| Summary: | Inertial navigation is always available as a base for multisensor navigation systems on, because it requires no external signals. However, measurement errors persist and grow with time so accurate calibration is crucial. Large systematic errors are present in the micro-electro-mechanical sensors (MEMS) whose low cost brings inertial navigation to many new applications. Using factory-calibrated MEMS another navigation technology can calibrate these errors with in-run estimation using a Kalman filter (KF). However, the raw systematic errors of low-cost MEMS are often too large for stable performance. This thesis contributes to knowledge in three areas. First, it takes a simple GNSS-inertial KF and examines the levels of the various systematic errors which cause the integration to fail. This allows the user to know how well calibrated the sensors need to be to use in-run calibration. Second, the thesis examines how the end-user could conduct a calibration: it analyses one method in detail showing how imperfections in the procedure affect the results and comparing calculation methods. This is important as frequently calibration methods are only validated by demonstrating consistent results for one particular sensor. These two are primarily accomplished using statistical Monte Carlo simulations. Third, techniques are examined by which an array of inertial sensors could be used to produce an output which is better than the simple array average. This includes methods that reduce the array’s sensitivity to environmental conditions, this is important because the sensors’ calibration typically depends strongly on temperature. Also included in the thesis are descriptions of experimental hardware and experiments which have been carried to support and unify the other parts of the thesis. Overall, this thesis’ contributions will help make low-cost inertial navigation systems more accurate and will allow system designers to concentrate effort where it will make the most difference.
|
|---|
