Information fusion for cooperative vehicles

Abstract

The cooperative perception by an inter-vehicle network promises a multitude of improvements for advanced driver assistance systems. First, for the simultaneous estimation of the vehicle ego pose and the road network infrastructure, we propose a local, highly model-based fusion architecture for digital map and video information. Second, to extend the vehicle's field of view, we include remote information concerning the state of each participating vehicle and its object detections. For the object-level fusion of these detections a centralized tracking process by Kalman filtering is employed. The algorithms are evaluated in simulated vehicle network scenarios which are based on real sensor data.