الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 191 |  |  |
| | | from | 191 | | |
Abstract
The use of MEMS-based IMUs has been developed to provide an accurate
estimation of the position, velocity and attitude of a moving target when integrated
with a GPS system through KF. But there is a challenge that the inertial sensors
measurements errors accumulate with time leading to inaccuracy in the estimation
of the navigation solution. So, there was a need for integrating INS with external
sources such as GPS to prevent the navigation solution from degradation while using
INS as a stand-alone system during GPS signal outages due to jamming,
interference, signal blockage and multipaths then the accuracy of the navigation
solution will depend on the inertial sensors measurements only. There are two main
types of INS errors: deterministic and stochastic errors. The deterministic errors can
be removed by calibration techniques in laboratory conditions. The stochastic errors
need to be modelled using random processes such as a 1
st order GM process, QN,
WN, BI, RW and DR which can be estimated by several stochastic error modelling
techniques such as ACF, AV and GMWM. The thesis aims to improve the stochastic
error modelling of the inertial sensors to provide an accurate navigation solution
while using INS as a stand-alone system. Laboratory static data collection processes
will be executed using two different MEMS-based IMUs, ADIS and SPATIAL
modules. Three different noise analysis techniques will be utilized to expose,
examine, and reveal the inertial sensor stochastic error part. Previously, a 1st order GM process was used only to model the stochastic error part of the inertial sensors
which could be estimated using the ACF approach. This is not enough to accurately
model this error part for various inertial sensors. Two different stochastic error
modelling methods will be used in this thesis namely the AV and GMWM. The pros
and cons of each method in identifying and estimating the parameters of the
stochastic processes will be discussed and a comparison between them will be
applied to get the best method in modelling the stochastic part of the inertial sensors
after performing a fitting test which would be GMWM method. After using GMWM,
a contamination test will be applied to identify the data that was contaminated by
outliers. Then, the best model that fits the desired signal will be selected from fortyfour candidate models using the WIC approach. Then, dynamic in-field datasets will
be collected to use with the modified navigation algorithm. Finally, a modified
loosely coupled INS/GPS integration navigation algorithm will be proposed based
on the selected best model from GMWM approach. The modified navigation
algorithm consists of 39 states based on the best model with the smallest value of
WIC. A comparison between the standard navigation algorithm that consists of 15
states based on a 1st order GM process estimated by the ACF method and the
modified navigation algorithm that consists of 39 states based on the selected best
model will be applied during GPS outage periods to study the accuracy of the
navigation solution getting from each one. The position errors from the modified navigation algorithm are less than those from the standard navigation algorithm. This
means that the quality of the INS-based navigation solution (when the INS is used
as a stand-alone system during GPS signal outage) is improved by using the
modified 39-states integrated navigation algorithm associated with the stochastic
noise parameters estimated using the GMWM.