IJCAS

Independent joint control by disturbance observer schemes requires the information of an inertial mass and an angular acceleration of each joint of a robot manipulator in a real-time fashion. When both the inertial mass and acceleration information is available, we can easily estimate the joint disturbances. However, it is difficult and costly to mount a torque sensor on each joint and accelerometer. In addition, when the information obtained through the optical encoder is estimated by the finite difference method, the estimated angular velocity and acceleration may be noisy. In this paper, a costly effective data-driven approach for the identification of a link parameter of a robot is presented. Based on the input and output data, a joint model is estimated by the recursive least square (RLS) algorithm. Using the identified models, a state observer in the discrete domain is designed to estimate the joint acceleration signal of a robot manipulator. A combined structure of RLS and state observer is implemented in digital control hardware to estimate the inertial mass as well as the angular acceleration of each joint in a real-time fashion without adding any cost. The performance of estimating the angular acceleration and inertial mass of each joint of a robot manipulator are empirically compared with the finite difference method under the same hardware condition.

Keywords: Data driven approach, joint acceleration and mass identification, RLS, robot manipulator, state observer