| Summary: | 碩士 === 國立成功大學 === 交通管理學系碩博士班 === 90 === Abstract
Train driver scheduling is the later stage of operation planning process for railway industry. The preceding planning includes Time tabling, Train diagramming, and Vehicle scheduling. Owing to the restrictions of salary, reasonable working and rest time and driving regulations, train driver problem has become more complex than other kinds of public transportation driver problems. Therefore it’s quite necessary to find out an algorithm combined with rapid computation to replace manual operations.
In this thesis, we perform a driver problem of Taiwan Railway Administration in Kaohsiung depot as a case study, including Crew Scheduling and Crew Rostering. In crew scheduling, the problem can be divided into two sections- (1) Shift Generation and (2) Shift Selection. In shift generation, we use a network heuristic to eliminate restrictions and produce a set of legal shifts. By the way of parameter control, potential shifts with high performances are generated. Shift selection is defined as a Set-Covering Problem (SCP). According to the properties of multi-objective, genetic algorithms is applied to choose a suitable shift set to cover all train trips. In crew rostering, we regard it as an improved Traveling Salesman Problem (TSP). In terms of performing all shifts once, we solve the problem under the objective of minimum cycle by genetic algorithms.
In the empirical study, this research integrates TRA’s current data and rules and combines algorithms with operational principles. Finally, we can make up some conclusions as follows:
(1) In real world, optimal methods aren’t an ideal way because of numerous limits of model formulation.
(2) Genetic algorithms accompanied with some heuristics can simultaneously consider multiple performance indices and produce many good solutions in a short time. Consequently, they are more suitable for practical problems.
(3) This research proves that we can use genetic algorithms combined with heuristics to replace traditional manual operations and obtain better results.
Keywords: Crew scheduling, Crew rostering, Genetic algorithms, Set-covering problem, Traveling salesman problem.
|
|---|
