Multiple-model tracking for the detection of lane change maneuvers

Abstract

Volkswagen research has developed a system for vehicle surround perception which integrates different sensor data of the environment into a combined description by using a single model Kalman tracker. This paper deals with the extension of the tracking system by means of an interacting multiple-model algorithm (IMM) to improve the tracking stability during curves and to detect lane changes of the observed target vehicle. The applied IMM-tracker uses specialized models for lateral and longitudinal motion that are partly affected by curvature estimation. The technique is tested with recorded sequences of measurement data and shows robust tracking and well-fitting classification of the dynamical behavior of the targets.