Parameter Identification of the Acrobot Using the Unscented Kalman Filter

摘要

This paper considers parameter identification of a 2 link underactuated Acrobat robot called Acrobot. According to parameter identification studies, the method using Extended Kalman Filter (EKF) has been proposed. The Acrobot considers here is assumed to be influenced kinematic and viscouse friction, and it's equation of motion includes discontinuous terms. Unfortunately, EKF can't be executed such an indifferentiable nonlinear systems. Then we proposed a new method using the Unscented Kalman Filter (UKF) for parameter identification of the Acrobot. UKF is the state estimation method for nonlinear systems developed in recent years, and this method is applicable even when there is a discontinuity in nonlinear systems. The proposed method's effectiveness has been clarified through computer simulation and experiment.