I've just learned that the Tesla cybertruck at this time does not have autopilot. One speculation is that it is because the CT's steer by wire innovation. Thinking about this further it could be autopilot software is more complicated because of CT's variable turn ratio. That is just driving down the straight away road the steering ratio is small but when making sharp turns the ratio is larger. This is what makes CT's ability to completely turn around ability without the normal hand over hand manipulation of the steering wheel possible. So my question to software control designers is how much more complex and difficult is it to write autopilot software with this added feature of variable turns ratio steering?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Cybertruck Autp-Pilot
- Thread starter Ridgeline
- Start date
Or you can look at from the aspect that's been told, just going to take a little bit to port it.
I'm pretty sure that steering and variable steering have NOTHING to do with it. How do you think that it turned the wheels before, it has always steered by wire. RW steering is just something that makes the steering "a little more"
It wouldn't surprise me if the steering interface to the autopilot is exactly the same as the old one.
It took Tesla about 6 months when the ultrasonic sensors removed from the cars last year. I highly expect that for the vast majority of buyers, that FSD will already be there. It is even possible that by the time that Foundation starts shipping to real customers, it may be there.
I think you are correct that auto pilot is steering by wire. but I think it's different now from a control loop analysis. Without variable turns ratio, the output to the control element (steering actuator) is linear with respect to the error (difference in where the car needs to be versus where it is). But with the variable turns ratio the output is not linear with the error. In variable ratio, for a sharper turn the controller would need less output when it approaches the end of its stroke of moving the car where it needs to be. Perhaps though the corrections are taking place so fast (errors are in millimeters and corrections occur in micro seconds) that it makes no difference, ie the output just keeps pulsing the steering until the car gets to where it needs to be in the road. I could be wrong about all this, it's complicated and I tend to think about control valves that are controlling flow through pipes. If anybody knows anything about the control algorithms of auto pilot I'd be interested. Real life situation, it took me quite a while to trust auto pilot to go around that tight curve at 60 mph coming off the freeway.
A) Most all of the automatic functions are now or will be driven by AI. It just to be retrained.
B) If the steering was indeed in legacy code, all you would need to do is a little conversion routine. Linear to exponential routine are easy. At an absolute minimum, a simple lookup table works.
Who knows, even the fly-by-wire wheel may already have the algorithm built into it, to feel better to the driver.
prettig
Active Member
No AP (auto steer) could be a dealbracker, i use it all the time on my other Tesla's. As long as we get it it's ok to wait. But no AP would be a serious issue for me. Should be standard (like every other Tesla). I do not need/want FSD, just AP.
Took CT delivery a week ago. No FSD or Lane Assist. Tech at service center said I'd have it OTA in another week "or so." Can't guarantee his info is accurate but this is what he said. Did get OTA update on 4/6/24 to 2024.8.8 but just bug fixes nd minor enhancements.
Tech probably just made you happy with the response, I doubt if they had any more info than this forum. They often get their information from the forums.
