A Multisensor Multiobject Tracking System for an Autonomous Vehicle Driving in an Urban Environment

This paper presents the tracking system of Boss, Carnegie Mellon University's winning entry in the DARPA Urban Challenge in 2007. We present the key challenges for implementing the tracking system, the design principles that guided its implementation, the software architecture of the tracking system and the sensor setup used by Boss. The system has been shown to work robustly in many different situations, including intersection handling, distance keeping or driving on open parking lots. The design principles and tracking architecture are formulated in a general way and may be used for the development of driver assistance systems which have to deal with the same situations.