應用速度場之輔助控制於上肢復健外骨骼機器人

Abstract

中風是造成成人失能一種常見的神經性損傷，其倖存者常伴隨上肢和下肢的運動障礙。據臨床研究結果顯示，中風患者可透過密集參與復健療程以回復其運動能力。將上肢復健外骨骼機器人引入療程，除了可降低醫療人力成本，更可實現密集且精確的治療。應用機器人於復健療程之控制策略可分為被動、輔助及主動控制。其中輔助控制的實現，需同時考量病患之運動意圖與欲達成的復健任務，藉此提供病患所需的協助達成目標。然而，於相關文獻中，復健任務之模型多為與時間相依之軌跡，導致其限制使用者自主控制的自由。本研究提出適用於上肢復健外骨骼機器人之輔助控制方法。首先，使用者的運動意圖可由應用卡爾曼濾波器之交互作用扭矩觀測器，經導抗模型轉換所獲得。接著，本研究提出基於速度場表示之任務模型，該模型可經由給定之任務軌跡生成，並且只與使用者之位置資訊相依，讓使用者不受限於需在特定時間達成特定動作。而提出之整合主動意圖與輔助的方法，可以根據使用者執行任務的投入程度與表現，於執行復健任務中調整機器人之輔助程度，並達成於考量任務的同時促進使用者之主動意圖。最後，透過Lyapunov穩定性分析，確保提出之輔助控制方法的穩定性與效能。本研究提出之輔助控制方法已應用於健康受試者之臨床試驗予以驗證。結果顯示與相關文獻方法相比，可於減少使用者完成任務之時間與施力的同時，促進使用者主動運動之意圖並精確達成給定之各項復健任務。於未來展望，為驗證提出系統之療效，需將本系統應用於中風病患之臨床試驗。

Stroke is a prevalent source of neurological impairment which causes disability in adults. The survivors of it commonly suffer from motor impairments on both upper and lower limb motion. According to clinical studies, the patients can regain their motor ability by intensively involving in rehabilitation therapy. Introducing upper limb rehabilitation exoskeleton robot into the therapy can not only reduce the labor cost but also provide intensive and accurate treatment. The control strategy for applying robot in rehabilitation therapy can be classified as passive, active, active-resistive and active-assistive control. To implement active-assistive control, both motion intention of the patients and the given rehabilitation task should be taken into consideration so that the controller can provide necessary assistance to achieve the goal. However, most of the rehabilitation task model in related works is a time-dependent trajectory, which limits the freedom of subjects to control actively. An active-assistive controller based on interactive torque observer is proposed for upper limb rehabilitation exoskeleton robot, NTUH-II. First, the motion intention of the subjects can be obtained by utilizing an adaptive Kalman filter based interactive torque observer. Next, we propose a velocity field based task model which can be generated via a given task trajectory. The model only depends on the location information of the subject so that the subject is not limited to achieve the motion at a specific time instant. The proposed active-assistive method can integrate the active and assistive motion based on the performance and the active involvement of the subject. The integration result considers not only the human active intention but also the given task such that the subject can perform the task more actively and accurately. Finally, the stability and effectiveness of the proposed active-assistive control system are verified by Lyapunov stability analysis. Various experiments are conducted on three healthy subjects and NTUH-II to verify the proposed active-assistive control system. The results show that compared with the related works, not only the execution time but also the subjects' exertion can be reduced when performing the given rehabilitation tasks. In addition, the proposed method can promote subjects' intention of active motion and assist them to accomplish the tasks accurately. In the future work, the effectiveness of the proposed system for stroke patients is required to be validated through clinical studies.