Development, Calibration, and Validation of a Large-Scale Traffic Simulation Model: Belgium Road Network

Mehrabani, Behzad Bamdad; Sgambi, Luca; Maerivoet, Sven; Snelder, Maaike · 2023 · Crossref

DOI: 10.52825/scp.v4i.199

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of developing large-scale traffic simulation models, specifically focusing on the Belgian road network. Traditional travel demand models, such as the classical four-step model, require extensive and costly household survey data, which is often unavailable for large-scale applications. To overcome this, the authors propose a "probabilistic travel demand model" that relies on readily available aggregate data, including city populations, inter-city distances, yearly passenger-kilometers traveled, and yearly truck trips. The study details the development, calibration, and validation of a mesoscopic traffic simulation using the SUMO software, aiming to provide a robust tool for transportation planners to test various scenarios, such as the impact of autonomous vehicles or abnormal conditions. The methodology involves constructing a network based on OpenStreetMap data, limited to highways and provincial/regional roads, with 60 cities selected as trip centroids. The probabilistic demand model calculates daily trip volumes (approximately 3.3 million trips) and assigns origins and destinations using weighted probability distributions based on population and distance, calibrated by a parameter $\lambda$. These hourly Origin-Destination matrices are input into SUMO, where Dynamic User Equilibrium (DUE) assignment is performed. Calibration was conducted in two stages: adjusting traffic flow parameters using established queuing model values and tuning DUE parameters (such as the routing algorithm and route choice model) to match real-world traffic counts from 50 detectors. Validation was performed by comparing simulated travel times against real-world data extracted from the Google Maps Distance Matrix API for 3,600 Origin-Destination pairs during the morning peak hour. The results demonstrate that the calibrated model accurately represents real-world traffic conditions. Calibration of the dynamic traffic assignment yielded an average GEH statistic of 4.9, with 78% of observations falling below the acceptable threshold of 7.5. The optimal calibration parameters included a $\lambda$ value of 0.25 for the demand model and the use of Dijkstra’s algorithm with a Logit route choice model over ten iterations. Validation showed a strong correlation between simulated and real travel times, with an $R^2$ value of 0.93. The model correctly identified congestion locations, reporting an average vehicle speed of 61 km/h and an average travel time of 79.2 minutes during the peak hour. The significance of this work lies in providing a viable alternative to data-intensive classical models for large-scale network simulation. By achieving high accuracy with minimal input data, the proposed framework enables researchers and policymakers to efficiently simulate and analyze traffic scenarios at a national level. The authors conclude that this model is a superior representation of reality and recommend future studies to expand the probabilistic demand model to other regions and incorporate international freight traffic for greater precision.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• mental model of traffic

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model