本論文使用長短期記憶神經網路作為回授控制以改善人形機器人的步態穩定度，並將理論套用到人形機器人David Junior III以測試效果。在硬體上，無線傳輸晶片取代了笨重的有線傳輸；高採樣速率ADC轉換器與雙足上之壓力感測器結合以量測壓力中心。嶄新的電腦控制核心與電路板能夠將以往預先決定的步態改為即時運算。為了訓練神經網路，四項特徵包括兩方向的加速度以及壓力中心軌跡被選擇當作訓練資料。相對應的最佳解則由粒子群演算法來尋找。透過不同的最佳化器選擇及結果比較，本篇論文所選擇之LSTM網路架構能夠適當地表現訓練資料間的非線性關係且同時能夠避免過擬合現象。另外為了能夠在機器人上達到即時運算的結果，本論文亦建立了以linux與Ubuntu作業系統為基底之機器人控制系統。本論文實驗先在Webot上模擬驗證方法論之效果，再轉移至實機人形機器人上測試。最後，實驗結果顯示，透過本論文所提方法能夠使人形機器人具備自主調節步態之能力且能夠在極短時間內做出反應，以達到高效之動態回授能力。

This thesis applies long-short term memory (LSTM) to build a feedback system for a bipedal humanoid robot. A newly designed humanoid robot, David Junior III, is built for testing the algorithm proposed. To make transmitting data convenient, a wireless communication chip replaces the bulky handmade wires used in the previous David series. Besides, a high sampling rate ADC combined with pressure sensors ensures correct and sufficient pressure data. A brand new computer core is developed to change gait pattern generation from pre-determined offline to online processes. For training the network, four sequential features including center of pressure (COP) and acceleration are acquired by pressure sensors on both foot soles and inertial measure unit on the trunk. While these features are mapped to the inputs of the network, the output of the network is mapped by the optimal solutions generated from the Particle Swarm Optimization. The structure of the proposed LSTM feedback network is composed of four layers and model is cautiously chosen to avoid over-fitting. To make the robot able to calculate the feedback data online, a fast computation system based on Linux and Ubuntu is built. The effect of the methodology provided by this thesis is first tested on the robot simulation environment Webot, then tested on the real bipedal humanoid robot, David Junior III. The experimental results demonstrate the methodology equips the robot with the ability to self-adjust. Also, the feedback system is proved to be fast and effective.

[1] Website of Pepper, accessed on June 21, 2018. [Online]. Available:
https://www.ald.softbankrobotics.com/en/cool-robots/pepper
[2] Website of DARwIn-OP, accessed on June 21, 2018. [Online]. Available:
http://support.robotis.com/en/product/darwin-op.htm
[3] Website of Robi, accessed on June 21, 2018. [Online] Available:
http://www.hellorobi.co.uk/
[4] Website of Atlas, accessed on June 21, 2018. [Online] Available:
https://www.bostondynamics.com/atlas
[5] Website of HRP, accessed on June 21, 2018. [Online] Available:
http://www.drc-hubo.com/
[6] Website of DRC-HUBO, accessed on June 21, 2018. [Online] Available:
http://www.hellorobi.co.uk/
[7] T. Sato, S. Sakaino, E. Ohashi, and K. Ohnishi, “Walking Trajectory Planning on Stairs Using Virtual Slope for Biped Robots,” IEEE Transactions on Industrial Electronics, vol. 58, no. 4, pp. 1385-1396, 2011.
[8] M. Shimojo, T. Araki, A. Ming, and M. Ishikawa, “A ZMP Sensor for a Biped Robot,” in Proceedings of 2006 IEEE International Conference on Robotics and Automation, Tokyo, Japan, May 2006, pp. 1200-1205.
[9] H. Shin and B. Kim, “Energy-Efficient Gait Planning and Control for Biped Robots Utilizing the Allowable ZMP Region,” IEEE Transactions on Robotics, vol. 30, no. 4, pp. 986-993, 2014.
[10] M. Crisóstomo, A. P. Coimbra and J. Ferreira, “ZMP Trajectory Reference for the Sagittal Plane Control of a Biped Robot Based on a Human CoP and Gait,” in Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, St. Louis, USA ,Oct. 2009, pp. 1588-1593.
[11] P. Sardain and G. Bessonnet, “Zero Moment Point—Measurements From a Human Walker Wearing Robot Feet as Shoes,” IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, vol. 34, no. 5, pp. 638-648, 2004.
[12] P. Sardain and G. Bessonnet, “Forces Acting on a Biped Robot. Center of Pressure—Zero Moment Point,” IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, vol. 34, no. 5, pp. 630-637, 2004.

