Online Gyro Bias Estimation from Single Vector Measurements Using Regression Models ☆

This paper addresses the problem of on-line consistent estimation of gyro bias using the measurements of a single vector and the biased angular velocity - both in the body-fixed frame. We propose two globally convergent gyro bias observers using new regression models, which are capable to deal with the cases of constant and time-varying reference vectors, respectively. Indeed, there are quite a few works discussing the latter case. To address this, we derive a nonlinear regression model, based on which a convexified gradient descent observer is designed, providing globally asymptotically convergent estimates to the gyro bias under some sufficient excitation conditions. The proposed schemes are illustrated by some numerical simulations.