The MEMS IMU Error Modeling Analysis Using Support Vector Machines

It's well known that the accuracy of the inertial navigation systems will rapidly degrades with time because of the measure sensor's error. Several variance techniques have been devised for the error modelling of this error by way of weighting functions, PSD, ARMA and NNs, etc. In this pap...

Full description

Saved in:
Bibliographic Details
Published in2009 Second International Symposium on Knowledge Acquisition and Modeling Vol. 1; pp. 335 - 337
Main Authors Guoqiang Xu, Xiuyun Meng
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.11.2009
Subjects
AR model
Electronic mail
Error analysis
Knowledge acquisition
MEMS IMU
Micromechanical devices
Navigation
NN model
Stochastic resonance
Support vector machine classification
Support vector machines
SVM/SVR
Training data
White noise
Online AccessGet full text

Cover

More Information
Summary:It's well known that the accuracy of the inertial navigation systems will rapidly degrades with time because of the measure sensor's error. Several variance techniques have been devised for the error modelling of this error by way of weighting functions, PSD, ARMA and NNs, etc. In this paper, we use the SVM(support vector machine) technique to predict the future noise coming from the measure sensors especially the gyro. Then we compare the resulting noise data with the one coming from the ARMA model and NNs model. Finally the three models are compensated to the output data from the IMU to compute the position errors and attitude angle errors. The results indicate that the SVR model (support vector regression) shows more stable feature and is more adequate for long time navigation than the AR model and NNs model.
ISBN:9780769538884
0769538886
DOI:10.1109/KAM.2009.287