Deterministic and stochastic error modeling of inertial sensors and magnetometers
This thesis focuses on the deterministic and stochastic modeling and model parameter estimation of two commonly employed inertial measurement units. Each unit comprises a tri-axial accelerometer, a tri-axial gyroscope, and a tri-axial magnetometer. In the first part of the thesis, deterministic modeling and calibration of the units are performed, based on real test data acquired from a flight motion simulator. The deterministic modeling and identification of accelerometers is performed based on a traditional model. A novel technique is proposed for the deterministic modeling of the gyroscopes, relaxing the test bed requirement and enabling their in-use calibration. This is followed by the presentation of a new sensor measurement model for magnetometers that improves the calibration error by modeling the orientation-dependent magnetic disturbances in a gimbaled angular position control machine. Model-based Levenberg-Marquardt and modelfree evolutionary optimization algorithms are adopted to estimate the calibration parameters of sensors. In the second part of the thesis, stochastic error modeling of the two inertial sensor units is addressed. Maximum likelihood estimation is employed for estimating the parameters of the different noise components of the sensors, after the dominant noise components are identified. Evolutionary and gradient-based optimization algorithms are implemented to maximize the likelihood function, namely particle swarm optimization and gradient-ascent optimization. The performance of the proposed algorithm is verified through experiments and the results are compared to the classical Allan variance technique. The results obtained with the proposed approach have higher accuracy and require a smaller sample data size, resulting in calibration experiments of shorter duration. Finally, the two sensor units are compared in terms of repeatability, present measurement noise, and unaided navigation performance.