I don't think this is correct. Tesla claimed there was some sort of calibration needed when the parallel parking feature came out and I guess that was true but I never heard nor experienced that with the other functions of AP. People should be careful with their speculation because new people might think it's true.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
A flight instructor teaches Tesla Autopilot
- Thread starter Papafox
- Start date
I don't think this is correct. Tesla claimed there was some sort of calibration needed when the parallel parking feature came out and I guess that was true but I never heard nor experienced that with the other functions of AP. People should be careful with their speculation because new people might think it's true.
Boatguy
Active Member
FWIW, the service center did not push back on the request for calibration, they put it on the schedule.
Boatguy, you're in a unique position to do some safe experiments and share your results with us. When driving manually, how does your Tesla look positioned in the lane, according to the dashboard indications? Is it favoring the left or centered? When there is no traffic around you, how does your Tesla under AS look positioned in the lane, according to your eyes through the windshield and through the dashboard indications? Please conduct these experiments before you take your Tesla to service and then let us know what Tesla says about the car. We can all learn from your experience here.
FWIW, I just finished driving 700 miles on the highway this past week, and I am very impressed with the improvements I have seen in AutoSteer even since January. There was a winding road (not highway but divided) that AS would never handle before and it handles it well now. I let the autopilot steer 90% of the time (I chose not to use AP in situations where the blue lane markers were fading in and out, due to bad paint, water on highway, or glare. In many cases, I had reliable lane markers even with wet roads on highways with good paint and no glare.
Not once did I feel uncomfortable about the distance between my Tesla and the vehicle in the lane to my right. I never received a red obstacle cone from the front or rear sensors when near other traffic. If a yellow popped up, the Tesla tended to slowly work towards the other side of the lane a bit to give additional breathing room. Part of my new comfort was due to my getting used to seeing what the car looked like when centered in the lane (as opposed to favoring the left side, as I was used to with manual driving).
This said, please don't become overconfident with your autopilot driving. I would not trust AutoSteer in a situation where traffic is substantial and paint, water on highway, or glare caused the blue lane edge markers on the dashboard to be anything other than solid blue. The exception is with turnoffs when I am in the right lane. I do see the right lane marker disappear as expected for a few seconds at such times and the Tesla faithfully tracks a safe distance from the left lane marker until the right lane marker reappears.
Not once did I feel uncomfortable about the distance between my Tesla and the vehicle in the lane to my right. I never received a red obstacle cone from the front or rear sensors when near other traffic. If a yellow popped up, the Tesla tended to slowly work towards the other side of the lane a bit to give additional breathing room. Part of my new comfort was due to my getting used to seeing what the car looked like when centered in the lane (as opposed to favoring the left side, as I was used to with manual driving).
This said, please don't become overconfident with your autopilot driving. I would not trust AutoSteer in a situation where traffic is substantial and paint, water on highway, or glare caused the blue lane edge markers on the dashboard to be anything other than solid blue. The exception is with turnoffs when I am in the right lane. I do see the right lane marker disappear as expected for a few seconds at such times and the Tesla faithfully tracks a safe distance from the left lane marker until the right lane marker reappears.
I also tested this yesterday and you are absolutely right. My bias is the the left but I think it's more than me being seated in the drivers seat. It also feels safer to have the extra buffer to my right lane blind spot area and it appears other cars are biased the same way.
msnow, you've made an insightful observation here that perhaps other cars are also biased to stay somewhat to the left of the lane's center. This phenomenon would explain why we are far more likely to be concerned about a vehicle in the lane to the right of us than to the left of us.
Traffic Favoring Left Side of Lanes
So, I reviewed my dashcam videos from last week's trips, and they confirm that many drivers favor the left side of their lanes. Thanks msnow for the suggestion. I also saw a tendency to favor the inside of turns, a tendency that the Tesla AutoSteer tends to mimic.
So, I reviewed my dashcam videos from last week's trips, and they confirm that many drivers favor the left side of their lanes. Thanks msnow for the suggestion. I also saw a tendency to favor the inside of turns, a tendency that the Tesla AutoSteer tends to mimic.
