Multi-Agent Based Crew Pairings

Crew scheduling is a complex, large-scale and continual work, characterized of multivariable, close coupling and multi-objectives and so on. But in most of domestic airlines, the process of crew scheduling is still on the stage of the manual level or depends on oversea's system. Therefore, that how to increase the computer application in the process of crew scheduling comes to be an important job of demestic airline to enhance competition and to control cost.This thesis presents a modal for the crew scheduling based on the multi-agent technology, an advanced A.I. one. First, the characterization of the crew scheduling is analysied, especially the realization technology of crew pairing in this process. The basic principles of typical algorithms which used in the crew pairing are evaluated based on the study of the capabilities of these algorithms. Considering the huge data quantity and the plentiful objectives of the crew pairing, genetic algorithm is selected to slove the problem, which can realize the overall searching efficiently. Next,the mathematical model , built according to the target function, is optimized and scheduled by the means of genetic algorithm. After the chromosome coding method is designed, fitness function is established corresponding to the crew pairing objects and the arithmetic operators of genetic algorithm is designed. Then, the crew pairing modal based on the multi-agent is built. The agents cooperate with each other to realize the operators function. The design of the modal, including the system flow, system basic constitution and the realization of most agents is introduced. Finally, in the end of the dissertation, some further suggestions are given to improving the modal.