Prediction of driver's stop or go at yellow traffic signal from vehicle behavior

Many studies in the field of intelligent vehicles aim at preventing accidents or improving the traffic performance of vehicles. For realizing more effective advanced driver assistance systems estimating the driver's intention is important. This paper addresses the problem of estimating the driver's intent to stop or pass through when a signal turns yellow. This problem is addressed by proposing two different methods. One is for predicting whether the driver is going to stop when the signal changes to yellow before the driver actually determines or executes it, and the other is for estimating whether the driver has made the decision to stop after the driver has started to implement that decision. The two proposed methods are evaluated by using driving behavior data collected from actual roads to demonstrate their performance and characteristics.