Isfahan University of Technology
Department of Electrical and Computer Engineering
Issue date: May, 2017
Contributor: Narjes Davari
Supervisor: Dr. Asghar Gholami
Advisor:Dr.Mohammad Reza Taban
Inertial navigation as an independent navigation system is vastly used in different applications. However, due to its intrinsic specifications and external factors, estimation error in this system indefinitely increases with time. Thus, to limit the estimation error, one needs to use auxiliary sensors and integrated navigation methods. The integrated navigation algorithms should be sufficiently accurate and reliable. The aim of this study is to design and implement methods to increase the accuracy and reliability of integrated navigation system for underwater vehicles. The system includes an inertial measurement unit and a processor to perform calculations. Auxiliary navigation sensors include GPS, DVL, Depthmeter and Compass. The system is able to estimate a variety of vehicle trajectory outputs, including, location in terms of longitude, latitude and depth, speed in the north, east and vertical directions, attitude in terms of the roll, pitch and heading angles, and estimation error for these outputs. System performance is evaluated using practical experiments.
Parameter adjustment for the integration algorithm is too time consuming and difficult. Moreover, outliers that appear at unpredictable moments make it difficult to estimate the exact value of the parameters for the integration algorithm. The application of sensors with different sampling rates requires the design of multi-rate integration algorithms. Therefore in this research, data integration methods, such as adaptive filters, that match dynamic motion and environmental conditions are used. Also, adaptive integration algorithms for several sensors with different sampling rates based on the innovation sequence and the variational Bayesian approximation (MAESKF, MAEKF and VB-MAESKF) have been proposed and implemented. For MAESKF and MAEKF algorithms, computation of innovation sequence vector and covariance matrix, and measurement noise covariance matrix for multi-rate sensors has been modified. For the VB-MAESKF algorithm, using measurement vector and matrix, and coefficients vector for sensors data in various time steps, measurement noise covariance matrix and Kalman gain are estimated. In this study, practical experiments have shown that the proposed multi-rate adaptive integration algorithm based on Bayesian approximation is robust against the outliers of auxiliary sensors (especially DVL), and does not require other algorithms to identify and eliminate the outliers; and it provides high precision navigation compared to other navigation methods. Finally, a novel algorithm to detect and eliminate outlier data of auxiliary sensors based on statistical data parameters is proposed. In this method, assuming that data follows a Gaussian probability function in a time window, the statistical moments are calculated, and acceptance/ rejection range for the data is presented. The performance of this algorithm is evaluated through practical experiments. Also, in this research, the stability conditions of the estimation for the error state Kalman filter integration algorithm based on the square error limitation are presented.