Intelligent vehicles measure their local environment utilizing surround sensors. Algorithms use this information for driver assistance or vehicle control applications. Insufficient reliability and accuracy of measurements can lead to fatal consequences for occupants as well as other road users. Therefore, sensors must provide correct object information in all driving situations and conditions. This paper focuses on radar sensor object tracking performance in high dynamic driving situations (e.g. skid events). A study of radar sensor accuracy in these situations is presented to emphasize the problem. A non-linear vehicle motion model was modified and used to estimate parameters of the horizontal vehicle motion during the skid event. An extended Kalman filter was designed to track a static target object using pre-processed automotive radar signals. The filter was tested and validated under realistic conditions using a test vehicle equipped with state-of-the-art automotive sensors, which was forced to skid during an object tracking task. Finally, the enhanced observer tracking performance in contrast to the standard radar sensor output is presented and discussed.
Object tracking based on an extended Kalman filter in high dynamic driving situations
2017-10-01
693284 byte
Conference paper
Electronic Resource
English