Agent-Based Simulation of the Swiss Highway Network
Introduction
This page contains a short abstract, the report, the oral defense slides and screenshots of the master project Agent-Based Simulation of the Swiss Highway Network. The project was made by Laurent Grangier in winter semester 2005-2006 and was supervised by Michael Schumacher and Radu Jurca.
Abstract
Modern highways can become very inefficient when unforeseen events (e.g. accidents), or temporary congestion perturb their normal functioning. Traffic jams or accordions can significantly decrease the total throughput of the highway, and generate supplementary costs (time and fuel) to the drivers. Agent-based micro-simulations can help understand and eliminate such problems by a better management of existing resources.
In this project we developed an agent-based micro-simulation of the Swiss highway network. Individual vehicles are modeled as autonomous agents, and the highway network is split into segments of variable length that can sense and react (e.g. by changing to maximum allowed speed) to the momentary traffic flow. The behavior of vehicles is governed by flexible rules (e.g. for path selection, accelerating, breaking, changing lanes) inspired from reality.
The software implements a centralized architecture, using the Governing Environment paradigm. A first prototype can simulate up to 300.000 vehicles, however, further speed improvements could be obtained by distributing the computation.
Screenshots
