本論文旨在介绍並描述一種新型免用扭力感測器的動力輔助系統電動輪椅（PAWC），它比起目前市面上的輪椅有顯着更低的成本和更高的性能。本論文所提出的PAWC是建立在利用Lyapunov定理作稳定性分析與驗證的扭力估測器(Torque Observer)。本設計基本上是使用一般市面上很容易獲得的馬達編碼器(Encoder)作為速度感測器，它的基本構造是具有两個傳統的商業用途之光耦合器(Opticoupler)充當速度感測器，每一個編碼器均控制在兩次脈衝寬度調制中拾取馬達轉數產生的脈衝數來感測輪子的速度。除此之外，唯一的输入是由馬達驅動器產生的反饋電壓。這是一種新的自適應扭力估測器，估計使用者施力產生的扭力矩和預測PAWC系统的狀態，然後輸出一個控制信号，用于通過一個數位信號處理器的比例 – 積分(PI)控制器。先將這些信號透過PC電腦模擬，再將實際電氣信號透過數位信號處理器(DSP)晶片的模擬器和實驗平台，在實驗室建構動力輔助輪椅的雛型，以進行雛型驗證。本方法具有良好的性能、符合安全規定和稳定的動力輔助輪椅PAWC設計的要求。此外，本論文所提出的扭力估測器比傳统的扭力估測器更能準確的觀測發現系統的扭力矩。當估測到系統扭力相關參數後，依據估測到的參數給予施力者電動輔助力，在輸出輔助力方面為了能夠準確即時判斷施力者的需求，將前述估測到的參數透過人工智慧之模糊-類神經網路(FNN)求出適合的動力放大系數G。
最後，提出一個模糊類神經網路控制系統，用以決定助力G值，並降低系統輸出的動態誤差，其結果與傳統的PID控制系統及固定G值相比有較佳的暫態收斂速度及較穩定的速度調配。尤其是，模糊類神經網路控制方案的提出產生相同的淨扭矩力輔助，保證了系統的穩定性，舒適和用戶的安全。 有了這次的研究成果後，我們打算研究讓輪椅使用者的肌電圖信號即時輸入系統以檢測過程中的體能和肌肉疲勞程度的變化，以此為基礎，提供輪椅動力輔助及研究方法的改進。然後，將被擴展和增強的理論觀點製成，再一次提出更新PAWC系統。特別是，考慮透過無線通訊技術，實現在線版本(On-Line)允許遠程關懷，遠程治療和遠程護理照護。
本論文所提出的免用扭力感測器動力輔助電動輪椅PAWC展示出比傳統電動輪椅更好的性能，較低的生產成本和卓越的维修成效，因此比一般電動輪椅或動力輔助電動輪椅更有效能更能展現其優勢，為下一代動力輔助電動輪椅PAWC，提出一個全新的系统概念，將有可能是未來輪椅建構的候選機種。

ABSTRACT
The objective of this paper is to present, summarize, technically describe, experimentally verify and finally discuss possible future developments with regard to a novel sensorless power-assisted wheelchair (PAWC). The basic goal of a PAWC is to improve upon certain of the inherent limitations and disadvantages of traditional manual wheelchairs. The classical and time honored manual wheelchair is a well known medical assist around the world’s medical treatment establishments and private facilities for care of aged or physically convalescent persons. The fully powered alternative designs are in line with the modern development of electric vehicles. Although the utility and convenience of the fully automatic wheelchair cannot be denied, the relation of the fully electric wheelchair to optimal patient care requirements and maximally reduced rehabilitation times requires careful consideration. There is a well known phrase of the medical universe that tells us, “Use it or lose it.” This is a therapeutically meaningful saying that is known around the world, known even unto ancient times, with some analogous version being found in all known languages. The basic point is that a recuperating limb or similar organ needs use and exercise to return to full function. The fully automatic wheelchair, however, prevents such use. Awareness of this has led to various devices and systems which provide a degree of expedited mobility, but which in addition require a degree of personal exercise on the part of the patient. The original mechanical wheelchair is one of the earlier versions of such systems. Unfortunately, the traditional mechanical wheelchair is a mechanically inefficient device which, especially if use for a long time, has long term negative effects on the joints of the user. Thus therapists emerged with the concept of the power-assisted wheelchair, a class of wheelchair-type device which can provide full automatic mobility but, for the sake of therapeutic considerations, is designed to hold back from full power support. Instead, it provides a degree of mobility assistance. Early PAWC designs were simply powered wheelchairs which were controlled by the users via some sort of interactive control system, e.g. touch-sensitive wheels. Over the years of PAWC development, utilization of active participation on the part of the patient in powering the controlling the PAWC device has become more sophisticated. Current work is oriented toward systems which modify the requirements of patient participation according to the recovery progress of the patient. This is trending toward telemedical applications. Certainly a number of alternative approaches have been presented in recent years. They have typically involved complex and expensive circuitry and systems. The PAWC designs presented in this present study, however, are of significantly lower cost and higher performance