摘要：Joint torque feedback (JTF) is widely consideredas an effective technique for improving the performance ofrobot control. To implement JTF technique to robot controlschemes, it is essential to fully get the measurements orestimations of the output torque transmitted by the system.In previous research, output torque is commonly measuredby joint built-in torque sensors which will change the robotjoint mechanical structure, or through position encoder measurementswith the transmission model of the joint. In thispaper, a new filter method is adopted to improve the jointtorque accuracy based on both precise model and positionmeasurements. More specifically, an optimization method basedon Unscented Kalman Filter (UKF) algorithm is used togetherwith the harmonic drive compliance model to achieve higheroutput joint toque accuracy. The proposed method has beenexperimentally studied and its performance is compared withthe measurements of a commercial torque sensor. And theresults have verified the effectiveness of the proposed torqueestimation method.
