An Adaptive Kalman Filter for UAV Attitude Estimation

Attitude estimation plays an important role for the stable flight of a UAV. This paper proposes an adaptive Kalman filter (AKF) to estimate the attitude of a UAV by combining inaccurate angle measurements from a gyroscope and an accelerometer. An attenuation factor related to the innovation sequence in the Kalman filter is introduced to estimate the system noise and the measurement noise simultaneously, which improves the convergence and stability of the filter. Experimental results show that the adaptive Kalman filter algorithm can effectively track the rapid changes of the statistical characteristics of noise, which leads to a more robust and accurate attitude estimation.