Hi everyone,
Long time reader, first time poster. I was surprised to see that no one else was talking about this major autopilot bug that just hit me on my vacation. If the same happens to others I hope my post can help relieve you of some stress and get your car back to normal quickly. Autopilot was rendered useless for a couple hundred miles of my trip after stopping for a supercharge. It had been working flawlessly for the first 150 miles of the trip. After super charging, the car would constantly want to turn left and then would jerk back to the center of the lane when it realized that it was not centered. I tried repeatedly to engage autopilot for the remainder of the trip because I had been looking forward to not having to drive for the entire trip and was dreading having to finish the tour up and down California having to do my own steering. The cruise control part worked fine so that was better than nothing but once you are used to autopilot there is no going back!
I was pretty sure that there was a hardware bug that was going to require a trip to the service center, but it didn't make sense because the visualizations on the screen were all still accurate. The road was displayed correctly and it even showed my car riding and sometimes even crossing the divider. I thought maybe it was wind (even though wind had never affected me before), but even if the wind came from the left it would still try to pull left. Wheel alignment since I just got new tires? No. Car drove true when driving manually.
Later that day when supercharging again, I searched the internet for similar issues. Someone said that the service center had told them that their radar was faulty but then wound up replacing the whole computer. Darn, I thought. Let me at least try rebooting the screen. Nothing. Later I tried shutting down the whole car, but I didn't wait long enough because that would have worked. The next morning, after the car really had shut down, everything magically worked. Drove through Yosemite almost 100% on autopilot. Supercharged again on the other side of yosemite, and the car decided to take the oncoming lane on one of the switchbacks. Problem returned. This time I got video because I knew people would claim that I wasn't using autopilot correctly. I started to suspect that this was a supercharging problem. Confirmed it today with the third instance of it glitching after supercharging. In all cases a full shutdown (safety/security shutdown) and waiting for at least 3 minutes fixes the problem. Still not 100% sure, though, because the autopilot still seems a little jerky unless the car has been off for the whole night. So this still could be a problem specific to my car that only occurs with the heat generated from supercharging.
Why does supercharging reveal the problem? My best guess is that there is a PID loop (feedback loop) that is used to control the steering wheel motors and that loop is still running when the car is idle and the integral summation is still accumulating the error to such an amount that it can't get back to normal until it is reset to zero with a reboot. Why the loop would be running when the car is stopped, I don't know. But that's why it's a bug, right! Anyway, that's the feel that it has based on my experience with wrongly tuned PID loops.
Here's the video that shows the problem. Sorry that it's short but I didn't want to die. You can see me fighting the steering wheel until finally autopilot disengages because I fight it too much. You can also see me hugging the center line and the animation clearly showing me doing that. There was a time on the 395 when there were no cars around that I let it move fully into the left lane before it finally decide to correct back.
Oh, one other oddity is that when this happens, the problem is less pronounced at lower speeds. I could trick myself into thinking that the problem went away when I turned on autopilot in a small town that had a speed limit of 20 or 30 mph. I suppose this makes sense because the effect of a small turning error has less impact and so there is less need to correct for it.
Bottom line, if this happens to you, turn off the car for more then 3 minutes and don't despair. Also, this is a good reminder for all of us to keep your hands on the wheel and don't get too comfortable letting the car drive. It's amazing technology but it's still beta.
Here's the video: https://photos.app.goo.gl/wTHn7SEYDrKShamZA
Neil
Long time reader, first time poster. I was surprised to see that no one else was talking about this major autopilot bug that just hit me on my vacation. If the same happens to others I hope my post can help relieve you of some stress and get your car back to normal quickly. Autopilot was rendered useless for a couple hundred miles of my trip after stopping for a supercharge. It had been working flawlessly for the first 150 miles of the trip. After super charging, the car would constantly want to turn left and then would jerk back to the center of the lane when it realized that it was not centered. I tried repeatedly to engage autopilot for the remainder of the trip because I had been looking forward to not having to drive for the entire trip and was dreading having to finish the tour up and down California having to do my own steering. The cruise control part worked fine so that was better than nothing but once you are used to autopilot there is no going back!
I was pretty sure that there was a hardware bug that was going to require a trip to the service center, but it didn't make sense because the visualizations on the screen were all still accurate. The road was displayed correctly and it even showed my car riding and sometimes even crossing the divider. I thought maybe it was wind (even though wind had never affected me before), but even if the wind came from the left it would still try to pull left. Wheel alignment since I just got new tires? No. Car drove true when driving manually.
Later that day when supercharging again, I searched the internet for similar issues. Someone said that the service center had told them that their radar was faulty but then wound up replacing the whole computer. Darn, I thought. Let me at least try rebooting the screen. Nothing. Later I tried shutting down the whole car, but I didn't wait long enough because that would have worked. The next morning, after the car really had shut down, everything magically worked. Drove through Yosemite almost 100% on autopilot. Supercharged again on the other side of yosemite, and the car decided to take the oncoming lane on one of the switchbacks. Problem returned. This time I got video because I knew people would claim that I wasn't using autopilot correctly. I started to suspect that this was a supercharging problem. Confirmed it today with the third instance of it glitching after supercharging. In all cases a full shutdown (safety/security shutdown) and waiting for at least 3 minutes fixes the problem. Still not 100% sure, though, because the autopilot still seems a little jerky unless the car has been off for the whole night. So this still could be a problem specific to my car that only occurs with the heat generated from supercharging.
Why does supercharging reveal the problem? My best guess is that there is a PID loop (feedback loop) that is used to control the steering wheel motors and that loop is still running when the car is idle and the integral summation is still accumulating the error to such an amount that it can't get back to normal until it is reset to zero with a reboot. Why the loop would be running when the car is stopped, I don't know. But that's why it's a bug, right! Anyway, that's the feel that it has based on my experience with wrongly tuned PID loops.
Here's the video that shows the problem. Sorry that it's short but I didn't want to die. You can see me fighting the steering wheel until finally autopilot disengages because I fight it too much. You can also see me hugging the center line and the animation clearly showing me doing that. There was a time on the 395 when there were no cars around that I let it move fully into the left lane before it finally decide to correct back.
Oh, one other oddity is that when this happens, the problem is less pronounced at lower speeds. I could trick myself into thinking that the problem went away when I turned on autopilot in a small town that had a speed limit of 20 or 30 mph. I suppose this makes sense because the effect of a small turning error has less impact and so there is less need to correct for it.
Bottom line, if this happens to you, turn off the car for more then 3 minutes and don't despair. Also, this is a good reminder for all of us to keep your hands on the wheel and don't get too comfortable letting the car drive. It's amazing technology but it's still beta.
Here's the video: https://photos.app.goo.gl/wTHn7SEYDrKShamZA
Neil
