چكيده به لاتين
Although emergence of MEMS technology has decreased volume, weight, price and power consumption of inertial sensors and it plays a significant role in the market, but accuracy and precision of these sensors are not suitable for many applications; thus, accurate and expensive inertial sensors are still the only option for accurate navigation systems. Purpose of this thesis is to increase precision through hardware redundancy and improve accuracy through a novel method for complete calibration of inertial sensors. Hardware redundancy is particularly suitable for MEMS sensors due to their lower volume, weight, price and lower power consumption.
In order to eliminate random error and increase precision, a novel method is proposed for redundant sensors based on linear constrained least squares with updating estimation parameters. In this method, a new modeling is proposed for redundancy sensors as a linear constraint. Unlike conventional models, which are only used for modelling inertial sensors, the proposed redundant sensor model can be employed for other kinds of measurement like pressure or temperature. This model is employed in least squares form with updating parameters to combine data of inertial sensors. Solution of the linear constrained least squares with updating estimation parameters is obtained analytically and it is proved that the proposed estimator is unbiased. Increasing precision in this estimator is shown through analyzing covariance matrix.
In order to improve accuracy, a new method is proposed for calibrating inertial sensors through extension of least squares with updating coefficients. In the previous methods, there were constraints like requiring data of only 6 or 12 positions, being oriented along gravity vector and estimating a limited number of calibration coefficients. But in the proposed method, it has been shown analytically that measurement information of at least 4 positions is required witch they have no need to align in particular direction. However, all calibration coefficients including bias, scale factor, non-orthogonality and misalignment are estimated. In addition, 5, 6 or more number of different positions can be used for better estimation of calibration coefficients. The proposed estimation method is applied to the attitude determination algorithm using accelerometer and magnetometer. Results show that precision and accuracy are improved.
