A Dijkstra-Based Framework for Analyzing Urban Mobility from CDR-Derived OD Matrices: A Case Study of RMBH, Brazil

Resumo

Urban mobility planning in large metropolitan areas increasingly depends on high-resolution, dynamic data sources. In this context, mobile phone data, particularly Call Detail Records (CDRs), offer promising alternatives to traditional surveys for estimating origin–destination (OD) matrices. This study presents the development of a computational framework that integrates a CDR-derived OD matrix with graph-based algorithms to explore mobility dynamics in the Metropolitan Region of Belo Horizonte (RMBH), Brazil. The OD matrix used was generated from anonymized mobile phone records collected during a large-scale urban mobility study conducted in 2019 and 2021, made available by the State Infrastructure Secretariat of Minas Gerais (SEINFRA).The matrix covers 393 spatial zones, each represented by its geometric centroid. These zones form the nodes of an undirected, weighted graph, in which edges are defined by geographical adjacency and weighted by the distance between centroids, calculated using the Haversine formula to account for Earth’s curvature. To simulate individual or aggregated travel paths, a shortest-path algorithm based on Dijkstra’s method is implemented, allowing efficient computation of minimal-distance routes between any two zones. Although the current model uses geodesic distances, the framework can later incorporate additional weighting factors, such as real travel times or congestion.The framework enables the simulation of several aspects of urban mobility, including the reconstruction of typical travel flows, the identification of high-demand corridors, and the detection of structurally central zones within the network. By systematically applying Dijkstra’s algorithm to the full OD matrix, it becomes possible to visualize and quantify overlapping routes, highlight pressure points, and assess the spatial distribution of cumulative travel load.By combining real-world OD data with classic computational methods in graph theory, this research provides a flexible and scalable platform for urban mobility modeling. Future developments may include the integration of stochastic elements, such as Markov chains, and the simulation of infrastructure scenarios or adaptive traffic control strategies. The approach is reproducible and adaptable, and its application to RMBH may serve as a prototype for other metropolitan regions facing mobility challenges in data-scarce environments.