State Estimation
The state of the robot is defined by its two dimentional positions and its angle of rotation about the vertical axis. In order to accomplish most of the tasks, DSP needs an accurate estimate of this state. A dead reckoned state is created by integrating readings from the encoders on the wheel's motors and the gyro sensor. Since this is an integrated state, it is prone to error accumulation from the wheel slippage and in particular, gyro drift. The primary way we accounted for this error was by using Ladar measurements.

The Ladar gives a measurements of the distance to the walls around the robot. By extracting corners from these distances and matching them to the expected course corners, we can create a new measurement of the state. Additionally, by finding the minimum distance between the robot and the nearest wall, we obtain one component of the robot position and the orientation. The estimation of our state is updated by these measurements through a Kalman filter.

Can Grabbing
We found the can using an IR sensor on the robot. The robot turns slowly, reading IR distances until the can gives a smaller distance. That angle is recorded as the right side of the can. The robot keeps turning until the IR distance goes back up, giving the left side of the can. The average of the two sides gives us the center of the can. The robot can then drive (open loop) in that direction until the can is in the gripper and the gripper is activated.

Ball Finding
HSV color values were found for the blue and orange golf balls. When the robot enters a region that may have balls in it, the vision algorithm alternates between these values on each frame. If an object is found, the robot extimates its position. The ball's position relative to the robot is estimated with a map from the objects centroid on the image plane to a location floor. This is then mapped to the world coordinates using the current state. If the ball is within or very near to an allowable region, the robot will attempt to grab the ball.

Ball Capture
The robot uses a proportional control on the linear and angular velocity to line up the ball at known distance in front of the ball hopper. Once the ball is lined up, an open loop algorithm opens the gate drives forward and closes the gate to capture the ball.