[Paper Review] Building the road network for city-scale active transport simulation models

Afshin Jafari, Alan Both+3 more|arXiv (Cornell University)|Apr 7, 2021

Urban Transport and Accessibility44 references31 citations

TL;DR

This paper presents a flexible, open-source algorithm that generates city-scale road networks from OpenStreetMap (OSM) data for active transport simulation, balancing accuracy and performance by capturing detailed pedestrian and cycling infrastructure while simplifying complex geometries. The method achieves comparable routing accuracy to full OSM conversions but runs over twice as fast, offering an efficient, scalable solution for modeling walking and cycling in urban simulations.

ABSTRACT

In this paper, we introduce and test our algorithm to create a road network representation for city-scale active transportation simulation models. The algorithm relies on open and universal data to ensure applicability for different cities around the world. In addition to the major roads, their geometries and the road attributes typically used in transport modelling (e.g., speed limit, number of lanes, permitted travel modes), the algorithm also captures minor roads usually favoured by pedestrians and cyclists, along with road attributes such as bicycle-specific infrastructure, traffic signals, and road gradient. Furthermore, it simplifies the network's complex geometries and merges parallel roads if applicable to make it suitable for large-scale simulations. To examine the utility and performance of the algorithm, we used it to create a network representation for Greater Melbourne, Australia and compared the output with a network created using an existing transport simulation toolkit along with another network from an existing city-scale transport model from the Victorian government. Through simulation experiments with these networks, we illustrated that our algorithm achieves a very good balance between simulation accuracy and run-time. For routed trips on our network for walking and cycling it is of comparable accuracy to the common network conversion tools in terms of travel distances of the shortest paths while being more than two times faster when used for simulating different sample sizes. Therefore, our algorithm offers a flexible solution for building accurate and efficient road networks for city-scale active transport models for different cities around the world.

Motivation & Objective

• To develop a general-purpose, open-source algorithm for creating simulation-ready road networks from OSM data that support active transport modes.
• To address the lack of detailed, yet computationally efficient, road network representations for walking and cycling in city-scale transport simulations.
• To balance network detail (for accuracy in active transport modeling) with simplification (for faster simulation run-time).
• To provide a tool adaptable to different cities by relying on open, universal OSM data and configurable parameters.
• To evaluate the algorithm’s performance and accuracy against existing networks, including a full OSM conversion and a major-roads-only network.

Proposed method

• The algorithm extracts road network geometries and attributes (e.g., speed limit, lanes, bicycle infrastructure, gradient) from OSM using context-specific tagging filters based on regional guidelines and OSM TagInfo.
• It simplifies complex geometries such as roundabouts and intersections by reducing node count, using a configurable minimum link length (default 20m) to control detail vs. performance.
• Elevation data is incorporated to support nuanced cycling behavior modeling, including gradient-based routing.
• Public transport infrastructure is integrated into the network to support multi-modal trip modeling.
• A network cleaning step ensures all nodes are reachable, making the output simulation-ready.
• The algorithm is implemented in R with supporting scripts in SQL and Bash, and is publicly available on GitHub.

Experimental results

Research questions

• RQ1 How does the accuracy of shortest-path distances on our network compare to a full OSM conversion and a major-roads-only network for walking and cycling?
• RQ2 How does the simulation runtime of our network compare to existing network representations across varying sample sizes?
• RQ3 To what extent does network simplification affect the accuracy of active transport simulations while improving computational efficiency?
• RQ4 Can the algorithm produce a balanced network that maintains sufficient detail for active transport modeling without incurring excessive simulation overhead?

Key findings

• The algorithm-generated network (SON) achieved comparable shortest-path accuracy to the full OSM conversion (MGN), with average travel distance deviations of less than 10% for all modes.
• For simulated trips, SON was more than twice as fast as MGN across different sample sizes, significantly improving simulation runtime efficiency.
• The major-roads-only network (IVN) was approximately four times faster than MGN but showed significant deviations (over 10%) in walking and cycling path distances, reducing accuracy for active modes.
• SON provided a balanced trade-off between accuracy and performance, outperforming both IVN (in accuracy) and MGN (in speed) for active transport modeling.
• The algorithm’s flexibility allows users to adjust simplification parameters (e.g., minimum link length) to tune the network for specific needs, such as higher accuracy or faster simulation.
• The method demonstrated strong applicability across cities by relying on open, universal OSM data, though tagging inconsistencies across regions remain a challenge for full automation.