Evaluation of Torque-Sensorless Control for a Knee Exoskeleton Using Back-Drivable Actuators

Abstract

Exoskeleton robots have attracted significant attention in an increasingly aging global population with a declining birthrate. However, exoskeleton robots are not as popular as other service robots, such as communication robots and mobility robots. This is primarily because exoskeleton robots are in close contact with the human body, making safety and operability an issue. Reduction gears that show a high reduction ratio and high backdrivability have been proposed to solve these problems. Reduction gears with a high backdrivability can be used to estimate external force without using any additional devices, such as torque sensors. Therefore, this study introduces the development of an knee exoskeleton using a highly back-drivable gear. The size of a developed knee exoskeleton is approximately 220 W × 60 D x 65 H mm with a weight of approximately 1.2 kg. Moreover, we introduce an external force estimation method without a torque sensor. We used a multi encoder based disturbance observer (MEDOB) to estimate the external force. The experimental results of the external force estimation are presented. The root mean square error of the external force estimation without and with input assist torque were 0.162 and 0.128 N. m, respectively. We confirm that estimating external torque with high accuracy is possible with and without assisted control.