Optimizing for Reliable and Sustainable Public Transport

Abstract

To make a public transport network efficient, a lot of planning has to be done. This planning leads to complex optimization problems. This thesis is devoted to key issues in the optimization of public transport with the focus on reliability and sustainability. For the optimization for reliability we use historical measurements of run times to develop models for evaluation and optimization of reliability of a public transport network. For the sustainability we include environmental concerns in the optimization: electric vehicles with no exhaust from the engine as well as a mixed fleet with electric vehicles and other vehicles of varying degrees of exhaust. We aim to develop models and optimization methods for these problems which can be used in practice. These methods will be able to handle real-life sized instances and should perform better than currently used algorithms or heuristics. In this thesis, we have investigated four different subjects: Trip Runtime Determination: Based on measured trip runtimes we determine the optimal planned trip runtimes on stops in order to minimize average delay. We developed a new method which leads to a decrease of the average delay at stops with 60 percent. We also optimized on minimal passenger travel time instead of minimal average delay at stops. This halves the average waiting time for a passenger and reduces the average travel duration with up to 5 percent. Robust Vehicle Scheduling: In order to reduce knock-on delays in vehicle schedules, some buffer time is scheduled between trips. When the first trip is delayed, the second trip will leave on time. Usually, a fixed buffer time is used. We investigated this subject and developed a new model for assigning buffer times. We were able to improve the punctuality with up to 3 percent at no cost. Electric vehicles: We have developed a model to schedule electric vehicles, taking into account the State of Charge of the battery. This is because the battery is not large enough to drive during a whole day without recharging. We use a combination of lagrange relaxation and column generation to solve our model. Furthermore, we developed a model for solving the pricing problem in polynomial time, where we are only aware of NP-complete solution methods for this pricing problem. Vehicle scheduling during transition to a more sustainable public transport: We developed models for vehicle scheduling during the transition phase from traditional, diesel vehicles to more sustainable, cleaner vehicles. Typical for this transition phase is the mixture of old and new vehicles that should be used. For a mixture of older, less clean and newer, cleaner vehicles we show that by taking the exhaust characteristics of the different buses in the fleet into account, we can reduce the total exhaust by 15 percent without extra cost. For a mixture of diesel and electric vehicles, we show that taking the different characteristics into account saves up to 4 percent in cost.