本研究針對手指虛實整合力量回饋系統，提出結合環境參數辨識與自適應導納控制之控制架構，以改善固定參數導納控制於多變接觸環境下之限制。系統以力追蹤導納控制作為力量誤差與運動命令之轉換方法，根據虛擬環境目標作用力與實體端量測力之差異產生對應位移命令，並透過串聯彈性致動器（Series Elastic Actuator， SEA）量測實體端回饋力量。由於不同接觸環境具有不同等效剛性與阻尼特性，若導納參數維持固定，系統可能因環境變化而產生力量追蹤誤差增加、接觸力過衝或震盪等問題。因此，本文導入環境參數辨識，使導納控制器可依據接觸環境之估測結果調整虛擬阻尼與虛擬剛性。 在環境參數辨識方面，本文以 Kelvin–Voigt 模型描述虛擬接觸環境之等效阻尼與等效剛性，並利用位置、速度與目標環境作用力建立回歸模型。為改善僅使用即時資料時可能產生之收斂不足問題，本研究採用並行自適應控制方法，結合即時資料與歷史資料進行參數更新，使系統可在有限互動資料下取得虛擬接觸環境之等效參數估測結果，並作為導納參數調整之依據。 本研究將所提出之控制方法建構於嵌入式微控制器平台，並透過模擬與實驗驗證其可行性。環境參數辨識結果顯示，估測剛性可依不同虛擬物體條件進行修正，並收斂至設定值附近；阻尼估測則較容易受到速度訊號品質與未建模動態影響，因此較適合作為接觸動態特性之輔助參考。在力量追蹤實驗中，自適應導納控制於不同力量輸入型態下，皆可使虛擬環境回授力與 SEA 實體端量測力跟隨目標作用力之主要變化。相較於固定參數導納控制，所提出方法可依據不同接觸環境之估測結果調整控制響應，使系統在多變接觸條件下維持較一致之力量追蹤表現。 整體而言，本文所提出之控制架構結合虛擬接觸環境之等效參數辨識與導納參數自適應調整，使手指虛實整合力量回饋系統能依據不同接觸條件修正控制響應。研究結果顯示，並行自適應控制可提供環境參數估測資訊，而自適應導納控制可改善固定參數導納控制於多變接觸環境下之力量回饋表現，並可作為後續多自由度手指力量回饋與實體接觸環境辨識之控制基礎。;This study proposes a control framework that integrates environmental parameter identification and adaptive admittance control for a finger virtual–physical force feedback system. The objective is to address the limitations of fixed-parameter admittance control under varying contact environments. In the proposed system， force-tracking admittance control converts the force error between the target force from the virtual environment and the measured force at the physical side into a displacement command. The physical-side feedback force is measured through a series elastic actuator (SEA). Since different contact environments exhibit different equivalent stiffness and damping characteristics， fixed admittance parameters may lead to force tracking errors， contact force overshoot， or oscillation. For environmental parameter identification， the Kelvin–Voigt model is used to describe the equivalent stiffness and damping of the virtual contact environment. A regression model is established using position， velocity， and the target environmental force. To address the convergence limitation that may occur when only real-time data are used， a concurrent adaptive control method is adopted. By combining real-time data with historical data， the system can estimate the equivalent parameters of the virtual contact environment under limited interaction data. The estimated parameters are then used to adaptively adjust the virtual damping and stiffness of the admittance controller. The proposed method is implemented on an embedded microcontroller platform equipped with a series elastic actuator and verified through simulation and experiments. The results show that the estimated stiffness can be adjusted according to different virtual object conditions and converges closely to the preset values， while the damping estimate is more sensitive to velocity signal quality and unmodeled dynamics. Compared with fixed-parameter admittance control， the proposed method provides more consistent force tracking performance under varying contact conditions and reduces force oscillation in high-stiffness environments. The proposed framework can serve as a control basis for future multi-degree-of-freedom finger force feedback systems and physical contact environment identification.
