ball cannon???

[Uberbots]

This game was practically designed to be a shooter game. It was hinted to man a times in the kickoff that robot to robot contact (or robot to trailer for the purposes of scoring) would be very difficult to do and would require much time and coordination (and i don't believe the other team would be so keen about you scoring in their trailer)

The camera systems given to us along with the lazy suzan (and even the game video) is almost a direct notice to us: use a shooter! but the question is how you would make it effective?

unlike 2006 where you would fire at a stationary target, this year you must fire at multiple moving targets which you cant assume will be in front of or behind you. it would be near impossible for a pilot to aim the turret at an opposing robot, accounting for all the possible motion and angle, and fire a successful shot. We are, however, armed with an extremely powerful vision system this year, which can do advanced color and blob analysis at speeds inconceivable with the CMUcams.

The basic logic i have hammered out is you would have a camera mounted on a pan/tilt on top of your turret. This will give your cam a constant, known height. we are also assuming you have a well designed shooter with a control loop that can have an accurate firing distance and angle, as well as approaching those parameters in under a second. The camera is on the constant hunt for beacons that your team wants to shoot at. when it locates a beacon, there is a control loop that tells the camera to lock its pan/tilt such that the beacon is in the center(ish) of the feild of view. Servo/image error will then allow for the angle of the turret to be corrected for.

The interesting part comes with shooting. Both you and your target are (perpetually?) moving. to account for this, your software will need to be able to do motion prediction (probably using second-degree extrapolation of some sort), and using empirically or mathematically determined data, will be able to lock onto where the target will be. With a system like this in place, even a fast moving and evasive target will not be able to outsmart your turret... well unless its out of range or they move in a manner that second-degree extrapolation cannot predict.

Software is incredibly important this year! you can close your loops with far greater precision (double precision floating point values!!!!!) than you could in previous years.