Last edited:
Big Truck Favoring Left Side of Lane
Yesterday, while driving between Barstow and L.A., I came up upon a truck to my right that was driving close to the left side of its lane (please expand thumbnails into 1200 pixel images).
At this time my Tesla was in the middle of its lane as we were about to emerge from a shallow left turn, but as I came alongside the truck my Tesla painted yellow for the arcs on the truck's side and gravitated toward the left side of my lane.
Here you can see me nearly by the truck and my Tesla is over on the left side of my lane a bit now.
Once I was past the truck, it veered a ways into my lane! I was ready to take over as we passed the truck if it had veered earlier, but I believe that the Tesla would have gone as far as it could to the left side of my lane to prevent contact if the truck had veered earlier. Rest assured I was watching the truck like a hawk and ready to intervene if needed.
I think a veering vehicle situation like this is the type of situation that AutoSteer is best able to deal with. There would have been no need to enter the adjacent lane should the veer had taken place earlier, and the autopilot could have handled the entire save, I believe. Personally, I would have taken over and driven into the lane to my left if the truck had swerved earlier, because I had no idea how far the truck would go into my lane before it corrected.
Here is another advantage of the autopilot. Because my focus was not on steering (AS was doing that work), I could put more focus on situational awareness, i.e. watching the truck's position in its lane and being sure that traffic allowed an escape to the lane on my left if the truck veered into my lane. Driving on AutoPilot doesn't necessarily mean less attention paid to driving. Sometimes it means more time available for other critical elements of driving because I had one less task to perform.
Yesterday, while driving between Barstow and L.A., I came up upon a truck to my right that was driving close to the left side of its lane (please expand thumbnails into 1200 pixel images).
At this time my Tesla was in the middle of its lane as we were about to emerge from a shallow left turn, but as I came alongside the truck my Tesla painted yellow for the arcs on the truck's side and gravitated toward the left side of my lane.
Here you can see me nearly by the truck and my Tesla is over on the left side of my lane a bit now.
Once I was past the truck, it veered a ways into my lane! I was ready to take over as we passed the truck if it had veered earlier, but I believe that the Tesla would have gone as far as it could to the left side of my lane to prevent contact if the truck had veered earlier. Rest assured I was watching the truck like a hawk and ready to intervene if needed.
I think a veering vehicle situation like this is the type of situation that AutoSteer is best able to deal with. There would have been no need to enter the adjacent lane should the veer had taken place earlier, and the autopilot could have handled the entire save, I believe. Personally, I would have taken over and driven into the lane to my left if the truck had swerved earlier, because I had no idea how far the truck would go into my lane before it corrected.
Here is another advantage of the autopilot. Because my focus was not on steering (AS was doing that work), I could put more focus on situational awareness, i.e. watching the truck's position in its lane and being sure that traffic allowed an escape to the lane on my left if the truck veered into my lane. Driving on AutoPilot doesn't necessarily mean less attention paid to driving. Sometimes it means more time available for other critical elements of driving because I had one less task to perform.
Last edited:
The reason drivers favour the left side of the lane (in North America) is that when driving in the curb lane, the flat-causing garbage is mostly near the curb. Driving a bit to the left reduces the number of flats.
Boatguy
Active Member
I sailed about 20,000nm with an autopilot steering my boat and that was exactly the case. With the electronics on the helm, I could pay more attention to sails, radar, scanning the horizon, etc. etc.
I'm not getting the calibration done for another couple of weeks so I will also observe for the "favoring the left" of other drivers and where the MS thinks I am relative to the lane.
Not to get off-topic, but what does autopilot do on your boat? Rudder and...engine? Just curious.
Interestingly my son and I were discussing this yesterday on the way to school. He's very conscious of other drivers favoring the left side of the lane, probably because he usually sits in the right rear seat and he can see just how close we come to cars on our right side. Several times he's remarked that the autopilot should have pushed us away slightly from cars (trucks, etc.) in adjacent lanes. Going to have to mention this part of the discussion to him...
