(strongly disagree) to five (strongly agree). The results of questionnaire are shown in Table 3. DISCUSSION The experimental results show that the control strategy is successfully applied to a prototype robot arm designed to bandage a swaying arm. The results reveal that the joint sensors of the robot arm can be used to estimate the direction of the bandage's movement and follow it effectively, while the tension of the bandage can be controlled by the end effector to achieve the desired accuracy. The method of tension direction estimation is considered to be effective. The results of the first experiment indicate that the mean estimation error is 2.85°. Although the accuracy is not perfect, it can still meet the needs of the robot arm to follow the movement of the human arm and complete the bandaging task. The results of the third and final experiments show that the methods can effectively enable the robot to estimate the human arm's moving direction and to follow. The accuracy of the torque measurement and the error of the mass distribution model may be the main influencing factors of the error in tension direction estimation. The former is mainly due to measurement noise caused by jitter from the low stiffness of the SEA itself. The latter is unavoidable, and an effective calibration method is needed. In addition, the st rength of the st ructural par ts of the robot arm and the response speed of the end effector for force control are also potential influencing factors. The tension control method is prov5 4 3 2 1 en to be appropriate. The results of the second and third experiments show that the peak tension error is reduced by 67.2% using the proposed method, and it is able to meet the need of tension for the bandaging task. The proposed following strategy is considered to follow the movement of the injured arm. In the third experiment, the tension is always controlled at around 2.5 N, and the length error of the released bandage does not exceed a maximum of 10 mm. Although the following error is not perfect, it still meetshttps://doi.org/10.1109/MRA.2022.3228501 https://doi.org/10.1109/MRA.2022.3228501 https://www.doi.org/10.1109/MRA.2022.3228501 https://doi.org/10.1109/MRA.2022.3228501 https://doi.org/10.1109/MRA.2022.3228501