Zustandsschätzung mit chronologisch ungeordneten Sensordaten für die Fahrzeugumfelderfassung
Muntzinger, Marc M.
FacultiesFakultät für Ingenieurwissenschaften und Informatik
Schriftenreihe des Instituts für Mess-, Regel- und Mikrotechnik
LicenseStandard (Fassung vom 01.10.2008)
Future driver assistance systems are based on environment perception and situation analysis that must be reliable and fast. Sensor fusion techniques carry with them the problem of measurements from different sensors arriving at the processing unit out-of-sequence, i.e., the original temporal order of measurements is lost. This poses a problem in usual sensor fusion algorithms, which is normally solved via buffering of measurements to ensure the correct temporal ordering. However, buffering causes severe temporal delays that have to be avoided by all means in time-critical applications. Therefore the presented work analyses algorithms to handle out-of-sequence measurements (OOSM) by directly incorporating them into the state estimation without buffering. The performance, temporal gain and costs of different approaches are evaluated. A detailed complexity analysis of the algorithms is presented. The presented work proposes a novel data association technique as well, which incorporates out-of-sequence measurements. The Joint Probabilistic Data Association (JPDA) is extended in order to associate (possibly delayed) measurements with tracks in a cluttered environment. In addition, a novel risk assessment with incorporating covariance propagation into a real-time pre-crash application is presented. A powerful, yet applicable method for using not only state but also covariance information for triggering actuators is proposed. The performance of the different OOSM algorithms is demonstrated using a test vehicle setup with several radar sensors, where an impact sensor is used to evaluate the estimated time to collision (TTC). Furthermore, the results from the OOSM algorithms are evaluated against reference data from a highly accurate laser scanner. All in all, out-of-sequence algorithms are shown to significantly improve the tracking of vehicles especially in time-critical applications.
Subject HeadingsFahrerassistenzsystem [GND]
Driver assistance systems [LCSH]
Multisensor data fusion [LCSH]
Sensor fusion [LCSH]
Tracking (Engineering) [LCSH]