LAUSR

In this paper, we consider the inertial measurement units (IMUs) based attitude estimation, a common core problem in a large variety of robotic systems. In particular, we focus on the case where the IMU measurements are contaminated by narrow-band vibration noises caused by the system flexible mode, aerodynamic turbulence, or rotating parts (e.g., motors). Our proposed scheme is a complete solution to adaptively estimate the narrow-band noise and actively compensate it in the attitude estimation. More specifically, the scheme adopts a notch filter to attenuate the narrow-band noise and subsequently embeds such filter into a multiplicative extended Kalman filter framework to eliminate the induced transient error. Furthermore, a least mean square method is utilized to estimate the dominant noise frequency in real time, forming an adaptive notch filter that can effectively attenuate the narrow-band noise with time-varying dominant frequency. The complete algorithm is verified on an actual gimbal system with exhaustive experiments.