Agent-based modelling has proven as an effective approach to analyze, reconstruct and predict systems with many individual entities. Typical use cases include segregation processes, predator/prey models or supply chain optimization.
Most of these models work within the framework of a “patch-dominated” world where agents move within a world consisting of square-cells named patches. The development of network specific extensions for simulation software like the “network” and “gis” extensions of NetLogo have allowed to apply agent-based theories to spatial data. This allows to move from a patch-, to a network concept. However, there are still some shortcomings that are not satisfyingly solved within the NetLogo software that may hinder an efficient and realistic simulation of agent movements along spatial networks.
This work evaluates the role of scale and time in agent-based traffic models and offers suggested implementations to allow for easier and more realistic parameterization and runtime analysis of traffic models. It will also discuss performance issues with the shortest path calculation in NetLogo and provide a possible solution to increase the simulation speed.
The model and analysis derived in this project are based on a polyline shapefile that contains the street network of Salzburg city.
The work evolved into a GIS Development project rather than pure analysis. An entire new codebase for the NetLogo platform has been developed in order to cater the specific needs of traffic simulations. It considers time and space dimensions and offers a new performant routing algorithm.
Mayrhofer C. (2015): Performance, Scale & Time in Agent-based Traffic Modelling with NetLogo, In: Car, A., Griesebner, G., Jekel, T. & Strobl, J. eds. GI Forum. Berlin: Wichman Verlag, 567-570
Categories: GIS Analysis / Development
Start: Dec, 2014
End: Jul, 2015