than previous designs. The proposed PAWCs are presented as two alternative models. The first alternative is a more readily understandable design that uses a PI control loop. Of the two alternatives, it represents a simpler and cheaper version. It is also much more limited in terms of its evolutionary potential. Nevertheless, this earlier prototype was found to be completely acceptable in our basic laboratory experiments. The design is essentially sensorless, having only two conventional commercial opticoupler speed sensors, one for each of the two pulse width modulated motors. The only other input is voltage feedback produced by the motor drivers, making the system inexpensive and easy to manufacture. The novel adaptive torque observer estimates the impact torque and predicts the system state, outputting a control signal for the PI controller via a digital signal processor (DSP). For safety and reliability, the control loop for the system is built around a torque observer using the Lyapunov stability theorem. Experiments were conducted by simulation and also by use of a lab-built prototype. The presented experimental results will verify that the proposed PI-based design has good performance, successfully meeting laboratory requirements for safe and stable PAWC behavior. Further, the proposed torque observer is found more accurate than a conventional torque observer. Because the proposed sensorless PI-PAWC demonstrates superior performance, lower intrinsic cost and enhanced probability of maintenance-free operation than conventional PAWCs, the presented PI-based system is a likely candidate for next-generation PAWCs.
However, the PI-based PAWC is a relatively primitive device in terms of the ongoing digital revolution. Thus, the second alternative design presented in this study is built around a fuzzy neural network (FNN) which provides a more sophisticated method for determination of the momentary power assist factor G. Compared with the traditional PI control system’s determination of G, the FNN control system’s technical behavior results in reduced tracking error of both the reference model and plant output position. The proposed FNN control scheme yields essentially the same net torque assist as the PI control scheme. This guarantees system stability, user security and patient comfort for both control schemes. The FNN control system however is an extendable system which can more easily be expanded to interface with a wider range of additional digital devices.
Finally, this study concludes with discussion of the future possibilities of PAWC systems. It will be pointed out that all contemporary or near-future robotic/mechatronic patient-assist systems are intrinsically of limited product design life. Genetic engineering and related developments are expected in the foreseeable future to make obsolete all contemporary medical fields. Certainly, the speed of arrival of such technology is of great debate. Nevertheless, Moore’s observation that the number of transistors on integrated circuits doubles every two years will find an analog in bio-genetic technology. Accordingly, the trans-PAWC future should arrive reasonably quickly. However, as was the truth with early post-transistor digital electronics, the early stages of the developmental curve can be expected to be relatively slow because the various necessary collateral technologies need time to develop and coalesce. Thus, for the immediate practical future, PAWCs and similar devices will no doubt have an important role to play. This study is thus directed to the enhancement and optimization of that future, and is expected to help the transition to a future that is essentially independent of what we currently consider to me medicine.

REFERENCES
[1] L. Silva, T. Felzer, G. Neto, R. Nordmann and F. Dedini, Modeling a hands-free controlled power wheelchair, Lecture Notes in Computer Science, vol. 5105,pp.1261-1268, 2008.
