Resumen
We propose an optimization framework for urban bicycle network design. The model takes into account interests of the users (who travel along shortest paths) and the planners (available budget). An underlying network composed by street segments suitable to build cycling infrastructure is taken as input. Each network link has construction and user cost, both proportional to the distance. A network link without cycling infrastructure which is part of a path followed by users, has a larger user cost. A multi-commodity network flow mixed-integer mathematical program is proposed and applied to small-sized problem instances to validate the model. The formulation considers the discontinuities of the bicycle network, i.e. the users? paths which include segments without cycling infrastructure. Sensitivity analysis are performed with respect to budget levels and to penalization of user?s cost in links without cycling infrastructure. A metaheuristic is proposed to handle large-sized instances. As an additional feature (difficult to formulate in the exact model), the metaheuristic also minimizes the total number of discontinuities by including them into the objective function. The accuracy of the metaheuristic is estimated by comparing with exact results when possible. The methodology is tested using data from the city of Montevideo, Uruguay, including a large-sized underlying street network and origin-destination trips estimated from a household survey. Computational results are obtained with and without minimization of discontinuities, and they are compared with the current bicycle network of the city.