Download presentation
Presentation is loading. Please wait.
Published byJohan Littleford Modified over 10 years ago
1
Location Forum 2006, 7 November, 2006 School of Surveying & Spatial Information Systems The University of New South Wales, Australia Adaptive Kalman Filtering for GPS/INS Integration Weidong Ding This research is supported by the Australian Cooperative Research Centre for Spatial Information (CRC-SI) under project 1.3 ‘Integrated positioning and geo-referencing platform’.
2
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 2 presented by Weidong Ding GPS/INS integration Surveying, navigation, location based services, etc. Solution of position & attitude Long term accuracy, high update rate, robustness, INS calibration
3
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 3 presented by Weidong Ding Limitations of Kalman Filter Wrong parameters of system models and noise properties may result in the filter being suboptimal or even cause it to diverge.
4
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 4 presented by Weidong Ding Adaptive Kalman Filter Covariance scaling method By applying a scale factor to the predicted state covariance matrix to deliberately decrease the weight of the state prediction, to improve KF stableness. Multi-model adaptive estimation A group of KF filters; each has slightly different configurations. The output is the optimal combination of the outputs from individual filters. Adaptive stochastic modelling (Innovation based, Residual based) Uncertain stochastic modelling parameters are estimated on-line using the covariance information of the KF innovation and residual series. A new process noise scaling method is proposed.
5
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 5 presented by Weidong Ding Results of on-line stochastic modeling
6
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 6 presented by Weidong Ding Results using process noise scaling
7
School of Surveying & Spatial Information Systems The University of New South Wales, Australia School of Surveying and Spatial Information Systems Location Forum 2006, 7 November, 2006 7 presented by Weidong Ding Summary The online stochastic modelling method can estimate the individual elements of noise covariance matrix. However, it is vulnerable to the innovation and residual covariance estimation biases, and is not scalable to a large number of parameters. The covariance scaling method is more robust and suitable for practical implementations. The proposed covariance based adaptive process noise scaling method has demonstrated significant improvements on the filtering performance in the test. Optimal allocation of noise to each individual source is not possible using scaling factor methods, which is a topic for further investigation.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.