Sensor Performance Characteristic Pre-filter using Multiple Model Estimation for Navigation

Yunqian Ma and Shrikant Rao

Abstract:	This paper considers navigation systems that incorporate extrinsic sensors like cameras and lidars typically used in GNSS denied environments. Most of these systems execute odometry or SLAM algorithms using matched features extracted from the sensor data at each time step along with the assumption of a static environment. To ensure features corresponding to static parts of the environment are used for ego motion estimation, several methods of outlier rejection have been previously suggested. However, in real-life applications, there arise situations where simple outlier rejection is inadequate. One such example is the case of occlusion wherein moving objects cover a large part of the field of view of the sensor. We propose to address this problem using the concept of multi-model estimation in a pre-filter of the navigation system. This paper explains the need and use of multiple model motion estimation in pre-processing of extrinsic sensor data. An illustrative example with an outdoor dataset for camera sensor is provided. A discussion of the overall scheme for incorporation of the pre-filter into the existing Kalman filter based navigation processing is also included.
Published in:	Proceedings of the 2013 International Technical Meeting of The Institute of Navigation January 29 - 27, 2013 Catamaran Resort Hotel San Diego, California
Pages:	670 - 674
Cite this article:	Ma, Yunqian, Rao, Shrikant, "Sensor Performance Characteristic Pre-filter using Multiple Model Estimation for Navigation," Proceedings of the 2013 International Technical Meeting of The Institute of Navigation, San Diego, California, January 2013, pp. 670-674.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In