Air Traffic Optimization in Training Environment: A Comparison Between Genetic Algorithms and Simulated Annealing

One of the main stages in the training of an air traffic controller: practical work. The trainee must apply all theoretical knowledge concerning the separation rules decreed by ICAO to resolve a traffic situation presenting a certain number of conflicts generated by a group of aircraft (A/C). At the end of his performance, the solution proposed by the trainee is compared to that of the instructor or that generated by a control simulator. For a trainee wishing to train in the absence of an instructor or simulator, it is useful to offer them a simple tool that can provide an optimized solution. Thus, the use of metaheuristics in this tool is recommended and the challenge revolves around two main objectives, eliminating the risk of collisions between A/C while minimizing changes to initial flight plans. Regarding previous work, where genetic algorithms and simulated annealing were used to the problem, we present a comparison an analyse of the results obtained for both principals objectives by this two algorithms through a series of tests on exercises inspired by real practical work.