Prediction of driver behavior on a limited sensory setting

We present a system able to predict the future left lane change behavior of the ego-vehicle on highway, depending on detected information about the preceding vehicles. Our system learns the mapping between the current perceived scene (information about the ego-vehicle and the preceding vehicles) and the future driving behavior of the ego-vehicle. The main difference to related contributions comes from the fact that we want to perform behavior prediction with a limited sensory setting. On the one hand, we do not perform driver monitoring, and on the other hand, we only use frontal sensors. However, we show that it is possible to predict accurately the left lane change behavior of the driver. The system is tested on a real world scenario, using 50 minutes of streams recorded on German highways, and the prediction occurs at a horizon of 2s into the future. The quality of the prediction is measured using frame-based as well as event-based evaluation with a cross-validation approach. We compare a system with only front sensing (vision and radar detection) with a system including a simulated rear sensing.