Drift and diffusion based models of driver behavior
Institute of Physics, Carl-von-Ossietzky University, Department of Physics, 26111 Oldenburg, Germany
2 Institut für Verkehrsführung und Fahrzeugsteuerung, DLR Braunschweig, Lilienthalplatz 7, 38108 Braunschweig, Germany
3 OFFIS, R&D division Safety Critical Embedded Systems, Escherweg 2, 26121 Oldenburg, Germany
Revised: 1 March 2010
Published online: 4 June 2010
Based on experimental data taken from a driving simulator, we present different ways of how a stochastic model of the driver's behavior can be estimated directly from the data. The aim is to generate a drift and diffusion based stochastic model, which is equivalent to a potential based model. Depending on the data, different strategies are presented. Even in the case of non-equidistant or transient data, the method will be used to estimate appropriate models in one or higher dimensions.
© EDP Sciences, Società Italiana di Fisica, Springer-Verlag, 2010