[2] S. Katsura and K. Ohnishi, Human cooperative wheelchair for Haptic interaction based on dual compliance control, IEEE Trans. Ind. Electronics, vol.51, no.1, pp.221-228, 2004.
[3] D. Ding and R.A. Cooper, Electric powered wheelchairs, IEEE Control Syst. Mag.,vol.25, no.2, pp.22-34, 2005.
[4] T. Yasuda, K. Nakamura, A. Kawahara and K. Tanaka, Neural network with variable type connection weights for autonomous obstacle avoidance on a prototype of six-wheel type intelligent wheelchair, International Journal of Innovative Computing, Information and Control, vol.2, no.5, pp.1165 -1177, 2006.
[5] M. Athanasiou and J. Clark, A Bayesian network model for the diagnosis of the caring procedure for wheelchair users with spinal injury, Comput. Meth. And Programs in Biomedicine, vol.95, no.2, pp.S44-S54, 2009.
[6] K. L. Su, S. H. Chia, J. H. Guo and B. Y. Li, Develop a power predictive system for scheduling loading of mobile robots, ICIC Express Letters. vol.5, no.8(B), pp.2791-2798, 2011.
[7] M. Nash, D. Koppens, M. V. Haaren, A. Sherman, J. Lippiatt and J. Lewis, Power-assisted wheels ease energy costs and perceptual responses to wheelchair propulsion in persons with shoulder pain and spinal cord injury, Arch Phys Med Rehabil., vol.89, pp.2080 -2085, 2008.
[8] P. Giacobbi, C. Levy, F. Dietrich, S. Winkler, M. Tillman and J. Chow, Wheelchair users' perceptions of and experiences with power assist wheels, Am. J. Phys. Med.
Rehabil., vol.89, no.3, pp.225-234, 2010.
[9] S. Tong, W. Zhang and T. Wang, Robust stabilization conditions and observer-based controllers for fuzzy systems with input delay, International Journal of Innovative
Computing, Information and Control, vol.6, no.12, pp.5473-5484, 2010.
[10] H. Lakomy, I. Campbell and C. Williams, Treadmill performance and selected physiological characteristics of wheelchair athletes, Br. J. Sports Med., vol.21, no.3, pp.130-133, 1987.
[11] R. Glaser, M. Sawka, L. Laubach and A. Suryaprasad, Metabolic and cardiopulmonary responses to wheelchair and bicycle ergometry, J. Appl. Physiol., vol.46, no.6, pp.1066 -1070,1979.
[12] R. Cooper, T. Corfman, S. Fitzgerald, M. Boninger, D. Spaeth, W. Ammer and J. Arva, Performance assessment of a pushrim-activated power-assisted wheelchair control system, IEEE Trans. Control Syst. Technol., vol.10, no.1, pp.121-126, 2002.
[13] H. Gellman, D. Chandler, J. Petrasek, I. Sie, R. Adkins and R. Waters, Carpal tunnel syndrome in paraplegic patients, 6TJ Bone Joint Surg6T. vol.70, no.A(4), pp.517-519,1988.
[14] P. Nichols, P. Norman and J. Ennis, Wheelchair user’s shoulder? Shoulder pain in patients with spinal cord lesions, Scand J Rehabil Med., vol.11, no.1, pp.29 -32, 1979.
[15] L. van der Woude, E. Botden, I. Vriend and H. Veeger, Mechanical advantage in wheelchair lever propulsion: effect on physical strain and efficiency, J. Rehabil. Res. Dev., vol.34, no.3, pp.286 -289, 1997.
[16]R. Cooper, S. Fitzgerald, M. Boninger, K. Prins, A. Rentschler, J. Arva and T.O'Connor,12T Evaluation of a push rim-activated, power-assisted wheelchair, Arch. Phys. Med. Rehabil., vol.82, no.5, pp.702 -708, 2001.
[17] A. Kakimoto, H. Matsuda and Y. Sekiguchi, Development of power-assisted attendant-propelled wheelchair, 9TInt J Jpn Soc Precis Eng., vol.65, no.8,pp.1126-1130, 1999.
