Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 15: Poster Session
SOE 15.19: Poster
Mittwoch, 28. März 2012, 16:50–17:50, Poster F
Fluctuation Analysis of Traffic Data: Comparison of Numerical Simulations, Driving-Simulator Experiments, and Real-World Data — •Mirko Kämpf1, Jan W. Kantelhardt1, Matthew Fullerton2, and Fritz Busch2 — 1Institut f. Physik, Martin-Luther Universität Halle-Wittenberg, Halle (Saale), Germany — 2Lehrstuhl für Verkehrstechnik, TU München, Munich, Germany
The understanding of dynamical properties of traffic flow is essential for distinguishing traffic states or patterns and designing optimization rules and control systems. Here, we employ Detrended Fluctuation Analysis (DFA) to check if time series of velocity and density generated by a common traffic simulation model show the same scaling behaviour as real world data. In particular, we analyse average-level and individual-level speeds and densities simulated by a (microscopic) Wiedemann traffic model combined with a VISSIM lane changing model on a straight road with one or two lanes and no exits. The same scaling analysis is applied to data obtained from driving simulator experiments with groups of four drivers and to measured data from a multi-lane highway near Madrid. Our results show that the model describes the long-range interactions (regarding large groups of cars) fairly well. For small groups of cars, however, we find strong long-term persistent anti-correlations in the simulated velocity data, which are not observed in experimental traffic simulator data and real world data. This way, DFA has reviled that short-range interactions between cars are probably too strong in the current model.