Originally developed to generate behavior in autonomous robots, attractor dynamics encode basic behavioral tendencies with meaningful parameters that support optimizations through direct policy search. We combined attractor dynamics with a powerful evolutionary algorithm to arrive at driver models that capture behavioral patterns of real human drivers who employ various driving styles. These models support the development of driver assistance systems in several ways.
Index Terms:
driver models, driver assistance systems, dynamical systems, evolutionary algorithms
Citation:
Antonio Pellecchia, Christian Igel, Johann Edelbrunner, Gregor Sch?ner, "Making Driver Modeling Attractive," IEEE Intelligent Systems, vol. 20, no. 2, pp. 8-12, Mar./Apr. 2005, doi:10.1109/MIS.2005.31