On the important features for a well-shaped reduced network MFD estimation during network loading and recovery.

Monitoring traffic dynamics at both link and network levels is vital for effective traffic management and control in urban road networks. Traffic monitoring at the network level, though, is an extremely challenging task due to the amount of data (and consequently sensors) needed to represent a good coverage of the network traffic dynamics. This has motivated researchers to investigate where to locate sensors (e.g. loop detectors) within a network to capture the so-called Reduced Network Macroscopic Fundamental Diagram (NMFD). The existing approaches for reduced NMFD estimation have presented limitations, because of the reliance on solely traffic flow characteristics or single topological features. In response, in this paper, a framework has been proposed to take into account the simultaneous impact of topological and traffic flow features for the estimation of reduced NMFD. Contrary to existing works in literature, which have mainly focused on the loading period, this work considers both loading and unloading periods. The presented framework deploys regression models to identify the set of critical links within the network. The efficiency of the presented method has been validated under both single and multi-reservoir settings and has been compared to two methods that solely consider traffic characteristics. Our findings reveal that the proposed method demonstrates superior performance, particularly when dealing with low coverage rates of loop detectors (e.g. 1%). Moreover, results show that during the recovery period, traffic characteristics (e.g. queue distribution) play a more important role in estimating the NMFD. On the other side, a more significant impact has been observed for the topological features (e.g. speed limit distribution) during the loading period. • Critical links identification through historical FCD and network topological info. • Reduced operational network MFD estimation during loading and unloading periods. • Feature importance analysis using Random Forest Regression (RFR). • Investigating the combined impact of topological and traffic characteristics on NMFD. [ABSTRACT FROM AUTHOR]