Truck Lust is a real thing...

[TrafficEng]

Yesterday I did a 500 mile round trip from Atlanta to Alabama and back. 95% of the time on Autopilot (AP1) and it worked great. Most of the time while passing trucks, no problem. However, at least 4-5 times on the trip the AP would start to wobble, and veer from side to side (including towards the truck), to the point where I would have to take over. I tried really hard to figure out WHY this occurred, but no luck. Eventually, I just had to conclude that Truck Lust is a real thing!

In each case the lines were good; there were no vertical curves; just a random wobble in Autosteer while passing a truck. On a couple of occasions while it was occurring, I could see the dash display. I could see the blue lane lines on the dash wobbling, and going blue/grey/blue to indicate problems in the lane detection.So it was definitely a problem with the camera detecting the lines, even though the lines were clear.

I have about 10,000 miles on AP1, and so I'm pretty experienced in the limitations of AP1. In all other cases (blind rise, approaching stopped vehicles on a curve, etc) the behavior of AP at the edge is predictable. But not so with Truck Lust! Just be careful when passing all trucks (and hold the wheel a little tighter) because every so often it will give you a fright and do something crazy.

Is it possible it is a combination of shadows confusing the lane detection system? Any thoughts on the cause?

 

[artsci]

I don't have AP but thanks for the post. Fascinating. I don't think I'm quailified to offer any conjecture, but could changed aerodynamic effects of passing a large truck have an effect?

 

[TrafficEng]

I don't have AP but thanks for the post. Fascinating. I don't think I'm quailified to offer any conjecture, but could changed aerodynamic effects of passing a large truck have an effect?

I was with 3 other engineers in the car during this drive, and so we were all trying to observe/figure out what caused it. We did speculate on air buffer/cross winds having an effect. However, on this drive there was no wind (the trees were still) and we couldn't think how the aerodynamic effects could result in what we experienced. But yes, I guess possible, but not understood.

 

[Todd Burch]

I've got around 20,000 miles on AP and never once experienced "truck lust".

Some older versions of AP used to do a strong lane bias, but that's something different. I do find that sometimes trucks feel closer because humans naturally bias away from trucks when passing them more than AP does, but oddly I've never actually seen a tendency for the car to physically head closer to a truck.

 

[wdolson]

On my trip to California in October I didn't notice the car having any affinity for trucks, but it wanted to swerve into bridge railings about 1/3 of the time when I approached a bridge while in the left lane. It would start to approach the rail, then decide not to and go back to driving down the lane. A few times it lost the lane and was swerving around like a drunk.

This only happened in California and not in Oregon. It hasn't happened in Washington either.

 

[Vitold]

I suspect it is ultrasonics overriding camera input at times where parts of the truck become suddenly visible.

HW2 has upgraded ultrasonic sensors and side cameras which should help in such situations.

 

[Ormond]

It might be the intoxicating aroma of diesel fuel or long-distance envy.

 

[TrafficEng]

On my trip to California in October I didn't notice the car having any affinity for trucks, but it wanted to swerve into bridge railings about 1/3 of the time when I approached a bridge while in the left lane. It would start to approach the rail, then decide not to and go back to driving down the lane. A few times it lost the lane and was swerving around like a drunk.

This only happened in California and not in Oregon. It hasn't happened in Washington either.

My suspicion is a longitudinal shadow created by an 18 wheeler sometimes 'looks' like a line (hence the wobble). Perhaps shadows created by the bridge railings had a similar effect? Then other times, when the sun is in a different direction then no effect? Just speculating that your observation may fit this hypothesis?

 

[Rockster]

I have thousands of miles on AP1 and I concur that truck lust is real. I have no idea why it occurs, I've just learned to watch for it.

 

[kort677]

I've never experienced truck lust but I have had a couple of odd things happen with ghosts.
IMHO there are two take aways from these reports.
the first is that you must always be alert and able to assume control immediately
and second, the AP system is still only a BETA. the system is far from being 100% perfect.
be cautious when using the system.

 

[BerTX]

The thing that makes sense to me is that the programmers tried to move the car slightly away from the truck, as we all do when driving. As a vehicle passes, the camera and sensors start to lose the vehicle beside it. When the vehicle is lost, the car moves back to the center of the lane, and when it does, we perceive that as swerving into the truck.