[13] L. Humphrey, H. Hemami, K. Barin, and A. Krishnamurthy, “Simulated Responses to Support Surface Disturbances in a Humanoid Biped Model with a Vestibular-Like Apparatus,” IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, vol. 40, no. 1, pp. 109-119, 2010.
[14] K. Hwang, J. Lin and K. Yeh, “Learning to Adjust and Refine Gait Patterns for a Biped Robot,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 45, no. 12, pp. 1481-1490, 2015.
[15] M. Claros, J. J. Rodriguez and R. Soto, “Balance Control of a Biped Robot in the Coronal Plane Using Foot Sole CoP Trajectories,” in Proceedings of IEEE International Conference on Control and Automation, Taichung, Taiwan, June 2014, pp. 661-666.
[16] Central Pattern Generators and the Control of Rhythmic Movements, accessed on June 21, 2018. [Online] Available: https://www.ncbi.nlm.nih.gov/pubmed/11728329
[17] J. Or, “A hybrid CPG–ZMP Control System for Stable Walking of a Simulated Flexible Spine Humanoid Robot,” Neural Networks, vol. 23, no. 3, pp. 452-460, 2010.
[18] A. Ijspeert, “Central Pattern Generators for Locomotion Control in Animals and Robots: A review”, Neural Networks, vol. 21, no. 4, pp. 642-653, 2008.
[19] S. Kajita, F. Kanehiro, K. Kaneko, K. Yokoi, and H. Hirukawa, “The 3D Linear Inverted Pendulum Mode: A Simple Modeling for a Biped Walking Pattern Generation,” in Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, vol. 1, Hawaii, USA, Oct. 2001, pp. 239-246.
[20] Y. Hong, B. Lee, and J. Kim, “Command State-Based Modifiable Walking Pattern Generation on an Inclined Plane in Pitch and Roll Directions for Humanoid Robots,” IEEE/ASME Transactions on Mechatronics, vol. 16, no. 4, pp. 783-789, 2011.
[21] G. Yu, Y. Liu and X. Li, “Humanoid Robot Gait Planning based on 3D Linear Inverted Pendulum Model,” Journal of Computer Applications, vol. 32, no. 9, pp. 2643-2647, 2013.
[22] T. Komura, A. Nagano, H. Leung, and Y. Shinagawa, “Simulating Pathological Gait Using the Enhanced Linear Inverted Pendulum Model,” IEEE Transactions on Biomedical Engineering, vol. 52, no. 9, pp. 1502-1513, 2005.
[23] S. Kajita, F. Kanehiro, K. Kaneko, K. Fujiwara, K. Harada, K. Yokoi and H. Hirukawa, "Biped walking pattern generation by using preview control of zero-moment point", 2003 IEEE International Conference on Robotics and Automation (Cat. No.03CH37422).
[24] R. Kelly, V. Santibáñez and A. Loría, Control of robot manipulators in joint space. London: Springer, 2005.
[25] K. Hu, C. Ott, and D. Lee, “Learning and Generalization of Compensative Zero-Moment Point Trajectory for Biped Walking”, IEEE Transactions on Robotics, vol. 32, no. 3, pp. 717-725, 2016.
[26] T.-H. S. Li, Y.-T Su, S.-H Liu, J.-J Hu, and C.-C Chen, “Dynamic Balance Control for Biped Robot Walking Using Sensor Fusion, Kalman Filter, and Fuzzy Logic,” IEEE Transactions on Industrial Electronics, vol. 59, no. 11, pp. 4394-4408, 2012.
[27] M. Talebi and M. Farrokhi, “Adaptive Fuzzy-PD Controller for 3D Walking of Biped Robots,” in Proceedings of IEEE Conference of AI and Robotics, Qazvin, Iran, April 2015, pp. 1-7.
[28] C. Lin and E. Boldbaatar, "Fault Accommodation Control for a Biped Robot Using a Recurrent Wavelet Elman Neural Network", IEEE Systems Journal, vol. 11, no. 4, pp. 2882-2893, 2017.
[29] S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory”, Neural Computation, vol. 9, no. 8, pp. 1735-1780, 1997.
[30] T.-H S. Li, Y.-F. Ho, P.-H. Kuo, Y.-T. Ye, and L.-F Wu, “Natural Walking Reference Generation Based on Double-Link LIPM Gait Planning Algorithm”, IEEE Access, vol. 5, pp. 2459-2469, 2017.
[31] S. Kessentini and D. Barchiesi, “Particle Swarm Optimization with Adaptive Inertia Weight”, International Journal of Machine Learning and Computing, vol. 5, no. 5, pp. 368-373, 2015.
