An Adaptive Filter for IMU/Encoder Data Fusion for Acceleration Estimation in Robot Arms

Abstract

This paper develops an adaptive filter for fusing the noisy and biased measurement data from MEMS-based inertial measurement units and encoders for estimation of acceleration in robot arms. A discrete-time second-order model is derived for designing an adaptive Kalman filter (AKF). The output of the AKF is an unbiased but noisy estimate of the acceleration. To cancel the noise, a recursive least square filter is designed to filter the output of the designed AKF. It is shown that the resulting filtered signal is unbiased and noise-cancelled. Experimental results demonstrating the effectiveness of the developed adaptive filter are presented.