非奇異快速終端滑模超扭轉控制在下肢外骨骼機器人之半癱復健

Abstract

本論文以「非奇異快速終端滑模控制應用於髖膝外骨骼機器人之下肢半癱復健」為研究主軸，針對因中風或神經損傷導致單側下肢癱瘓的復健需求，設計並實作一款可主動驅動髖關節與膝關節之穿戴式下肢外骨骼系統。其主要目標在於模仿健康側肢體的動作，帶動癱瘓側完成對稱步態，進而達到復健輔助之效果。硬體方面，外骨骼之機構設計依照人體下肢比例進行調整，並採用直流無刷馬達、編碼器與多軸驅動器進行雙關節控制；通訊部分透過 RS-232通訊方式與控制端實現即時資料接收與命令輸出。控制方法方面，本研究採用無模型控制架構，並整合線性擴展狀態觀測器（Linear Extended State Observer, LESO）以即時估測總擾動，搭配非奇異快速終端滑模控制（Nonsingular Fast Terminal Sliding Mode Control, NFTSM）與超扭轉控制（Super-Twisting Control）提升系統追隨能力與抗擾性能。進一步地，本研究導入阻抗模型控制機制，以模擬人機交互過程中的生物力學行為，進行交互力補償與柔順性調節，強化步態控制的自然性與安全性。本研究進行多組實驗，包括慢速步行、變步距步行等不同情境，觀察 PID、NFTSM-ST 與 LESO-NFTSM-ST 三種控制器之表現。結果顯示，LESO 輔助之 NFTSM 控制策略不論在穩定性、收斂速度或誤差抑制方面皆優於傳統方法，可有效提升外骨骼於復健過程中的使用效益，具備實際應用潛力。

This thesis presents a study on “Nonsingular Fast Terminal Sliding Mode Control for Lower-Limb Exoskeleton Robot on Hemiplegia Rehabilitation.” A wearable exoskeleton robot is developed to actively drive the hip and knee joints, aiming to assist individuals with hemiplegia caused by stroke or neurological disorders to achieve symmetrical gait by following the movement trajectory of the healthy limb so that the impaired limb is improved accordingly.The hardware system is designed based on human lower limb proportions, employing brushless DC motors, encoders, and multi-axis motor drivers for dual-joint actuation. The control unit communicates with actuators is via RS-232 to ensure real-time data transmission. To estimate unknown disturbances, a model-free framework is adopted and enhanced with a Linear Extended State Observer (LESO) to estimate total disturbances in real time. Moreover, integrated with Nonsingular Fast Terminal Sliding Mode Control (NFTSM) and Super-Twisting Control (STC), the system is improved for tracking accuracy and disturbance rejection. Furthermore, an impedance control strategy is introduced to model the biomechanical interaction between the human and the exoskeleton, enabling compensation for interaction forces and improving motion compliance and safety.Experiments, including slow walking and variable step length walking, were conducted to evaluate and compare the performance of three control strategies: PID, NFTSM-ST, and LESO-NFTSM-ST with/without impedance control. The results demonstrate that the proposed LESO-assisted NFTSM controller significantly outperforms traditional methods in terms of stability, convergence speed, and tracking accuracy. The integration of impedance modeling further enhances the naturalness of gait assistance, confirming the practical feasibility and application potentials of the proposed system on hemiplegia rehabilitation.