A novel method to improve the accuracy of torque estimation for robotic joint with harmonic drive transmission

Joint torque feedback (JTF) is widely considered as an effective technique for improving the performance of robot control. To implement JTF technique to robot control schemes, it is essential to fully get the measurements or estimations of the output torque transmitted by the system. In previous research, output torque is commonly measured by joint built-in torque sensors which will change the robot joint mechanical structure, or through position encoder measurements with the transmission model of the joint. In this paper, a new filter method is adopted to improve the joint torque accuracy based on both precise model and position measurements. More specifically, an optimization method based on Unscented Kalman Filter (UKF) algorithm is used together with the harmonic drive compliance model to achieve higher output joint toque accuracy. The proposed method has been experimentally studied and its performance is compared with the measurements of a commercial torque sensor. And the results have verified the effectiveness of the proposed torque estimation method.