Congestion patterns of traffic studied on Nanjing city dual graph

Mapping the streets and street-crossings of the Nanjing city plan onto nodes and edges of a dual graph, we implement the traffic model of vehicles moving towards specified destinations with navigation and queuing. The navigation rules with local search within the next-nearst-neighborhood are applied both with and without dynamical feedback. The traffic is autonomously driven with a parameter which controls traffic density and maximum queue lengths are limited by the actual lengths of streets. The analysis of the simulated traffic load time series indicates three regimes, depending on the traffic density: free flow, regime with temporal congestions, and congested traffic. We determine the probability distributions of the waiting times in queues and travel times of vehicles, as well as the distributions of the local loads on nodes and edges. With the spectral analysis of the weighted matrix of traffic load on the edges of the dual graph, and mapping back to the real space, we identify geographically distinct regions which have different congestion patterns.