It's simply meant to be an implementation of a heuristic, based on the traveling salesman problem, that takes into account physical considerations (speed, acceleration, direction) and processing limitations (RAM, processor cycles) for both initial setup and decision-making at each step.
The speed/direction stuff reminds me of the kind of skating that hockey players do (is it more effective to go in one direction, stop, and turn around, or is it better to modify your line and preserve momentum? in this game, too, is it better to accelerate greatly and bounce off a wall behind your target, or approach more slowly in order to modify your line without an abrupt change in direction?).
The processing limitations are interesting too, and provide for an interesting optimization exercise.
Or, by "I don't understand", should I simply answer "it's fun"...? ;)