Coding Before Scrimmage

Tags: control
Personhours: 16
Coding Before Scrimmage By Cooper, Karina, Bhanaviya, and Trey

Task: Finish the temporary auto and work with drivers for teleop

Tonight, the night before the scrimmage, We worked on making the depositing of the stone and parking of the robot more reliable. Or as reliable as possible, as we are planning to use FrankenDroid, which is somewhat in need of repair, which I also did with the help of Trey, Bhanaviya and Karina. This had a few changes come with it, as while we solved the problems of when we started the auto, there were still many that cropped up.

Problem #1 - dragging the base

In the auto, we need to drag the build area into the taped off section in the corner. This poses a problem, as dragging it can lead to major inaccuracies in estimated positioning. This, however can be solved somewhat easily once we have a distance sensor, which we could use in junction with PID based turning. Though in theory I could have done it with just the PID turning. While I would have loved to test that, there was another problem--

Problem #2 - problem with hook

There was a problem with our hook. I tried every time I ran auto to get the hook to work. I changed the return value, I changed the physical positioning of where it started, yet nothing worked. This was interesting, as it does work in the teleop. In any case, it prevented us to actually dragging the base in this version of auto. Looking back on it, there was a possibility that I needed to set it as a timed state, like the gripper, since we were using a servo to control it. While its unlikely, it's possible.

Problem #3 - PID Tuning?

This was the major issue of tonight, which we haven't found the root of just quite yet. During the auto, at the third turn, where the robot turns to heat to the foundation, there is a ~25% chance that the PID does not check where it turns and it just continues wherever it turns to. This usually leads to it overshooting and then ramming in to the wall. There is a temporary fix, however. For now, it seems that if only happens after we upload the code to the robot, or if we run auto fresh off of it being off. That is to say, if we run the auto at least for a second and then reset and re-init, then it will be good. This is a good thing however, as any chance we get to fix the underlying code's problems, that means we won't have to make a work around after in the season.

Problem #4 - putting the block on the build platform

This was the major fixable problem in the code. During auto, we need to take a block from the quarry and put it on the foundation. The problem is when we actually go to deposit it. When we go to put it down, we need to be very accurate, which with FrankenDroid is not easy. With no distance sensors, the best we can do is to tune the exact movements. While this isn't the greatest solution, this will do for now. In the future, we will have a distance sensor so that we can know where we are exactly in relation to the base.

Next Steps

We need to implement the distance sensors and other sensors on the robot. Obviously we aren't going to be using FrankenDroid for too much longer. TomBot may bring new innovations like a telemetry wheel which will make auto more accurate.

Date | October 26, 2019