yes 8m is a lot. in the video i was using 3m. But it is different now
Two kinds of coordinate systems - the world and local. we run around in a position and heading in our big world but our arms are referenced to us locally.
Pure pursuit calculates an arc based on how far the pivot is from the line. So we need to build a circle around us to arc to the line. For that we use local coordinates. All the integral does is tweak the local heading fooling the system to turn a bit more without correcting that turn (since we are using distance to calculate the original arc).
So it works out pretty linear to 15 degrees being about 4cm of movement per 0.01 radians (0.57 degree). so 0.1 is about 40 cm off line compensation. of course it could be higher, but really doesnāt need to be.
The bottom number is the derivative or the rate of change of distance moving to the line. when driving to the line, we donāt want integral, the distance between samples change a lot - derivative is high. as we get close to the line, that distance barely changes so when it is less them 0.2, then we start building up the integral and force it over that last little bit by adding some angle in the local coordinate.
Now you too know the magic trick!
2 Likes
i guess it comes down to how many bugs to release. A few of us are driving it now, so wonāt be long. Just those unavoidable click the right sequence 400 times and up pops a bug kind of bug. Very soon.
3 Likes
Vili doing a field test as we decide which steer method to keep. PP is pretty smooooooth
2 Likes
Hy, very cool to get a new version.
Do you already know when the source code is available?
As written above. First release very soon. 
The speed of 3.4 km / h is very slow, the normal working speed is 6 or 7 km / h or more for a tractor or combine.
Can you test at working speeds? that would help decide which one is correct.
pp has always worked well, even without imu. from very slow up to 18mph.
For me as well. PP has always worked better for me by far, slow or fast.
1 Like
Speed seems to have always been the factor, I could never get PP to work very well for me at any speed with dual anyway, but I also need to go super slow for veg etc, Stanley just worked for everything really well, so just used that but as long as it works, makes no odds to me really! Open to any option!
Is the antenna positioned on the roof?
Yes, just at front of cab, 2WD style tractor.
I find that so very odd that with dual antenna you couldnāt PP to work. But a lot has changed. This PID loop on the Arduino now runs at 50hz, has no delays etc. I give @GoRoNb full credit for showing me the way. Such a vital and yet simple change
What was is doing that made it seem worse the Stanley.
Very excited to try all these changes, it sounds like this version has ALOT of big changes / fixes!
What was it doing? wandering back and forth? too slow to get to line?
I honestly cant actually remember it was such a long time ago I was trying it. I seem to think I could always āget it to workā at one speed, but then if i tried it at a much slower speed, the settings would all need changing to get it to be stable / work againā¦vs stanley which I could just set and forget, and would work for it allā¦
yes, it is a lot more speed sensitive. Looking at some Raven manuals and they have different gains for different speeds rather then trying to do it auto.
Iāve had exactly the same problems with pp and dual as darren. Stanley works fine for me with the same settings from 1 kmh for planting veg or seedbed preparation up to 14 kmh for tillage. With pp i had to adjust the settings for every single speed.
Iām so impatient on waiting for the new version. Videos look great!
Do you think this issue will still be the case if we go to PP only for dual users Brian? Or have things changed so much now its likely to be totally different anyway?
Important to get PP good to work is the right setting of look ahead time. With dual I was able to reduce lookahead and this forces the tractor much more to the line. But on sidehills I had to increase sidehill draft gain.