具有摩擦力估測之無感測器上肢復健機器人控制

Abstract

中風是一種常見的神經性損傷，中風倖存者通常伴有失能，主要是上肢和下肢運動障礙。臨床研究表明，患者可通過高密度以及高參與率的復健療程重建起運動能力。大多數中風患者需要持續復健一段時間，導致大量的人力和成本需求。引進外骨骼上肢復健機器人的治療，可實現更頻繁的治療並降低醫療成本。應用機器人輔助訓練中的控制策略可分為被動，輔助和主動控制。 本研究提出一套應用於無傳感器外骨骼上肢復健機器人的主動控制方法。由於NTUH-II外骨骼機器人的機械結構，摩擦力模型較一般的市售的機器人複雜。然而，此研究需建立精確的摩擦力模型，以便建立一個應用卡爾曼濾波器的交互作用扭矩觀測器來估計人與機器人之間的相互作用力。並使用相互作用力來轉換成機器人的運動軌跡，藉此使機器人跟隨患者的手臂運動。應用此控制方法讓患者可自由運動手臂以訓練其運動控制能力。 本研究提出的控制方法已經於三位健康受試者之臨床試驗予以驗證。結果顯示與相關文獻相比，它可以提高運動的平滑性，同時減少使用者的施力。與使用肌電訊號感測器和力/力矩感測器方法相比，本研究成果顯示，本方法可省去額外感測器的成本並同時達到良好的表現。

Stroke is one of the leading causes which acquired neurological impairments in adults. Stroke survivors usually remain some disabilities in which the most common disabilities are motor impairments on both upper and lower limb movement. Clinical study shows that when the rehabilitation is both intense and involves the patients, the patients might regain their motor ability. Rehabilitation therapy usually needs to be performed persistently throughout most of the stroke patients’ life, hence can be labor-intensive and costly. Exoskeletons robots have the potential to enable more frequent treatment and potentially reducing costs. The control strategy in robot-assisted training (RAT) can separate into passive, active-assisted and active mode control. In order to design the active mode control, the human-robot cognitive interaction control becomes one of the key topics. In this study, a sensorless control of upper limb rehabilitation exoskeleton, named NTUH-II is proposed. Due to mechanical structure of NTUH-II, the friction behavior is more complicated than the commercial robot. However, the accurate friction model is crucial for constructing an interaction torque observer using Kalman filter which is used to acquire human intention. Furthermore, taking the human intention to derive the desired motion trajectory of the exoskeleton. As a result, the robot arm will follow the user’s arm, and the patients can freely move their arm to improve their motor control. Various experiments have been conducted on three subjects which verify the performance of the proposed interactive torque observer based controller. The results show that it can improve the smoothness of the motion and reduce the subject’s effort compared with the related works. Moreover, compared with the use of electromyography sensor and force/torque sensor, this method can get rid of additional sensors while achieving good performance.