[apalrd]

We bagged the robot with a higher-speed top roller and played like this at Kettering. We then looked at several other robot's up close, and built a pincher-roller for Troy using the practice robot as a guide. This taught us two things:
1. pincher rollers are waaay better
2. the practice bot is really far off of design specs

Ours was a floating design, so the entire assembly was a four-bar linkage. The two horizontal bars supported the bottom plate and top roller, one vertical "bar" was the chassis, and another held the two pieces a fixed distance apart. Using bungee, we sprung it down to avoid carrying penalties, and put some little wheels on it so metal never touches the ground. We had an issue with carrying penalties at Troy, so we fixed it in software and haven't had problems since.

We actually had an interesting problem with ours: It had such a good grip we could no longer kick more then about 3 feet. So we fixed it in software and can now kick the full 33 feet (+ a few more) it could with the old roller.

We carefully tuned the depth of the top roller and bottom plate, as well as the distance between the two, to get the most grip while not allowing a piece of paper to go under the ball. Software helps with this: If the roller is allowed to run continuously, it will gradually suck the ball in more, and carry it. If you just kill the roller, then it will loose the ball. We implemented a duty-cycle based roller kill to pulse the roller when holding a ball (determined by a broken-beam sensor), and to continuously run the roller when going backwards (both joysticks are positive, reverse) to prevent loosing the ball. Timing this was tuned to 5 iterations on for a 20 iteration cycle.