本研究藉由肌電訊號、磁性編碼器、慣性感測器所得的訊號，透過機器學習的方法，建立人體在運動時的模型並判斷姿態，進而下達命令給馬達提供正確的輔助力。膝關節輔具使用特殊設計的四連桿，以配合腿部運動軌跡。透過雙眼視差法，使用網路攝影機取得腿部運動時的姿態，匯入Matlab建立各姿態時腿部的三維空間座標，再使用差分進化演算法(Differential Evolution)，求得最佳四連桿瞬心軌跡。最後，量測腳上的肌電訊號、大腿及上半身的慣性感測器、膝關節四連桿上的磁性編碼器，透過機器學習判斷當下姿態，作為輔具馬達出力的參考訊號。

This paper developed a method to design a four-bar linkage for the knee. We use two USB-camera to capture knee motion. With those date , using Matlab can calculate the instantaneous center(I.C.) of knee. Once obtain I.C. trajectory, we design an objective function and use initial condition to generate four-bar linkage. By comparing generated trajectory with measured trajectory, objective function will output an error. This error will be the index of Differential Evolution(DE). Finally, optimal four-bar linkage can be obtained with DE. This optimal four-bar linkage will assemble on the knee with reduction gear and servo motor. We also design an Electromyography(EMG) device to detect electrical activity produced by muscles. Additionally, using inertial measurement unit(IMU) to measure body angle and angular velocity which can tell us body position. Last but not least, an encoder will calculate the degree of knee by assembling on the reduction gear. By using these signal as input ,we will train a model which will output an index that tells servo motor how much torque to generate.

摘要 i
Abstract i
目錄 ii
圖目錄 iii
表目錄 iv
第一章 緒論 5
1.1研究動機與目的 5
1.2文獻回顧 5
第二章 輔具設計流程 10
2.1 雙眼視差求腿部姿態 10
2.2 膝關節四連桿設計 14
2.2.1曲線擬合 14
2.2.2規劃四連桿 16
2.2.3目標函數fobj 20
2.2.4差分進化演算法 30
2.2.5最佳四連桿選擇 33
2.2.6四連桿扭力分析 36
第三章 下肢輔具實作 40
3.1 設計目標 40
3.2 硬體架構 40
3.2.1輔具 40
3.3.2肌電訊號儀 46
3.2.3慣性感測器 47
3.2.4磁性編碼器 47
3.2.5微控制器 48
3.2.6伺服馬達 49
3.3 馬達控制器設計 50
3.3.1串聯彈性制動器 50
3.3.2系統鑑別及控制器設計 51
第四章 結論與未來工作 54
4.1 結論 54
4.2 未來工作 54
參考文獻 55

[1] Sankai, Y. (2010). HAL: Hybrid assistive limb based on cybernics. In Robotics Research (pp. 25-34). Springer, Berlin, Heidelberg.

[2] “ReWalk” ReWalk Robotic, [Online]. Available: http://rewalk.com/

[3] Zoss, A. B., Kazerooni, H., & Chu, A. (2006). Biomechanical design of the Berkeley lower extremity exoskeleton (BLEEX). IEEE/ASME Transactions On Mechatronics, 11(2), 128-138.

[4] “eLEGS” Berkeley Robotics & Human Engineering Laboratory,[Online].Available:http://bleex.me.berkeley.edu/research/exoskeleton/elegs%E2%84%A2/

[5] Freudenstein, F., & Woo, L. S. (1969). Kinematics of the human knee joint. Bulletin of Mathematical Biology, 31(2), 215-232.

[6] Blacharski, P. A., Somerset, J. H., & Murray, D. G. (1975). A three-dimensional study of the kinematics of the human knee. Journal of biomechanics, 8(6), 375IN7377-376IN8384.

[7] Acharyya, S. K., & Mandal, M. (2009). Performance of EAs for four-bar linkage synthesis. Mechanism and Machine Theory, 44(9), 1784-1794.

[8] Holland, J. H. (1992). Adaptation in natural and artificial systems: an introductory analysis with applications to biology, control, and artificial intelligence. MIT press.

[9] Kennedy, J. (2011). Particle swarm optimization. In Encyclopedia of machine learning (pp. 760-766). Springer US.

[10] Storn, R., & Price, K. (1997). Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces. Journal of global optimization, 11(4), 341-359.

[11] Zhou, H., & Cheung, E. H. (2001). Optimal synthesis of crank–rocker linkages for path generation using the orientation structural error of the fixed link. Mechanism and Machine Theory, 36(8), 973-982.