File Download
Supplementary
-
Citations:
- Appears in Collections:
postgraduate thesis: A study of GPS/INS integrated navigation with artificial neural network and K-means aided Kalman filter
Title | A study of GPS/INS integrated navigation with artificial neural network and K-means aided Kalman filter |
---|---|
Authors | |
Issue Date | 2017 |
Publisher | The University of Hong Kong (Pokfulam, Hong Kong) |
Citation | Chen, R. [陈睿]. (2017). A study of GPS/INS integrated navigation with artificial neural network and K-means aided Kalman filter. (Thesis). University of Hong Kong, Pokfulam, Hong Kong SAR. |
Abstract | This thesis concerns about two major issues for loosely coupled Integrated Navigation systems of Global Positioning System (GPS) and Inertial Navigation System (INS) with Adaptive Kalman Filter, namely, the INS bias elimination and Kalman Filter noise estimation.
To eliminate INS bias, reliable GPS positioning as the reference is required. Since there are various factors affecting GPS positioning performance and its reliability, it is hard to accurately model the relationship between these factors and GPS positioning performance. Therefore, a novel method based on Artificial Neural Network (ANN) is proposed to evaluate GPS positioning performance and choose reliable GPS positioning for INS accumulated bias elimination. On the other hand, bias prediction and elimination is typically modeled as autoregressive (AR) process, which is sensitive to impulsive noises. Therefore, a robust recursive Least M-Estimate (RLM) algorithm is employed for bias prediction/elimination to battle with impulsive noises. The proposed ANN deploys a 3-layer single output ANN with logistic sigmoid function as the activation function. Limited-memory BFGS method (L-BFGS) logistic regression is used to avoid the storage and reverse of large matrices. Various kinds of information related to GPS and INS positioning such as velocity, position, temperature, GPS satellites signal CN0, etc. are used as input for this ANN, Assistant GPS is used to obtain training samples. Compared with conventional GPS positioning performance evaluation method, the proposed method learns and models the latent stochastic relationship between input data and GPS positioning performance by a large number of supervised training. Compared with conventional Logistic Regression and Linear Regression, the proposed method is more accurate to evaluate GPS positioning performance. In addition, to train the proposed ANN, a Trajectory Matching algorithm is proposed to provide some of the input for ANN. With identified reliable GPS positioning selected by ANN, linear regression and Transversal RLM are used for INS bias elimination and reducing interferences from impulsive noises. The theory of bias in Kalman Filter is also studied.
For GPS/INS integration Adaptive Kalman Filter, a new mathematical method is proposed to make accurate noise covariances estimation. Since transient interferences and unknown noises exist, a novel method with Recursive K-Means clustering is proposed to automatically identify and discard transient high amplitude interferences. Hence, only steady and long lasting measurement errors are used to make noise covariances estimation. The theory of noise estimation in Kalman Filter is also studied.
From the road tests with proposed GPS/INS integration Adaptive Kalman Filter, the effectiveness of the proposed methods to eliminate INS bias and to accurately estimate noise covariances are verified.
|
Degree | Master of Philosophy |
Subject | Inertial navigation systems Global Positioning System Neural networks (Computer science) Kalman filtering |
Dept/Program | Electrical and Electronic Engineering |
Persistent Identifier | http://hdl.handle.net/10722/241424 |
HKU Library Item ID | b5864192 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Chen, Rui | - |
dc.contributor.author | 陈睿 | - |
dc.date.accessioned | 2017-06-13T02:07:50Z | - |
dc.date.available | 2017-06-13T02:07:50Z | - |
dc.date.issued | 2017 | - |
dc.identifier.citation | Chen, R. [陈睿]. (2017). A study of GPS/INS integrated navigation with artificial neural network and K-means aided Kalman filter. (Thesis). University of Hong Kong, Pokfulam, Hong Kong SAR. | - |
dc.identifier.uri | http://hdl.handle.net/10722/241424 | - |
dc.description.abstract | This thesis concerns about two major issues for loosely coupled Integrated Navigation systems of Global Positioning System (GPS) and Inertial Navigation System (INS) with Adaptive Kalman Filter, namely, the INS bias elimination and Kalman Filter noise estimation. To eliminate INS bias, reliable GPS positioning as the reference is required. Since there are various factors affecting GPS positioning performance and its reliability, it is hard to accurately model the relationship between these factors and GPS positioning performance. Therefore, a novel method based on Artificial Neural Network (ANN) is proposed to evaluate GPS positioning performance and choose reliable GPS positioning for INS accumulated bias elimination. On the other hand, bias prediction and elimination is typically modeled as autoregressive (AR) process, which is sensitive to impulsive noises. Therefore, a robust recursive Least M-Estimate (RLM) algorithm is employed for bias prediction/elimination to battle with impulsive noises. The proposed ANN deploys a 3-layer single output ANN with logistic sigmoid function as the activation function. Limited-memory BFGS method (L-BFGS) logistic regression is used to avoid the storage and reverse of large matrices. Various kinds of information related to GPS and INS positioning such as velocity, position, temperature, GPS satellites signal CN0, etc. are used as input for this ANN, Assistant GPS is used to obtain training samples. Compared with conventional GPS positioning performance evaluation method, the proposed method learns and models the latent stochastic relationship between input data and GPS positioning performance by a large number of supervised training. Compared with conventional Logistic Regression and Linear Regression, the proposed method is more accurate to evaluate GPS positioning performance. In addition, to train the proposed ANN, a Trajectory Matching algorithm is proposed to provide some of the input for ANN. With identified reliable GPS positioning selected by ANN, linear regression and Transversal RLM are used for INS bias elimination and reducing interferences from impulsive noises. The theory of bias in Kalman Filter is also studied. For GPS/INS integration Adaptive Kalman Filter, a new mathematical method is proposed to make accurate noise covariances estimation. Since transient interferences and unknown noises exist, a novel method with Recursive K-Means clustering is proposed to automatically identify and discard transient high amplitude interferences. Hence, only steady and long lasting measurement errors are used to make noise covariances estimation. The theory of noise estimation in Kalman Filter is also studied. From the road tests with proposed GPS/INS integration Adaptive Kalman Filter, the effectiveness of the proposed methods to eliminate INS bias and to accurately estimate noise covariances are verified. | - |
dc.language | eng | - |
dc.publisher | The University of Hong Kong (Pokfulam, Hong Kong) | - |
dc.relation.ispartof | HKU Theses Online (HKUTO) | - |
dc.rights | This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. | - |
dc.rights | The author retains all proprietary rights, (such as patent rights) and the right to use in future works. | - |
dc.subject.lcsh | Inertial navigation systems | - |
dc.subject.lcsh | Global Positioning System | - |
dc.subject.lcsh | Neural networks (Computer science) | - |
dc.subject.lcsh | Kalman filtering | - |
dc.title | A study of GPS/INS integrated navigation with artificial neural network and K-means aided Kalman filter | - |
dc.type | PG_Thesis | - |
dc.identifier.hkul | b5864192 | - |
dc.description.thesisname | Master of Philosophy | - |
dc.description.thesislevel | Master | - |
dc.description.thesisdiscipline | Electrical and Electronic Engineering | - |
dc.description.nature | published_or_final_version | - |
dc.identifier.mmsid | 991026390699703414 | - |