Pink v. Cyan Remote Scrimmage Code Post MortemTags: control
Task: review the code changes needed from our latest scrimmage
After looking over the progression of matches of the scrimmage, it is obvious that there needs to be some code changes. I’d like to go over in detail those issues, and how they’d be solved.
First off, the issue prevalent when the robot turns in auton- the PID loops aren’t tuned. The PID constants were from last year, and with the robot arm being much lighter right now, the kP was ridiculously high. Both the base and the turret overshoots really bad when doing right angle turns. While I could go into the specifics of how we tune our PID loops, I’m instead going to explain why we aren’t going to fix it yet. It's really simple-- the launcher is going to be as heavy, if not heavier than the arm of last year. At the very least it's going to be different, which means if we tune the PID right now, we’ll have to tune it again very soon when the launcher is mounted. Therefore, we will tune it then.
Next, and maybe a bit more broad this time, auton is busted. It's obvious from the many auton runs that ended up with the robot wildy veering off course. This is caused by the fact that for our long run to the other side of the field uses a wall following, and that I made and tuned it in 30 mins before the scrimmage. To fix this, we need to implement a more sensor-fusion based model, including vision and odometry to check each other off on, or improve the distance sensor’s reliability.
During teleop, it is evident that having wobble goals get stuck on the robot was not good. While this is another thing that is subject to change, the appropriate solution for this would have been making an articulation that puts the arm at the correct position, and having some logic in the call to the toggle articulation to prevent it if the arm was below a certain limit angle.
Fix all of the items listed in this post.