OPTIMIZATION OF VEHICLE ROUTING PROBLEM WITH GENETIC ALGORITHMS

V magistrski nalogi sta opredeljena pojma genetskih algoritmov ter problem dostavnih vozil. Podrobneje so predstavljeni operatorji genetskih algoritmov ter vrste problemov dostavnih vozil, katere je potrebno spoznati skozi izdelavo praktičnega dela naloge. V nalogi smo se osredotočili na implementacijo problema dostavnih vozil z več skladišči ter mehkimi časovnimi okni. Problem smo podrobno razdelali, ga implementirali in nato testirali. Rezultate testiranj s popravki smo med seboj primerjali ter prikazali analize. Z nalogo smo želeli izvedeti, kako rešiti problem dostavnih vozil s pomočjo genetskih algoritmov in preveriti, ali vhodni parametri, ki jih lahko spreminjamo, vplivajo na končne rezultate. Želeli smo izdelati rešitev, ki bo vrnila v končnem času sprejemljiv rezultat, ter oceniti učinkovitost rešitve. Da smo lahko odgovorili na vprašanja, smo implementirali genetski algoritem in na njem testirali različne nastavitve. Na podlagi pridobljenih rezultatov smo ugotovili, da se z genetskimi algoritmi problem dostavnih vozil da rešiti v končnem času, da na končne rezultate vplivajo vrednosti vhodnih parametrov in tako dobljene rešitve se lahko že v kratkem času izboljšajo tudi do 30 %.The Master Thesis examines genetic algorithm and the problem of delivery vehicles. It presents in detail genetic algorithm operators as well as those problems of delivery vehicles, which needed to be considered through the practical part of the thesis. The Thesis is focused on the implementation of the delivery vehicles with multi depots and soft time windows. The problem is analysed in detail, implemented and tested. The test results were compared and the analyses presented. The aim of the Thesis is to find a solution to the problem of delivery vehicles with the help of genetic algorithms and to check if the input parameters, which can be modified, can have an impact on the final results. The purpose of the work is to create a solution, which would produce acceptable results in the real time and to estimate the efficiency of the solution. In order to be able to answer the questions, the genetic algorithm was implemented and tested for different settings. Based on the obtained results it can be concluded, that the problem of delivery vehicles can be solved in the real time, that the input parameters value can have an impact on the final results and that solutions, obtained in such a way, can be improved up to 30 % already in a short period of time