Abstract
This paper presents a technique to estimate a switching condition of the driver between primitive tasks which compose
a part of driving scenarios. Overtaking behavior of the driver is focused and measured in the virtual expressway
implemented on a driving simulator. Measured overtaking behavior is segmented into several primitive tasks, and
the decision making models of the driver to switch the primitive tasks are constructed by logistic regression models.
In addition, many candidate variables which can be measured in the driving environment are tested by a chi squared
test if the variable plays an important role as the explanatory variable in the model. A quantitative interpretation of
the overtaking behavior is provided by considering the obtained model structure and the weights of corresponding
explanatory variables.
