid rear-ending someone after having been rear-ended itself. They seemed kind of sketchy in their provenance though, and I don't have them to hand to show.

Its not going to swerve out of its lane. Nor accelerate.
But it will brake if a vehicle intrudes sufficiently into your lane. I've had it do that when a truck started to wander into my lane when using NoA.

Changing lanes to avoid an accident, or even swerving, gets you into a classic AI ethical conundrum. I'm sure Tesla don't want to have to solve that one yet.
The conundrum is whether the AI should act to minimize risk to the car and its occupants or compute in the risk to whatever it might hit when it swerves. Even assuming the AI has enough data to quantify the risks.
For example: Would you, as a driver, swerve to avoid hitting a car in the rear if the alternative was to plough into a group of kids crossing the street?
 
Last edited:
Hence my question. I wonder if anyone knows... seems like it will need to if a robotaxi future is real and not just more Musk hyperbole.
If the car is acting as a robotaxi it will be full time FSD/AP.
Its not going to swerve out of its lane. Nor accelerate.
But it will brake if a vehicle intrudes sufficiently into your lane. I've had it do that when a truck started to wander into my lane when using NoA.

Changing lanes to avoid an accident, or even swerving, gets you into a classic AI ethical conundrum. I'm sure Tesla don't want to have to solve that one yet.
The conundrum is whether the AI should act to minimize risk to the car and its occupants or compute in the risk to whatever it might hit when it swerves. Even assuming the AI has enough data to quantify the risks.
It will swerve if it's on autopilot though.
 
Last edited:
  • Helpful
Reactions: jebinc
Its not going to swerve out of its lane. Nor accelerate.
But it will brake if a vehicle intrudes sufficiently into your lane. I've had it do that when a truck started to wander into my lane when using NoA.

Changing lanes to avoid an accident, or even swerving, gets you into a classic AI ethical conundrum. I'm sure Tesla don't want to have to solve that one yet.
The conundrum is whether the AI should act to minimize risk to the car and its occupants or compute in the risk to whatever it might hit when it swerves. Even assuming the AI has enough data to quantify the risks.

I agree with you completely, which is why I went out of my way to say only that the videos I've seen claim to demonstrate that behavior, and indeed expressed my skepticism as to the accuracy of those claims. However, those videos are out there, and wanted to help provide context in case someone here happened to post one as a "wow look at what it can do".
 
  • Helpful
Reactions: jebinc
I'm not sure about swerving but mine definitely lane corrects when not on AP. I've had it misread lane markings a couple of times at the same intersction and cause the car to move over. It screams at me at the same time and is pretty alarming considering I'm driving along at an intersection I've probably drive in over 1500 times (best estimate). If there was a group of school kids there it would have been an interesting experiment!
I think Elon is talking his usual bluster about a fleet of robotaxies in that short a period of time, but given the speed of tech advances who knows?
 
  • Like
Reactions: jebinc
It will swerve if it's on autopilot though.
Under certain circumstances, it'll even swerve when it's *not* on autopilot!

I had no idea this was a thing, until it happened to me last week. I posted a thread about it, in fact.

So my wife and I are cruising along on a very nice four lane road. Lanes are well marked with fresh paint, no other traffic.

Suddenly get really horrible audible alarms, with the message, "Imminent Collision Detected: Corrective Steering Applied." I never did feel any actual steering input, though.
 
  • Like
Reactions: Nocturnal
Side Collision Avoidance should cause it to steer away from a side collision with or without without AP enabled I think. There have been a few videos of that in action. It seems to require the intruding car to have a lot of overlap with your Tesla. Plenty of videos of bad lane changers hitting or nearly hitting Teslas with a small rear to front overlap.
 
Under no current circumstances will the car accelerate on its own to avoid anything. From the side, the car will move. No one knows how aggressive that movement can be. From the front, the car will slow down. The way AEB - which is active at all times - is described by Tesla, it will not bring the car to a full stop but only decelerate. From the rear, the car takes no action and provides no indication of an impending rear impact.

When I was rear-ended, Tesla did ask me how the safety functions of the car performed. I was at a dead stop in the right-most lane with room enough in front and to the side that the car could have moved away, and I recommended that the car be programmed to detect a potential rear impact and take such action. Maybe some day, but not now.
 
I've had my Model 3 for a year now. Some time after I began using Auto Pilot, my car was driving on the highway (My hands were on and following the moves of the wheel) when a speeding pickup tried to merge into my lane, from the left, before he'd completely passed me. My Model 3 noticed the situation before I did and immediately moved into the lane to the right. The interesting thing is that my car then sped up and pulled in front of the pickup! Surprised the hell out of me. I was grateful for the the save, but speeding up and pulling in front of the aggressive pickup was not ideal. No use adding to the local stress level. But, I was so grateful for the save. This incident happened several OTA software/firmware updates ago. I'd been using AP for awhile and had learned to follow the moves of the steering wheel -- only taking control occasionally. I noticed the pickup's near impact with me _after_ my car took evasive action. I remember looking over my righthand shoulder as my car moved into the right lane and realized that the lane had been clear, but I was unaware that the righthand lane was a safe haven before being blindsided by the pickup's unpredictable behavior. My car was aware of the safe escape path when I wasn't. I didn't override the AP moves because they proved to be ahead of my in-the-moment awareness of a bad situation. I've no idea what emergency moves that AP has now, and nothing like that has happen since--thank my lucky stars.
 
  • Informative
Reactions: jebinc
Changing lanes to avoid an accident, or even swerving, gets you into a classic AI ethical conundrum.
Yes and no. The example you cite is clearly a toughie. But if the lane to swerve into is open, then there is no ethical decision. Every utilitarian philosopher would support a swerve into an open lane where there are no adverse consequences.

I recall someone posted a dashcam video a few months back of their 3 swerving from lane 2 into lane 3 as a pickup truck spun out in lane 1, encroaching on the car's path in lane 2. The poster said he could not have checked lane 3 fast enough himself to make the maneuver safely, but the car had ongoing situational awareness and knew it could swerve safely. The driver felt Tesla saved him in this situation.
 
  • Helpful
Reactions: jebinc