A stochastic multiple area approach for public transport network design

This paper proposes a new method for public transport (PT) network optimization that considers the entire PT chain from door-to-door and allows flexible line alignments. The approach optimizes different speed levels, e.g., bus and tram, regional train, etc., sequentially, starting with the fastest service. To reduce computing times, each speed level for the geographic area under consideration is divided into several planning areas. If computing times for planning areas are short enough, computing times for network design in entire areas can be handled as well since they are only linearly dependent on the total size of the area based on the suggested approach. For each planning area, the approach uses a network reduction process that requires comparatively few network evaluations. The network reduction process starts with a network of the shortest lines. Then, lines are deleted, merged or shortened sequentially using the ant colony optimization algorithm. A genetic algorithm simultaneously optimizes service frequencies and vehicle sizes. During the network reduction process, total operating and travel time costs are minimized. For network evaluations, a headway-based stochastic multiple route assignment is used. The reduction approach was compared to existing approaches by applying it to Mandl's Swiss benchmark problem. Based on the comparison, it can be stated that the approach developed shows promising results in terms of optimized fleet size and user costs. The multiple area approach and the reduction process were tested together in a larger case study in Winterthur, Switzerland.