Optimal City Street Network Design Under Uncertainty

Abstract

In this article, we address the problem of city street network design - specifically that of using one-way vs. two-way streets - from a different perspective than existing published literature. While at least one author acknowledges that motorist familiarity should be a factor in network design, this has not been empirically modeled. Instead of implicitly assuming motorists travel optimal paths, we explicitly model motorist unfamiliarity and uncertainty with an area. Furthermore, while the published research uses VMT or similar metrics to evaluate network design, we propose the ratio of actual VMT to optimal VMT as a more appropriate metric, with a target ratio of unity. We develop two simple idealized city street grids: one in which all streets are two-way, and a second of perfectly alternating one-way streets. Motorists are simulated traveling on both grids while varying the level of unfamiliarity and uncertainty. For each motorist, the ratio of actual to optimal VMT is measured and recorded. Our results suggest that travel efficiency for visiting motorists unfamiliar with an area will always be highest for one-way street networks. The policy this suggests is that one-way network city street designs should be preferred when there are likely to be a high proportion of motorists who are unfamiliar with the area. This conclusion goes against the prevailing wisdom, since most analysis evaluates network designs based on minimizing VMT, assuming motorists travel optimal paths.

Keywords

• monte carlo simulation
• transportation modeling
• transportation network design