Modeling Driver Behavior in a Straight-line Emergency Situation 2005-01-2700
Setting up a digital driver behavior model represents one major approach to design and validate active safety devices during their development phase. The present contribution introduces our predictive model of driver actions on car controls (steering wheel and pedals). Its core is based on the combined use of neural networks and genetic algorithms. This model was set up using a behavioral database gathered from a test track experiment. This paper details the different stages followed to construct our model. It shows validation results using another database and discusses its effective robustness.