Abstract:
A new algorithm that generated a set of paths between a pair of origin–destination nodes in a transportation network for the purpose of generating a measure of accessibility on the level of route choice was designed, developed, and tested. The proposed algorithm incorporated the well-known issue of path overlap in the process of generating the path choice set. This algorithm fit naturally into the class of iterative penalty-based Kth-shortest-path algorithms; in this class the link penalty terms are designed to reflect the amount of overlap between the paths already generated. With the proposed algorithm, paths were generated in order of decreasing utility and corrected by a path size correction factor; it was thus highly efficient in the sense that a comparatively small number of paths could result in a broad spectrum of desirable choices. The algorithm was developed in response to the Valencia paradox, which arose from using logsums from the existing algorithm for choice set generation as a route-level accessibility measure for the bicycle network in San Francisco, California. The Valencia paradox occurs when an accessibility measure decreases following an improvement to actual network accessibility. A detailed case study demonstrated the effectiveness of the proposed algorithm in minimizing this kind of paradoxical result and generating a route-level accessibility measure suitable for making fine-grained planning decisions.
