A GENETIC ALGORITHM FOR THE POST TAKE-OFF MANOEUVRE OPTIMIZATION *

This work introduces and compares dif- ferent optimization methods to find the best longitudinal manoeuvre required to an aircraft to reach, after take- off, in the minimum time a given altitude at a given speed. A longitudinal flight simulator based on non- linear equations, integrated numerically with a Runge- kutta method, has been built. The optimization algorithm used in this case is based on G.A. theory (Genetic Algorithms), in comparison with a simple random-search algorithm and a random- search and gradient-descending algorithm. Good performances have been observed as long as the cost function of the G.A. is well adapted to the problem. A number of situations have been analized and compared with the results and performances given by simpler algo- rithms based on random search. Common optimization algorithms based on gradient descending cannot be used in this case, since there are too many sub-optimal solu- tions. Results are presented as comparisons between the time histories related to the elevator angle. The results obtained with the GA and random search calculation programs are practically identical; but in the second case the calculation time to obtain the optimizated manoeu- vre is much lower.