Modeling and Predicting Driver Behavior in ACT-R

wareness, deciding when to initiate and terminate maneuvers, as well as manage other tasks that are not immediately relevant to the primary driving task (e.g. enjoying the environment or talking on a cell phone). These are characterized by specifying the important input information as well as the various payoff structures for the tasks involved and letting the ACT-R architecture constrain the interaction between these cognitive processes. We compare the predictions of the model to data collected in a highway-driving experiment in our fixed-based simulator. Eye movements were monitored in our experiments and used as a rational surrogate for attention shifts. Results show that the model is capable of predicting attentional demands and reproducing drivers' attention shifts before and during lane changes as well as during the various phases of curve negotiation. The second primary branch of work is the a priori prediction of driver distraction due to secondary in-vehicle devices. This w