Bruce.
The real questions are, is it safer to be centered in the lane when the rest of the cars aren't? Do UK drivers bias right? Do we think Tesla should make this a configurable setting?
My guess is that Tesla will continue pretty much on their current path: starting with the car in the center of the lane, except for curves, where it will favor the inside of the turn as other drivers do. In the long-run, Tesla is looking towards an autonomous-driving future, and in that future it would be best if cars ran in the center of their lanes, rather than favoring the left side. That said, I think you will see AutoSteer continue to become more accommodating when it see the vehicle in the right lane is favoring the left side of the lane. Seriously, when I passed that truck that was favoring the left side very dramatically, the Tesla slowly gravitated to the left side of my lane and there never was an uncomfortable time for me until I saw the truck make a big swerve after I was already past him.
Boatguy
Active Member
Bruce - the AP on the sail boat steers (hydraulic ram to rudder) to a desired waypoint (adjusting for set/drift created by current or leeway), magnetic heading, or relative wind angle (true or apparent).
We drove about 120mi today and the MS is staying centered in the lane. My issue is apparently just discomfort with being in the center of the lane when the adjacent car is hugging the left side of the lane, a situation in which I would also drift left. This is particularly pronounced when I'm over taking a car that is hugging the left as the AS does not seem to sense the left bias through the radar so it doesn't see it or respond until the car is next to me.
I had one total AS failure today as we crested a hill with a mild left turn at the top and then descent to the left. The AS couldn't see lane lines past the crest and I felt it was going to just go straight off the turn at about 55mph so I took the wheel and kept us on track.
Overall I think there is probably no need to calibrate the car, but there is probably a need to calibrate the driver!
We drove about 120mi today and the MS is staying centered in the lane. My issue is apparently just discomfort with being in the center of the lane when the adjacent car is hugging the left side of the lane, a situation in which I would also drift left. This is particularly pronounced when I'm over taking a car that is hugging the left as the AS does not seem to sense the left bias through the radar so it doesn't see it or respond until the car is next to me.
I had one total AS failure today as we crested a hill with a mild left turn at the top and then descent to the left. The AS couldn't see lane lines past the crest and I felt it was going to just go straight off the turn at about 55mph so I took the wheel and kept us on track.
Overall I think there is probably no need to calibrate the car, but there is probably a need to calibrate the driver!
Boatguy
Active Member
I think this is a big challenge for autonomous driving systems. As experienced drivers, we pick up on clues about what a car may do. We know to adjust to erratic drivers, like maybe your trucker was nodding off, or a drunk, or someone who tailgates, or makes a lot of quick lane changes. Our perceptions go beyond pure physics and are a factor in our driving decisions. DNN's have a long way to go before they can do that.
After watching this video of an AutoSteer save, I am of the impression that AutoSteer can act quickly enough to prevent a collision if the driver in the next lane swerves a foot or so into your lane. I think AutoSteer's reactions might be faster than my own, but once the Tesla reaches the far side of its lane, it's up to the driver to continue the save if the offending vehicle continues to proceed further into the lane. As more saves emerge on video, we'll get a better idea of the capbilities of AutoSteer in critical situations.
Boatguy
Active Member
I've decided that my car does not need to be calibrated.
The MS stays reasonably well centered, it hunts at times, but overall it's my desire to bias away from a car I'm passing (whether on the right or left) that is at odds with the AP fixation on the center of the lane.
There seems to be a transition period as a vehicle passes from the radar in front to the ultrasonic sensors on the sides, during which lateral distance to the other car is not well established and I'm theorizing was the result of the original incident that triggered my concern. Just as I was passing a car, when it's left rear was closest to my right front, the MS veered slightly to the right, exactly when I would have been veering to the left.
In any event, I don't think this is about calibration, it's about the need for continued refinement of a product that is very much a beta and has lots of holes in it's ability to steer the car unattended. Cresting hills, sloped driveways, driving into the sun, these are normal parts of driving, but parts that still baffle an automated system. The AP is very useful, but I suspect it's going to need a couple more turns of the software, and the silicon, before it's a truly reliable partner in the cockpit.
