A Bi-Objective Airport Gate Scheduling with Controllable Processing Times Using Harmony Search and NSGA-II Algorithms

Abstract

Optimizing gate scheduling at airports is an old, but also a broad problem. The main purpose of this problem is to find an assignment for the flights arriving at and departing from an airport, while satisfying a set of constraints.A closer look at the literature in this research line shows thatin almost all studies airport gate processing time has been considered as a fix parameter. In this research, however, we investigate a more realistic situation in which airport gate processing time is a controllable. It is also assumed that the possible compression/expansion processing time of a flight can be continuously controlled, i.e. it can be any number in a given interval.Doing sohas some positive effectswhich lead to increasing the total performance at airports’ terminals. Depending on the situation, different objectives become important.. Therefore, a model which simultaneously (1) minimize the total cost of tardiness, earliness, delay andthe compression as well as the expansion costs of job processing time, and (2) minimize passengers overcrowding on gate is presented. In this study, we first propose a mixed-integer programming model for the formulated problem. Due to complexity of problem, two multi-objective meta-heuristic algorithms, i.e. multi-objective harmony search algorithm (MOHSA) and non-dominated sorting genetic algorithm II (NSGA-II) are applied in order to generate Pareto solutions. For calibrating the parameter of the algorithms, Taguchi method is used and three optimal levels of the algorithm’s performance are selected. The algorithms are tested with real-life data from Mehrabad International Airport for nine medium size test problems. The experimental results show that NSGA-II has better convergence near the true Pareto-optimal front as compared to MOHSA; however, MOHSA finds a better spread in the entire Pareto-optimal region.Finally, it is possible to apply some practical constraints into the model and also test them with even large real-life problems instances.