Designing Rapid Transit Networks from the Results of a Survey

Abstract

In this work we present a two-stage approach for designing rapid transit networks. It is based
on another approach that we described elsewhere. In the first stage, the stations and links to be
constructed are selected by solving an integer linear programming model that maximizes an
estimation of the number of trips through the rapid transit network. In the second stage, a set of
lines is generated by utilizing a greedy heuristic procedure that, taking into consideration the
transfers that should be made by the users to arrive at their destinations, attempts to maximize
a more accurate estimation for the number of trips. This new estimation is done by means
of a modification of the well-known Floyd-Warshall algorithm. The main contributions are
a novel way of computing the expected number of trips by making use of the results from a
survey amongst the potential users of the rapid transit network, as well as the contemplation
of the possibility of linking certain pairs of station locations by more than one line. Some
computational experiments on several randomly generated instances are also reported.