Comparing Meta-Heuristic Algorithms for Transit Network Design

This work presents a comparative study of the performance of different meta-heuristic algorithms used within a simulation-based optimisation model that is used to design public transit networks. Simulation-based optimisation focuses on solving decision-based problems by integrating optimisation and simulation. The model combines multi-objective meta-heuristic optimisation algorithm with an activity-based travel simulation to design public transport networks. The simulation evaluates different network solutions, while the meta-heuristic framework identifies optimal ones. The model is robust, thus allowing for integrating different meta-heuristic algorithms, thereby making it feasible to compare the performance of multiple algorithms relative to the network design problem. Four different algorithms, the non-dominated sorting algorithm-II (NSGA-II), strength Pareto evolutionary algorithm (SPEA2), non-dominated sorting algorithm-III (NSGA-III) and the indicator-based evolutionary algorithm (IBEA), are then compared against each other within the simulation-based network design framework using four multi-objective quality indicators including hypervolume, generational distance and maximum Pareto front error. The obtained results show that NSGA-II yields better results than the other algorithms.