Abstract
This paper describes a motion control for parallel mechanism. A motion control is commonly based on an acceleration controller. High performance acceleration controller is realized by the disturbance observer. In parallel mechanism, high performance acceleration controller is also expected to be improved in workspace. However, it is difficult to calculate its forward kinematics analytically. In a conventional method, it takes more time to compute its forward kinematics compared with serial mechanism, unless passive joint angles are measured directly. Thus, it is possible to reduce sampling rate of controller. In this paper, the conventional acceleration controller based on disturbance observer is extended to parallel mechanism. A new observer is proposed which estimates not only disturbance torque, but also forward kinematics through identity observer. The estimated joint angular velocity is transformed to workspace velocity by jacobian matrix. And forward kinematics is obtained by integrating workspace velocity. Moreover, estimated joint angle which is calculated by inverse kinematics, is used to reduce estimated error. By implementing the feedback of estimated disturbance to input torque, disturbance torque is compensated without inverse dynamics calculation. High performance acceleration controller in parallel mechanism is realized as well as in serial mechanism. A new parallel mechanism is also developed for the biped robot which has light legs. The validity of the proposed method is confirmed by the experimental results.