But I don't have the thousands of miles of observation of others, so I'll have to do more study.

 

[EinSV]

Like other posters above, I have never experienced truck lust (I am on AP 1, v.8.0). There are rare occasions where I take over from AP because the vehicle in the next lane is too close to the lane line or the lane width is narrow and normal variation feels too close for comfort. I still haven't seen video footage that clearly shows "truck lust" as distinct from normal lane movement.

 

[Joe F]

Tesla has the logs, and is surely aware of the issue. Can't imagine a careful analysis of the data hasn't been done by now. Sounds as though it happens too often to be considered a corner case.

While waiting for AP2 S/W activation, I'm wondering if truck lust occurs at night or in overcast conditions? Would help rule out high contrast considerations.

 

[NikeWings]

No truck lust here either but there is an absolute tendency (for me anyway) to drive left off center, and AP off course, corrects for this. I proved it to myself with the dash cam recoding my AP and non AP tracks or otherwise, I too would have been in the veering (not lusting) for trucks camp. That said, I remain a wheel gripper for sure while passing trucks.

 

[bonaire]

Tesla has a solution ... the AP 2.0. Have you considered a trade-in? More cameras and other new features will be coming.

 

[Electroman]

Although I have not experienced truck lust in my 15K+ of AP miles, I kind of concur with TrafficEng that it is a case of unable to pricesly read the lane position in between the two vehicles.

When it has lane marking on one side (let us say left side) and loses markings on the right side as in a lane drop, it always tends to moves a few inches to the right hunting for the lane on the right.

I am guessing that is what is happening in a Truck without a side skirt

 

[kort677]

Tesla has a solution ... the AP 2.0. Have you considered a trade-in? More cameras and other new features will be coming.

for me it is financially wiser to keep the car I own and drive it intelligently with vigilance and within the parameters of the car's AP abilities. AP2 IMHO is not ready for prime time and at this point not worth upgrading to.

 

[⚡️ELECTROMAN⚡️]

Yesterday I did a 500 mile round trip from Atlanta to Alabama and back. 95% of the time on Autopilot (AP1) and it worked great. Most of the time while passing trucks, no problem. However, at least 4-5 times on the trip the AP would start to wobble, and veer from side to side (including towards the truck), to the point where I would have to take over. I tried really hard to figure out WHY this occurred, but no luck. Eventually, I just had to conclude that Truck Lust is a real thing!

In each case the lines were good; there were no vertical curves; just a random wobble in Autosteer while passing a truck. On a couple of occasions while it was occurring, I could see the dash display. I could see the blue lane lines on the dash wobbling, and going blue/grey/blue to indicate problems in the lane detection.So it was definitely a problem with the camera detecting the lines, even though the lines were clear.

I have about 10,000 miles on AP1, and so I'm pretty experienced in the limitations of AP1. In all other cases (blind rise, approaching stopped vehicles on a curve, etc) the behavior of AP at the edge is predictable. But not so with Truck Lust! Just be careful when passing all trucks (and hold the wheel a little tighter) because every so often it will give you a fright and do something crazy.

Is it possible it is a combination of shadows confusing the lane detection system? Any thoughts on the cause?

The most obvious question to me is, what firmware version does your car have installed? I've seen a couple of people reporting that truck lust might be a thing of the past with this latest update: 8.0.2.52.22 Firmware 8.0

 

[TrafficEng]

The most obvious question to me is, what firmware version does your car have installed? I've seen a couple of people reporting that truck lust might be a thing of the past with this latest update: 8.0.2.52.22 Firmware 8.0

My software is 8.0.2.50.114. I'll see if the 'truck lust' improves with the update, when I get it (hopefully soon!)

Tesla has the logs, and is surely aware of the issue. Can't imagine a careful analysis of the data hasn't been done by now. Sounds as though it happens too often to be considered a corner case.

While waiting for AP2 S/W activation, I'm wondering if truck lust occurs at night or in overcast conditions? Would help rule out high contrast considerations.

The truck lust experiences from Thursday were during good weather, during the day, with sunny/overcast conditions.

 

[Todd Burch]

Surely someone experiencing truck lust has a dashcam? Can someone post a video?