.png){kind=icon}
.png){kind=icon}
Return
Traffic estimation in unobserved network locations using data-driven macroscopic models
.png){kind=icon}
PRE
AI
DOI：10.1080/23249935.2025.2511820.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper presents the Macroscopic Traffic Estimator (MaTE), a model that leverages computational graphs, neural networks, macroscopic flow theory and multi-source spatiotemporal data from traffic counters and probe vehicles to estimate traffic flow and travel time in roads without direct measurements. This estimation problem is critical in applications where the sensor coverage is low and the planned interventions have network-wide impacts. Grounded in macroscopic flow theory, MaTE has fully interpretable parameters, with network flow adhering to conservation constraints and travel time increasing monotonically with link flow. Using logit-based stochastic traffic assignment as the principle for routing flow behaviour makes the model fully differentiable with respect to the model parameters, which facilitates the application of automatic differentiation tools and computational graphs to learn parameters from vast amounts of spatiotemporal data. Neural networks and polynomial kernel functions are also incorporated to capture link flow interactions and enrich the mapping of traffic flows into travel times.MaTEalso adds destination choice and trip generation layers, allowing it to be effectively trained without requiring access to historical origin-destination (O-D) matrices. Experiments on synthetic data show that the model can accurately estimate travel time and traffic flow in out-of-sample links. Results obtained using real-world multi-source data from a large-scale transportation network in Fresno, CA suggest thatMaTEoutperforms data-driven benchmarks, especially in travel time estimation. The estimated parameters ofMaTEare also informative about the spatio-temporal changes in travel demand and level of service in the transportation network.
Keywords:
Computational graphs
traffic flow estimation
logit-based stochastic traffic assignment
origin-destination demand estimation
large-scale network modelling
Journal
IF:
3.1
Papers:
927
Citations:
2.2K