A greedy-navigator approach to navigable city plans
- 187 Downloads
We use a set of four theoretical navigability indices for street maps to investigate the shape of the resulting street networks, if they are grown by optimizing these indices. The indices compare the performance of simulated navigators (having a partial information about the surroundings, like humans in many real situations) to the performance of optimally navigating individuals. We show that our simple greedy shortcut construction strategy generates the emerging structures that are different from real road network, but not inconceivable. The resulting city plans, for all navigation indices, share common qualitative properties such as the tendency for triangular blocks to appear, while the more quantitative features, such as degree distributions and clustering, are characteristically different depending on the type of metrics and routing strategies. We show that it is the type of metrics used which determines the overall shapes characterized by structural heterogeneity, but the routing schemes contribute to more subtle details of locality, which is more emphasized in case of unrestricted connections when the edge crossing is allowed.
KeywordsEuropean Physical Journal Special Topic Degree Distribution Minimum Span Tree City Plan Connection Probability
Unable to display preview. Download preview PDF.
- 2.A.K. Hartmann, M. Weigt, Phase Transitions in Combinatorial Optimization Problems (Wiley-VCH, 2001)Google Scholar
- 4.R. Thomas, S. Donikian, Spatial Cognition V, LNAI 4387 (2007), p. 421Google Scholar
- 13.W. Liu, A. Zeng, Y. Zhou,e-print [arXiv:1112.0241]Google Scholar
- 16.This rule is applied since we consider the road-like structure, where each vertex corresponds to the junctions of roads. Therefore, a new crossing point generated by a new edge can be considered as a new vertex, which effectively changes the system size, so we intentionally avoid this situationGoogle Scholar
- 19.Note that there are two slightly different kinds of clustering coefficients, i.e., the one averaged over the individual vertices’ clustering coefficient which is used for C(k) in this work, and the other taking the ratio of numbers of triangles to the numbers of vertex triads used for C△Google Scholar