Application of genetic algorithms in the aerospace domain

This report contains the paper that was accepted for presentation at EUFIT '96, the Fourth European Congress on Intelligent Techniques and Soft Computing, Aachen, Germany, September 2-5, 1996. In the paper, two case studies, performed at NLR and investigating the applicability of genetic algorithms in complex optimisation problems, are presented. The first case study concerns the use of algorithms to support planning in Air Traffic Management (ATM). Based on given starting conditions of aircraft, they must reach their exit conditions through flying trajectories such that no conflicts occur. Trajectories are modelled by means of a Mode Of Flight based representation, i.e. trajectories are specified using straight lines, curves, accelerations and altitudes. Alternative trajectories are investigated from the first conflict on. Problem specific knowledge has been incorporated in the mutation and recombination operators. The importance of recombination increases when scaling up to larger problems. Problem specific genetic operators were also used in the second case, which concerns parameter optimalisation for examining airfoil pressure distribution on aircraft wings. The objective function was taken from an earlier experiment with non-linear optimisation methods which did not always find a solution. Although the genetic algorithm was able to find a solution, it was found that the initial population had a great influence on the quality of the final solution. Conclusions that can be drawn from these case studies are that the large amount of variables is difficult to manage and the necessary computing power can be quite large. On the other hand, when the initial problem is investigated and some domain dependent knowledge is added, promising results are obtained. APPLICATION OF GENETIC ALGORITHMS IN THE AEROSPACE DOMAIN H.H. Hesselink, H. Kuiper, J.M. van den Akker Informatics Division National Aerospace Laboratory NLR, The Netherlands email: {hessel,kuiperh,vdakker}@nlr.nl