Abstract
Keywords - Sensor fusion, fault detection, isolation, estimation, diagnosis
Abstract - In recent years an increasing number of electronic systems in modern passenger vehicles can be observed. Especially, systems sensing and influencing the vehicle motion (e.g. GPS based navigation and ABS, ESP respectively) are rather popular. The variety of sensors included in each of these systems offers the possibility of sensor fusion with a potential use of the retrieved information for other systems in the vehicle.
Additionally, direct or analytical sensor redundancy can be used in order to perform a plausibility check of the sensor information. This feature will become crucial especially for employment of sensor information in safety critical applications.
A first realization of such a system will only consider unidirectional usage of fused sensor information, i.e. the information supplying vehicle systems will rely on plausibility checks of the proposed system but not on fused information, which, for example, can be estimated by means of vehicle models. However, for future realizations, one focus of sensor fusion will be to use fused information in order to cut down on sensors and thus to save costs.
A main feature of a central sensor plausibility check is an improvement of fault detection: with additional sensor information the sensitivity of fault detecting algorithms can be improved. Furthermore, the concentration of plausibility check and estimation of non-measured physical information in one system will reduce the overall effort on software and will facilitate the realization of new functions. An important requirement for such a system is a simple development- and application process. The system should be scalable in the sense that new sensors easily can be integrated in the proposed concept.
Introduction of a central sensor plausibility check will be a gradual process. One possible introduction scenario is presented. Nevertheless, the final architecture of a central sensor fusion is still open. Some open problems will be discussed at the end of the paper.