Route generation algorithms for public transit network design

This study contains demand-based route generation algorithms to solve transit route network design problem (TRNDP) and a model of Adana Urban Public Transportation network. The proposed route generation algorithms are tested on Mandl's Swiss Road network and the four large networks proposed by Mumford and on Adana Urban Public Transportation network. The algorithms use local search methods, Hill Climbing (HC) and Tabu Search (TS) and Evolutionary Algorithm (EA). EA has both single-objective optimization and multi-objective optimization options. Local search methods HC and TS are only employed as Single-Objective optimization because of their structure. The main work presents two new initial route generation algorithms, adding a swap operator in modification solution method and a new crossover operation for EA. Experimental results are evaluated using average travel time (ATT), route length (cost), and zero transfer percentage parameters. In general, better results are obtained on all networks compared to the similar previous work. For example, for the largest artificial network, the best solution found for the first objective has approximately 3.6 % less ATT and 26.72 % less cost compared to previous work. Also, new solutions which have lower cost and ATT than existing bus routes, are proposed for Adana Urban Public Transportation network.