Demand estimation and spatio-temporal clustering for urban road networks

Abstract

Urban road networks play a key role in mobility amongst the critical infrastructure of city, which is a strong time-variant system with uncertainty. In this paper, for the purpose of understanding the traffic congestion propagation patterns, demand estimation and spatio-temporal clustering was performed with a case study on the central area of Nanjing, China. Firstly, a four-hour time-dependent origindestination traffic demand is calibrated by utilizing the Adaptive Fine-tuning (AFT) algorithm, and aiming at minimizing the error between microscopic simulated results by SUMO and real-world Radio Frequency Identification (RFID) data. Then, spatio-temporal clustering was performed to illustrate the dynamic feature on congestion propagation by implementing the spectral clustering approach. Results demonstrate that the calibrated dynamic origin-destination matrix can illustrate a good match with RFID data, and the proposed spectral clustering approach is fast and feasible for the partitioning of urban road network.