[18] Y. Liu and Z. Wang, A prediction-based data collection method in wireless sensor network using Kalman filter, ICIC Express Letters Part B: Applications, vol.2, no.6, pp.1439 -1446, 2011.
[19] L. Yao and Y. Chen, A Type-2 fuzzy controller for automatic guided vehicle wall-following control, ICIC Express Letters Part B: Applications, vol.1, no.1, pp.77-83, 2010.
[20] A. Takahashi and R. Oguro, Torque and speed estimation with parameter identification of line-start induction motor operated value, International Journal of Innovative Computing, Information and Control, vol.5, no.12(A), pp.4551-4559, 2009.
[21] T. Saboori, M. Menhaj and B. Karimi, A nonlinear adaptive observer based on neural networks for nonlinear systems including secure communication problems, International Journal of Innovative Computing, Information and Control, vol.6, no.11,pp.4771-4782, 2010.
[22] H. Chuang, Y. Chuang and C. Chen, Development of two-quadrant PMDC motor drives with regenerative braking on electric vehicles, ICIC Express Letters, vol.5, no.9(B), pp.3321 -3328, 2011.
[23] D. Luenberger, An introduction of observer, IEEE Trans. Autom. Control, vol.16, no.6,pp.596 -602, 1971.
[24] B. Shafai, C. Pi, S. Nork and S. Linder, Proportional integral adaptive observer forparameter and disturbance estimations, IEEE Trans. Ind. Electronic, vol.4, no.1, pp.4694-4699, 2002.
[25] H. J. Ho and T. C. Chen, Motorized CPM/CAM physiotherapy device with sliding-mode fuzzy neural network control loop, Comput. Meth. Programs Biomed,Vol. 96, pp. 96-107, 2009.
[26] R. Cooper, T. Fletcher-Shaw, R. Robertson, Model reference adaptive control of heart rate during wheelchair ergometry, IEEE Trans. Control Syst. Technol., vol.6, no.4,
pp.507-514, 1998.
[27] Y. Oonishi, O. Sehoon and Y. Hori, A New control method for power-assisted wheelchair based on the surface myoelectric signal, IEEE Trans. Ind. Electron., vol.57, no.9, pp.3191-3196, 2010.
[28] S. Choi, J. Ko, I. Kim, J. Park and S. Hong, Precise position control using a PMSM with a disturbance observer containing a system parameter compensator, Proc. Electric Power Appl., vol.152, no.6, pp.1573-1577, 2005.
[29] N. Li, Y. Han and C. He, Capacitive sensor design for an automatic car-wiper system,ICIC Express Letters Part B: Applications, vol.2, no.4, pp.879-884, 2011.
[30] L. Shiu, C. Lee and C. Yang, Obstacle-free robot deployment algorithm for wireless sensor networks, ICIC Express Letters Part B: Applications., vol.2, no.2, pp.325-330,2011.
[31] S. Zhang, G. Wu and S. Zheng, Speed following driver model based on self-adaptive Chaotic Genetic PID and preview control, ICIC Express Letters, vol.5, no.11,pp.4041-4046, 2011.
[32] T. J. Ren and T. C. Chen, Modeling and control of a power-assisted mobile vehicle based on torque observer, IET Control Theory Appl., vol.1, no.5, pp.1405-1412, 2007.
[33]T. J. Ren and T. C. Chen, “Robust speed-controlled induction motor drive based on recurrent neural network,” Electr. Power Syst. Res., vol. 76, pp. 1064-1074, 2006.
[34]C. H. Wang, H. L. Liu and T. C. Lin, “Direct Adaptive Fuzzy-Neural Control With State Observer and Supervisory Controller for Unknown Nonlinear Dynamical Systems,” IEEE Trans. Fuzzy Syst., vol. 10, pp. 39-49, 2002.
[35]T. C. Chen and T. T. Sheu, “Model Reference Robust Speed Controlled for Induction-Motor Drive With Time Delay Based on Neural Network,” IEEE Trans.Syst. Man Cybern. Part A Syst. Hum., vol. 31, no. 6, pp. 746-753, 2000.