Boatguy, thanks for the great post. Yes, the process of becoming comfortable with AutoSteer during highway driving is a two-way street. Part of the process involves your becoming more comfortable with the behavior of a Tesla when it is driving down the center of a highway lane, rather than favoring the left side as you are used to doing. Part of the process also depends upon the manufacturer slowly updating the autopilot software so that its behavior is more in line with techniques that make drivers comfortable.
You are wise to approach AutoSteer with caution. In some situations where the Tesla has difficulty seeing clearly-defined lane edges, the autopilot's abilities can be compromised somewhat. In time, you will learn to read the road as the autopilot does, and you will feel confident in its behavior in some conditions but not in others. Your judgement in such matters is still developing.
To put things into boating perspective, imagine that you have an assistant for some long-distance sailing. If your assistant is too green, you won't feel comfortable going down below for some shut-eye under most any conditions. If the assistant if somewhat competent, you might feel fine trusting him with a beam reach in light winds but not trusting him in conditions that could more-easily result in an accidental jibe. So, what you're doing is looking at both the competence of your assistant and the particular conditions present at that moment and making a decision based upon the two. Change the word "assistant" to "AutoSteer" and "boat" to "Tesla" and the principle is the same.
Since the Tesla autopilot is still in beta mode, Tesla is also saying, "Don't go below at all, stay on deck and remain ready to take over in a moment if needed."
About two weeks ago in Norway, Elon Musk announced that using the deployment of an airbag to indicate an accident has occurred, Teslas on autopilot have only half as many accidents as Teslas being hand-driven. This is an impressive statistic, especially when you realize how new the autopilot is and how far it is yet to evolve. Personally, I have found that when lane markings are good, I can drive all day without taking over. I'm still watching very carefully, but my level of confidence in my assistant's skills is high when the conditions are good (good road markings).
You are wise to approach AutoSteer with caution. In some situations where the Tesla has difficulty seeing clearly-defined lane edges, the autopilot's abilities can be compromised somewhat. In time, you will learn to read the road as the autopilot does, and you will feel confident in its behavior in some conditions but not in others. Your judgement in such matters is still developing.
To put things into boating perspective, imagine that you have an assistant for some long-distance sailing. If your assistant is too green, you won't feel comfortable going down below for some shut-eye under most any conditions. If the assistant if somewhat competent, you might feel fine trusting him with a beam reach in light winds but not trusting him in conditions that could more-easily result in an accidental jibe. So, what you're doing is looking at both the competence of your assistant and the particular conditions present at that moment and making a decision based upon the two. Change the word "assistant" to "AutoSteer" and "boat" to "Tesla" and the principle is the same.
Since the Tesla autopilot is still in beta mode, Tesla is also saying, "Don't go below at all, stay on deck and remain ready to take over in a moment if needed."
About two weeks ago in Norway, Elon Musk announced that using the deployment of an airbag to indicate an accident has occurred, Teslas on autopilot have only half as many accidents as Teslas being hand-driven. This is an impressive statistic, especially when you realize how new the autopilot is and how far it is yet to evolve. Personally, I have found that when lane markings are good, I can drive all day without taking over. I'm still watching very carefully, but my level of confidence in my assistant's skills is high when the conditions are good (good road markings).
Bonlaw
Member
Because it occurred so fast, I don't have full confidence in my recollection. But this is what I believe I experienced.
I was driving in the far left HOV lane on AP. I hear a loud motorcycle approaching from the rear. The Tesla moved itself over a couple of feet to allow the lane splitting motorcycle to pass and after it did, the Tesla moved a few feet to center again.
Did his happen?
I was driving in the far left HOV lane on AP. I hear a loud motorcycle approaching from the rear. The Tesla moved itself over a couple of feet to allow the lane splitting motorcycle to pass and after it did, the Tesla moved a few feet to center again.
Did his happen?
Similar threads
