[Jon Stratis]

I'm not sure if I can remember our entire design process, but here goes...

We spent 1 night the first week prototyping ways to shoot the ball (a wheeled shooter, a catapult, etc). I got to work with the group that was focusing on a wheeled shooter, and we talked about a couple of things:
- How do aerodynamics affect a ball in flight? (think knuckle ball versus curve ball, versus slider, etc)
- How does spin affect a ball contacting a stationary surface? (think about how a pool player can change the angle of reflection on a bank or kick shot)
- What was our goal and backup plan for the robot?

The team came up with these answers:
- A ball with no spin tends not to travel in a straight line as well as one with vertical spin. Side spin can cause a ball to curve from one side to the other.
- A ball with back spin can cause it to reflect downwards off the backboard. Top spin will cause it to reflect straight off the backboard and miss the basket.
- Our main goal was to score from the key. Our backup plan was to have enough power to act as an effective feeder robot playing defense.

From this, we built a quick prototype with some spare 2x4's and old wheels. We wanted to be able to test differing amounts of backspin on the ball (while maintaining enough speed to hit the wall in a "realistic" fashion), so we put wheels on top and bottom.

The initial build had 1 wheel on top and 1 on bottom (each powered by a drill). We quickly noticed that the ball tended to "squirt" out one side or the other if it wasn't perfectly placed, and thus not shoot straight. We fixed this by going 2 wheels on top and 2 on bottom, which helped to center the ball and make the shot consistent side to side. We could alter the strength of the two drills to vary the backspin on the balls.

It turned out that this prototype worked really well. The shots were more consistent than anything else we built. So, we went with that as our basic design.

In building the "final product", we decided to get smaller wheels with better grip, and to add some heavy flywheels to help reduce variation in the shots (since they would decrease the amount each shot slowed the wheels, we could shoot more rapidly). We also added some encoders onto it to allow us to more easily monitor the speed and set a more precise speed for each shot.

In order to figure out how fast we wanted the shooter to go, we turned to empirical evidence. We knew our prototype was about 80% of what we wanted to shoot from the key. So, one of the mentors brought in a strobe light with a variable speed setting. We taped a dark line on the wheel, and brought it up to speed. By adjusting the speed of the strobe, we could make it appear like the line stopped moving, thus giving us the rotational frequency of the wheel. A little math, and we could figure out how fast we wanted our real wheels to spin, and work backwards from there to figure out the desired gear ratio.

After that, it was really just a whole lot of testing with the actual robot to figure out what the best values were for the code.

The end result was a very consistent shooter. If we placed the robot in a single location and set the speed to be constant, we could feed the same ball in over and over again, making a basket each time.

The only thing we didn't take into account very well was the variable density of the balls, and how "soft" and "hard" balls would shoot differently. We'll get that next time there's a shooting game!