[meta] Update object positions using laser scan data

[HALtheWise]

The vehicle should be using its laser and (if necessary) camera to improve its estimate of the locations of navigation buoys and both docks.

[HALtheWise]

I think the best way to store the location of stuff on the course is with ROS's tf (transform) system. We can define a coordinate frame attached to each challenge object, then just give navigation commands relative to the course objects. For example, the dock for the Docking task could have a tf frame stuck to it, and that makes maneuvering to it a simple matter of targeting a point defined in that coordinate system.

A typical run would look like:

• Before the round, we initialize the locations of field elements as accurately as we can to GPS coordinates
• When the round starts, we use our GPS to estimate our own position, and operate using the previously assumed locations.
• When we get close to an object, we look for things in our LIDAR (and possibly camera) data that are near where we expect the object to be. We use this information to update our estimate of the object's position. This step may require a Kalman filter, or could just be done heuristically.
• To complete a task, we navigate to a series of waypoints defined in the frame of the task object(s).

What do people think?

[HALtheWise]

I've started writing up some more details in the wiki, feedback appreciated.

[NathanYee]

For finding lines: http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.RANSACRegressor.html