Parameter sensitivity of the genetic algorithm for flight planning optimization

Abstract

Increase in fuel costs have pushed airlines to look for ways to cut costs substantially. One area which costs can be minimized is through the usage of optimal flight routes. Such flight routes can be characterized as having the shortest distance between two points as well as an optimum flight altitude to allow short flight times. The current flight route optimizer used by Flight Focus Pte Ltd, a provider of flight computers, is Dijkstra’s Algorithm. Although this algorithm is able to compute the global optimal route based on specified cost functions, the process is time-consuming especially when considering a large search domain subject to several cost functions. The Genetic Algorithm (GA) is another method of optimization which has the potential to be able to do 3-D optimization in a shorter time period. Functional tests done using GA have been benchmarked to be able to obtain results within 5% of the global optimum obtained by DA. In order to improve the results obtained by GA, studies on the various parameters were done so as to evaluate how each parameter affects the speed and quality of the results obtained. Tests have shown that excessive elitism rates have the effect of speeding up the computation but at the expense of results quality while mutation has the reverse effect.