Bayesian Network-based Identification of Driver Lane- changing Intents Using Eye Tracking and Vehicle-based Data

A Bayesian network decision-making method is proposed by combining driver's eye-tracking data and vehicle-based data together to identify driver lane-changing intents. First, experiments are conducted in a driving simulator with eye-tracker device to obtain the data when a subject driver makes lane-changing maneuvers. Second, collected data are analyzed in machine learning method using Bayesian decision-making approach to predict driver's lane-changing intents. Last, to show the benefits of our proposed method, comparison experiments are made between the data fusion way and only using eye tracking data or vehicle-based data. The results show that the Bayesian network with data fusion method performs better than using single information to recognize driver's lane-changing intents. At the same time, thresholds of Lane-changing probability and vehicle-based data as restricting condition choosing work is discussed in order to select the best identification parameter. Topic: Active safety and driver assistance systems, Driver modeling