A System Architecture for Heterogeneous Signal Data Fusion, Integrity Monitoring and Estimation of Signal Quality

Abstract

A large number of today’s automobiles, right down to the compact car segment, are equipped with vehicle dynamics control and driver assistance systems. In general, each one of these functions has been developed with its own dedicated set of sensors, and applied independently of other sensors installed in the same vehicle. As a result, redundant measurements are performed, the advantages of which are currently only utilized in a few cases. The increasing powerfulness of microprocessors and the availability of bus systems in vehicles offer a basis for a central processing of the large quantity of data already available. In this article, criteria are shown for selecting and combining methods for data fusion, integrity monitoring and signal quality estimation from a system design and integration point of view, i.e. which integrity and error detection algorithm fits the requirements for working with signals from cost-effective series sensors and offering useful results for the application together with driving dynamics control and driver assistance systems. Also, the ease of series application and usability in existing system architectures will be considered.The structure of a system architecture for centralized, consistent fusion of random pieces of data is shown and evaluated using an example. The sensors used here are acceleration and yaw rate sensors produced using micro-electro-mechanical system (MEMS) technology combined to an inertial measurement unit (IMU) with 3 degrees of freedom, a single-channel (L1) GPS receiver that issues raw data (pseudo ranges and carrier phase measurement), as well as odometry sensors measuring angle pulses from all four wheels and the steering wheel angle. In addition, an evaluation of the signal quality based on usage of redundancies is shown. In general, many different, proven methods for evaluating the integrity of signals or a sensor fusion system exist. Although first definitions of, and requirements for integrity in the automotive sector exist, no appropriate algorithm concept, and no system architecture has yet been defined, and the existing methods only partly fulfill the requirements. Hence, in this article a top-down approach will be shown, beginning at the requirements of automotive integrity, and leading to the definition of an automotive integrity and accuracy benchmark, as well as a system architecture suited for the integration into existing and future system setups. For this purpose, a description of the signal quality by means of integrity and accuracy is shown, and components for such a description are shown by way of example.