Evolutionary computation in air traffic control planning

Air Traffic Control is involved in the real-time planning of aircraft trajectories. This problem is a heavily constrained optimization problem. We concentrate on free-route planning, in which aircraft are not required to fly over way points. The choice of a proper representation for this real-world problem is non-trivial. We propose a two level representation: one level on which the evolutionary operators work, and a derived level on which we do calculations. Furthermore we show that a specific choice of the fitness function is important for finding good solutions to large problem instances. We use a hybrid approach in the sense that we use knowledge about air traffic control by using a number of heuristics. We have built a prototype of a planning tool, and this resulted in a flexible tool for generating a free-route planning of minimal cost, for a number of aircraft.