Advancing Macroscopic Fundamental Diagram Estimation: Addressing Heterogeneity in Urban Road Networks

The fundamental diagram (FD) relates the traffic flow, density, and speed of a single link and is used to improve the traffic conditions at the link level. However, in urban road networks, where there are numerous interconnections between links, optimization at the individual link level may negatively affect the overall traffic conditions across the network. The development of the Macroscopic Fundamental Diagram (MFD) addresses this issue by demonstrating the relationship between the average flow and density at the network level, leading to the development of various MFD-based strategies aimed at optimizing traffic conditions and alleviating congestion and delays in urban road networks. Nevertheless, the traffic conditions in urban road networks are unevenly distributed, which negatively impacts the accuracy of MFD estimation. Therefore, this dissertation aims to increase the accuracy of MFD estimation incorporating heterogeneous distribution of traffic conditions. Traffic data is collected from various links and aggregated to estimate the MFD for the entire road network. Given the time and resource constraints associated with data collection, gathering data from all road network links is often impractical. As a result, traffic data from a selected sample of links is utilized to estimate the MFD. The traffic states of links are dynamic and exhibit variability in space and time, often showing an inhomogeneous distribution. This variability can significantly influence the representa- tiveness of the sampled links, potentially leading to inaccurate MFD estimates. There is a lack of systematic methodology for selecting representative links for MFD estimation. To address this issue, the first study developed a method that effectively incorporates the heterogeneity of traffic states in selecting sample links, which substantially improved the accuracy of MFD estimation. The route choice behavior of drivers is the primary contributor to the heterogeneity observed in traffic state distribution. Links that drivers prefer typically experience higher demand, often reaching their capacity. Such links significantly influence the average traffic state of the network and are thus critical to MFD estimation. Traditionally, link length has been used as a parameter for aggregating traffic data for MFD estimation. This approach, however, fails to account for the criticality of links, treating critical and non-critical links as equivalent. Consequently, reliance solely on link length for data aggregation can lead to underestimating critical links alongside non-critical ones, thereby impacting the accuracy of MFD estimation. The third study introduced a method for assigning weights to each link based on their criticality. This ensures that critical links maintain their significance and contribute meaningfully to MFD estimation. The enhanced role of critical links in MFD estimation accentuates the shortcomings inherent in link length-based data aggregation. In summary, the findings of this dissertation offer valuable insights into previously overlooked areas of research regarding the incorporation of heterogeneity in MFD estimation, representing a significant advancement towards achieving accurate MFD estimation, which is essential for the effective and efficient management of urban road networks.