Todays Advanced Driver Assistance Systems (ADAS) change from single, independent functionalities to integrated multipurpose applications. The current method to reach these multipurpose applications is the combination of the existing independent applications. But the independent applications do not provide the universality to change basic assumptions or parameters, always requiring a redesign for integrating functionalities. Furthermore, the applications are designed for a clearly defined scenario as well as environment, making an all situation support merely impossible. In addition, the complexity and performance requirements on the system level are not manageable if isolated functions and modules are simply put together. Hence, another way has to be taken in order to reach the goal of an accident avoiding vehicle that can support the driver with superior abilities in all situations. To this end, a number of novel algorithms and approaches are presented in this thesis to overcome existing limitations. More specifically, the emphasis is placed on genericity, parameterization, and simple extensibility of the algorithms to emphasis the knowledge exchange in the system context. Additionally, the integration of all algorithms in a biological motivated system design proves the benefits of a generic system architecture. In general, the robust identification of the path requires diverse knowledge of many environmental characteristics. In particular, the path is the area in front of the vehicle which was identified to satisfy the current task (e.g. identification of the ego-lane, extraction of the stop position). There fore, the path depends on a number of different features (as, e.g., the lanes, other vehicles, symbolic information, driving rules). All of these information sources are needed to robustly provide the path task dependently in all situations and also all environments. Therefore, the first aspect of this thesis handles the classification of the current driving scene in order to be able to parameterize the system depending on its current surrounding. A reliable classification of the scene is presented requiring a single image only, which is based on a computational model mimicking the characteristics of the human visual pathway. Hence, the system gets the ability to adapt its processing at run time to the environment.
Robust path identification for driver assistance
2011
172 Seiten, 64 Bilder, 15 Tabellen, 148 Quellen
Theses
English
Robust path identification for driver assistance
TIBKAT | 2011
|DRIVER ASSISTANCE SYSTEM WITH PATH CLEARANCE DETERMINATION
European Patent Office | 2016
|Driver assistance system with path clearance determination
European Patent Office | 2018
|DRIVER ASSISTANCE DEVICE, DRIVER ASSISTANCE SYSTEM, AND DRIVER ASSISTANCE METHOD
European Patent Office | 2025
|Driver Assistance Device, Driver Assistance Method, and Driver Assistance System
European Patent Office | 2021
|