Estimating Driver Awareness of Crossing Pedestrians While Turning Left Based on Vehicle Behavior Using Bayesian Approach

In this paper, we propose a method for estimating, from the vehicle behavior, a target driver´s awareness of pedestrians that are crossing or are about to cross a pedestrian crosswalk on the driver´s left while the driver is poised to turn left at a target intersection (left-hand traffic is supposed). To construct such an estimation model based on statistical learning, a large number of training examples are required, however, in reality, it is difficult to prepare a large number of examples for a particular driver turning left at a single intersection. In this study, we constructed a model using a Bayesian approach by learning from few examples at the target intersection using a prior distribution of the model parameters derived from large amounts of data gathered at other intersections. We then report the evaluation results using vehicle behavior data obtained from the driving of actual vehicles on public roads.