Recognition of Lane Change Intentions Fusing Features of Driving Situation, Driver Behavior, and Vehicle Movement by Means of Neural Networks

Abstract

The work presented aims at an early and reliable prediction of lane change maneuvers intended by the driver. For that purpose, an artificial neural network is proposed fusing features modeling the environmental situation that influences the formation of intentions, the gaze behavior of the driver preparing an intended maneuver and the movement of the vehicle. The sensor data required are provided by a multisensor setup comprising automotive radar and camera sensors. The whole prediction algorithm was put into practice as a real-time application and was integrated in a test vehicle. With this system, a naturalistic driving study was conducted on urban roads. The naturalistic driving data obtained were finally used for the parametrization of the algorithm by means of machine learning and for the evaluation of the prediction performance of the algorithm, respectively.