Study On Compiling Optimization Method Of Crew Rostering Of Urban Rail Transit

As an important part of operation and management of urban rail transit, crew rostering is the last major step to implement train operation plan. Based on crew scheduling, it is a working plan for train drivers over a period of time. At present in the practice of operation and management of urban rail transit in China, crew rostering is always handmade by engineers with their rich experience. However, this type of approaches is time-consuming and hard to take many compiling objectives into account. For convenient operations, the handmade crew rostering is always obeying a fixed cycle. It is really not very flexible but suitable to the handmade compiling method.According to the existing research, some crew rostering optimization models are proposed through analyzing the components and influencing factors of crew rostering in this dissertation. The main contributions of the dissertation are summarized as follows.(1) Establishing period-cycle pattern and non-cycle pattern crew rostering models. Different from the fixed cycle pattern, the periodicity of external conditions and differences between weekday operation and weekend operation are considered in the period-cycle pattern, and rostering units are regarded as cycle elements. In addition, based on cycle pattern model, non-cycle pattern model has no periodicity and it is more flexible than the cycle one.(2) Proposing two optimization objectives, workload equilibrium and all rest at weekends. Working equilibrium requires that everyone takes about the same workload. Based on working equilibrium, the objective of all rest at weekends tries to ensure drivers to rest for two days at weekends so that they can attend those important social activities which are often organized at weekends, it will make the crew rostering more hu and attractive.(3) Decomposing a crew rostering model into two stages, rostering partition and rostering allocation. On partition stage, the same type of shifts are not distinguished, while the specific shifts are allocated to the result of partition stage on allocation stage.(4) Designing an intelligence algorithm to solve the proposed crew rostering models. The particle swarm optimization with a simulated evolution algorithm nested inside is proposed to solve the models of different patterns, objectives and stages.The feasibility and superiority of models and algorithms are verified with a real case, it shows that workload of the result is more balanced than the existing crew rostering with the objective of workload equilibrium. If the objective is all rest at weekends, a high percentage of full rest weekends can be found in the result and it is efficiently optimized comparing to the existing crew rostering.