[32] “RoboCup Federation official website”, Robocup.org, 2018, accessed on June 21, 2018. [Online]. Available: http://www.robocup.org/. [Accessed: 22- May- 2018].
[33] e. Systems, "nRF24L01 / 2.4GHz RF / Products / Home - Ultra Low Power Wireless Solutions from NORDIC SEMICONDUCTOR", Nordicsemi.com, 2018, accessed on June 21, 2018. [Online].Available: http://www.nordicsemi.com/eng/Products/2.4GHz-RF/nRF24L01. [Accessed: 22- May- 2018].
[34] Load Cell, accessed on June 21, 2018. [Online] Available:
http://www.playrobot.com/load/723-load-cell-10kg-straight-bar-tal220.html
[35] HX711, accessed on June 21, 2018. [Online] Available:
https://learn.sparkfun.com/tutorials/load-cell-amplifier-hx711-breakout-hookup-guide
[36] Arduino Mega 2560, accessed on June 21, 2018. [Online] Available: https://www.arduino.cc/en/main/arduinoBoardMega2560
[37] Sparkfun.com, 2018, accessed on June 21, 2018. [Online]. Available https://www.sparkfun.com/datasheets/Components/nRF24L01_prelim_prod_spec_1_2.pdf. [Accessed: 21- June - 2018].
[38] STM32-F103ZET6, accessed on June 21, 2018. [Online] Available:
http://www.st.com/web/en/catalog/mmc/FM141/SC1169/SS1031
[39] USB2Dynamixel, accessed on June 21, 2018. [Online] Available: http://wiki.cybedroid.com/index.php?title=USB2Dynamixel
[40] “The leading operating system for PCs, IoT devices, servers and the cloud | Ubuntu”, Ubuntu.com, 2018, accessed on June 21, 2018. [Online]. Available: https://www.ubuntu.com/. [Accessed: 21- June - 2018].
[41] PICO 880GA-I7-4650U, accessed on June 21, 2018. [Online] Available:
http://www.axiomtek.com.tw/Default.aspx?MenuId=Products&FunctionId=ProductView&ItemId=8135&upcat=137
[42] M. Vukobratovic, B. Brovac, D. Surla, and D.Stokic, Biped Locomotion: dynamics, stability, control and application, Springer-Verlag, Berlin Germany, 1990.
[43] M. H. P. Dekker, Zero-Moment Point Method for Stable Biped Walking, Eindhoven University of Technology, Jul. 2009.
[44] J. Ferreira, M. Crisostomo, A. Coimbra, and B. Ribeiro, “Control of a Biped Robot with Support Vector Regression in Sagittal Plane,” IEEE Transactions on Instrumentation and Measurement, vol. 58, no. 9, pp. 3167-3176, 2009.
[45] D. Braun, J. Mitchell, and M. Goldfarb, “Actuated Dynamic Walking in a Seven-Link Biped Robot,” IEEE/ASME Transactions on Mechatronics, vol. 17, no. 1, pp. 147-156, 2012.
[46] C. Liu, D. Wang, and Q. Chen, “Central Pattern Generator Inspired Control for Adaptive Walking of Biped Robots,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 43, no. 5, pp. 1206-1215, 2013.
[47] J. Han, “Bipedal Walking for a Full-sized Humanoid Robot Utilizing Sinusoidal Feet Trajectories and Its Energy Consumption,” Ph.D. dissertation, Dept. Philosophy in Mech. Eng., Virginia Polytechnic Institute and State Univ. , Apr. 2012.
[48] S. Buss and J. Kim, “Selectively Damped Least Squares for Inverse Kinematics,” Journal of Graphics Tools, vol. 10, no. 3, pp. 37-49, 2005.
[49] R. Eglese, “Simulated annealing: A tool for operational research,” European Journal of Operational Research, vol. 46, no. 3, pp. 271-281, 1990.
[50] J. F. Martínez and E. G. Gonzalo, “The PSO family: deduction, stochastic analysis and comparison”, Swarm Intelligence, vol. 3, no. 4, pp. 245-273, 2009.
[51] Y. Shi and R. Eberhart, “Empirical study of particle swarm optimization”, in Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat. No. 99TH8406).
[52] T. Chow and Y. Fang, “A recurrent neural-network-based real-time learning control strategy applying to nonlinear systems with unknown dynamics”, IEEE Transactions on Industrial Electronics, vol. 45, no. 1, pp. 151-161, 1998.
[53] Yu Wang, “A new concept using LSTM Neural Networks for dynamic system identification”, in Proc. 2017 American Control Conference (ACC), 2017.
[54] Keras Documentation, Keras.io, 2018, accessed on June 21, 2018. [Online]. Available: https://keras.io/. [Accessed: 21- June - 2018].