Simultaneous tracking and shape estimation using a multi-layer laserscanner

Environment awareness is an essential prerequisite for the operation of automated vehicles. In particular, reliable knowledge about the motion and geometry of surrounding objects is vital for the inference of safe and comfortable driving strategies. The availability of detailed contour measurements, e.g. from laser scanners, allows for precise object geometry estimates which has benefits for the tracking itself as well as higher-level processing. In this paper, we propose a batch formulation of the object tracking problem which allows to simultaneously estimate object motion and shape as an accumulated 3D point cloud. We show that our method is able to produce highly accurate motion estimates and consistent shape reconstructions from only three layers of scan data.