The impact of complexity in the built environment on vehicular routing behavior: Insights from an empirical study of taxi mobility in Beijing, China
Chaogui Kang, Zheren Liu
TL;DR
The paper tackles the mismatch between traditional shortest-path routing and real-world driving in complex urban environments by introducing an anchor-based cross-nested logit model that incorporates complexity in the built environment. It leverages large-scale Beijing taxi mobility data and rich ambient features to quantify how road-network topology, land use, and visibility influence route choice, beyond travel time and distance. The study demonstrates a cumulative gain of about 12% in explanatory power over conventional models, with complexity contributing roughly 5% and the anchor structure about 7–8%, while revealing distinct patterns across time, trip distance, and occupancy. These findings have practical implications for urban transport interventions and planning, supporting the use of digital twins and reinforcement-learning approaches to optimize routing and zoning in real-time.
Abstract
The modeling of disaggregated vehicular mobility and its associations with the ambient urban built environment is essential for developing operative transport intervention and urban optimization plans. However, established vehicular route choice models failed to fully consider the bounded behavioral rationality and the complex characteristics of the urban built environment affecting drivers' route choice preference. Therefore, the spatio-temporal characteristics of vehicular mobility patterns were not fully explained, which limited the granular implementation of relevant transport interventions. To address this limitation, we proposed a vehicular route choice model that mimics the anchoring effect and the exposure preference while driving. The proposed model enables us to quantitatively examine the impact of the built environment on vehicular routing behavior, which has been largely neglected in previous studies. Results show that the proposed model performs 12% better than the conventional vehicular route choice model based on the shortest path principle. Our empirical analysis of taxi drivers' routing behavior patterns in Beijing, China uncovers that drivers are inclined to choose routes with shorter time duration and with less loss at traversal intersections. Counterintuitively, we also found that drivers heavily rely on circuitous ring roads and expressways to deliver passengers, which are unexpectedly longer than the shortest paths. Moreover, characteristics of the urban built environment including road eccentricity, centrality, average road length, land use diversity, sky visibility, and building coverage can affect drivers' route choice behaviors, accounting for about 5% of the increase in the proposed model's performance. We also refine the above explorations according to the modeling results of trips that differ in departure time, travel distance, and occupation status.