Application of Constrained Enumeration Approach to Multimodal Choice Set Generation

Collected data often include only information about chosen routes. To gain insight into travelers’ route choice behavior or to predict route shares, one must know the set of alternatives from which travelers have chosen their routes. An alternative approach to choice set generation in mixed multimodal networks is presented. This new algorithm—a run-based, constrained enumeration method that uses branch-and-bound techniques—is suitable for both estimation and prediction. One key characteristic of the algorithm is a set of constraints that reflects observed travel behavior. The proposed algorithm for choice set generation can be applied to a complete multimodal network at once. However, by exploiting knowledge about the structure of multimodal trips, the separate application of the algorithm to partial networks and consecutive concatenation of subroutes into complete door-to-door routes substantially reduce computation times without resulting in incomplete choice sets. This algorithm for choice set generation has been calibrated for and successfully applied to a real-size, mixed multimodal transport network in the Netherlands. A comparison of generated choice sets with reported chosen and known alternatives indicated that the algorithm can generate these alternatives, with high coverage levels as a result. This result clearly indicates that this constrained enumeration approach meets the requirements for choice set generation and thus offers interesting perspectives for route choice analysis and the prediction of route shares. Furthermore, the separate application of the algorithm to partial networks and the consecutive concatenation of subroutes into complete door-to-door trips substantially do not result in incomplete choice sets.