Uncertainty Aware Sensor Fusion for a GNSS-based Collision Avoidance System

The performance that can be achieved by safety critical systems which are based on
information on their absolute position inherently depends on the accuracy and the
update rate that is provided by the navigation system. A Collision Avoidance System
(CAS) that is intended to prevent collisions between vehicles by an autonomous braking
resp. evasion maneuver is such a safety critical system and will be subject of
consideration in this paper. In this context, the employed sensor fusion algorithm that
provides a navigation solution with high update rate based on an inertial navigation
system (INS) as well as on a Global Navigation Satellite System (GNSS) will be
presented. Furthermore, a method to assess the actual accuracy provided by GNSS resp.
sensor fusion will be introduced. Subsequently, the paper shows how this information
can be incorporated into sensor fusion. Obviously, the feasibility of CAS inherently
depends on the accuracy that can be obtained from GNSS. Therefore, CAS will be
developed and tested in a terrestrial Galileo test and development environment (GATE)
near Aachen (Germany) where a positional accuracy of about 0.8 meters can be
achieved. Finally, first simulation results based on measurement data will be presented
to evaluate the sensor fusion algorithm as well as the method to assess the actual
navigation accuracy.