Constructing a network fundamental diagram using a sample of connected vehicles and stationary sensors. (December 2021)

• Record Type: Journal Article
• Authors: Du, Jianhe; Rakha, Hesham A.
• Is Part Of: Transportation research interdisciplinary perspectives. Volume 12(2022)
• Publication Date: 2021-12
• Subjects: Network Fundamental Diagram -- Macroscopic Fundamental Diagram -- Connected Vehicles -- Synthetic O-D Estimation; Transportation -- Periodicals; 388.05
• DOI: 10.1016/j.trip.2021.100471
• Languages: English
• ISSNs: 2590-1982

Abstract

Highlights: Estimating NFD using a small percentage (10%) of probe vehicle and limited locations with traffic count data (15% of the road network with count data available). Highly accurate NFD estimation (RMSE < 0.01). Market penetration rate (MPR) calculated from estimated synthetic OD using limited count data. Tested with real-world network (Los Angeles). Abstract: The Network Fundamental Diagram (NFD can be used to monitor and alleviate traffic congestion. Theoretically, one needs the trajectories of all vehicles in order to accurately estimate the NFD, which is not feasible. One solution is to construct the NFD using a sample of vehicles serving as probes together with an estimated market penetration rate (MPR). In a previous paper, we built a model to first estimate the MPR and then incorporated it in the NFD estimation. That model used data from probe vehicles (trip travel time, origin, destination, and route) and a few traffic count sensors on a hypothetical grid network. The results indicated that sensor locations have a significant impact on the accuracy of the estimated MPR and NFD. In this paper, we extend the model by adding more spatial criteria and test the approach on a real-world calibrated network, namely downtown Los Angeles. We examine the effect of selecting different probe vehicle samples on the model performance. After identifying the spatial rules for selecting sensor locations and the criteria to select probe vehicles, we present a methodology for estimating the NFD producing an acceptable NFD root mean square error. The significance of this study is that first, the model is validated using a real-world calibrated network; second, the data needed for the estimation are highly available, given that with connectivity the location and speed of a sample of vehicles is available; and third, the computational load to select detector locations and probe vehicles is minimum.