Integrity Based Data Fusion of Redundant Fusion Filters for Vehicle Dynamic State Estimation in Automated Driving

A reliable and highly-available estimation of the vehicle's dynamic state, which is providing integrity information, is essential for automated driving. A possible solution for this challenge is a modular structure consisting of a first level with redundant multi-sensor data fusion filters and a second level processing the first level's outputs. In this work, we propose a concept for the second fusion level performing a data fusion based on the first level's integrity information. We present an implementation of our concept using three fusion filters which input sensor data from a Global Navigation Satellite Signal receiver, an Inertial Measurement Unit and odometry sensors. Outputs are the estimated state including the position, velocity and heading (yaw angle) of the vehicle. This implementation is evaluated using measurements collected in four typical scenarios for automated driving with different environmental conditions especially regarding the satellite reception. In all tested scenarios, the proposed second fusion level fulfills the integrity requirements. The provided solution outperforms the inputs' solutions in the challenging urban scenario with respect to the accuracy while reaching a similar to increased availability depending on the estimated state under consideration.