以學習演算法為基礎之中風病患關節協調訓練應用於外骨骼式上肢復健機器人

Abstract

對於中風病患而言，在自由度與肌肉力量改善之後，往往伴隨著不正常的關節力量反應，例如協同作用或是肌肉痙攣。這種不正常的協同作用經常阻礙病人以至於無法執行一個協調性的運動或是日常生活中的功能性動作。因此為了提升機器人輔助復健的應用，我們提出了關節協調性這個議題以及相對應的訓練以便於讓病人可以透過重複的練習改善不正常的力量共同反應。此外，學習演算法被運用來即時調整模式選擇因子以便於平衡各軸出力的不平衡來誘發病人更加協調的表現。 整體控制系統架設並呈現於近期開發的上肢外骨骼復健機器人NTUH-II。相對應的運動模型，人體動態模型以及力量力矩資訊之處理將被完成地整合以便於達到順應式之人機互動。最後為了驗證我們的設計，將透過NTUH-II呈現實驗數據並展示整體訓練演算法的效果與表現。

The stroke patients, after range of motions and muscle strengths recovered, are usually left with abnormal torque reflections such as synergies and spasticity. The abnormal synergies tend to obstruct the patients from performing coordinated motions or functional tasks in activity of daily living. Therefore, we enhance the application of robot-assisted rehabilitation by addressing issues of inter-joint coordination with its corresponding training for patients to reduce abnormal torque coupling effects via repetitive practices. In addition, learning algorithm is applied for online tuning of mode selection gains to compensate the imbalanced joint torques to induce more coordinated performance in rehabilitation process. The control system of the training is implemented and addressed on the recently developed upper limb exoskeleton rehabilitation robot NTUH-II. The extended kinematic model, human arm dynamic model and the processing of force/torque information are well integrated to achieve the compliant motion with human interaction. Finally, to validate our design, several experimental results demonstrate the performance of the training algorithm via NTUH-II.