[36]H. J. Ho and T. C. Chen, “Implementation of CAM/CPM physiotherapy device with a virtual spring,” Comput. Biol. Med., vol. 38, pp. 923-930, 2008.
[37] D. R. Ankrum, “Question to ask when interpreting surface electromyography (SEMG)research,” Proceedings of the IEA 2000/HFES congress, vol. 5, pp. 530-533, 2000.
[38]R. Riener, M. Frey, T. Proll, F. Regenfelder and R. Burgkar, “Phantom-based multimodal interactions for medical education and training: the Munich knee joint simulator,” IEEE Trans. Inf. Technol. Biomed., vol. 8, no. 2, pp. 208-216, 2004.
[39]D. E. Toutoungi, T. W. Lu, A. Leardini, F. Catani and J. J. O’Connor, “Cruciate ligament forces in the human knee during rehabilitation exercises,” Clin. Biomech., vol. 15, no. 3, pp. 176-187, 2000.
[40]G. M. Ginty, J. J. Jrrgang and D. Pezzullo, “Biomechanical considerations for rehabilitation of the knee,” Clin. Biomech. vol. 15, pp. 160-166, 2000.
[41]J. C. Upper, B. L. Ramage, D. T. Corr, D. A. Hart and J. L. Ronsky, “Measure knee joint laxity: A review of applicable models and the need for new approaches to minimize variability,” Clin. Biomech., vol. 22, no. 1, pp. 1-13, 2007.
[42]R. R. Neptune and S. A. Kautz, “Knee joint loading in forward versus backward pedaling: implications for rehabilitation strategies,” Clin. Biomech., vol. 15, no. 7,
pp. 528-535, 2000.
[43]S. Jung and T. C. Hsia, “Neural Network Impedance Force Control of Robot Manipulator,” IEEE Trans. Ind. Electron., vol. 45, pp. 451-462, 1998.
[44]M. Zerikat, M. Bendjebbar and N. Benouzza, “Dynamic Fuzzy-neural Network Controller for induction Motor Drive,” Word Academy of Science, Engineering and Technology, vol. 10, pp. 278-283, 2005.
[45]G. N. Reising and M. H. Daniels, “A Study of Hogan’s Model of Counselor Development and Supervision,” Journal of Counseling psychology, vol. 30, No. 2, pp.235-244, 1983.
[46]Y. Izuno, R. Takeda, and M. Nakaoka, “New fuzzy reasoning-based high performance speed/position servo control schemes incorporating ultrasonic motor,” IEEE Trans. on Industry Applications, vol. 28, no. 3, pp. 613-618, May 1992.
[47]A. Alessandri, C. Cervellera, and M.Sanguineti, “Design of asymptotic estimators: an approach based on neural networks and nonlinear programming,” IEEE Trans. on Neural Networks, vol. 18, no. 1, pp. 86-96, January 2007.
[48]Standard neural network models for control systems,” IEEE Trans. on Neural Networks, vol. 18, no. 5, pp. 1376-1391,September 2007.
[49]V. M. Becerra, F. R. Garces, S. J. Nasuto, and W. Holderbaum, “An efficient parameterization of dynamic neural networks for nonlinear system identification,”
IEEE Trans. on Neural Networks, vol. 16, no. 4, pp. 983-988, July 2005.
[50]C. M. Lin, and C. F. Hsu, “Recurrent neural network based adaptive -backstepping control for induction servomotors,” IEEE Trans. on Industrial Electronics, vol. 52, no.6, pp. 1677-1684, December 2005.
[51]C. C. Ku and K. Y. Lee, “Diagonal recurrent neural networks for dynamic systems control,” IEEE Trans. on Neural Networks, vol. 6, no. 1, pp. 144-156, 1995.
[52]M. A. Brdys and G. J. Kulawski, “Dynamic neural controllers for induction motor,”IEEE Trans. on Neural Networks, vol. 10, no.2, pp. 340-355, 1999.
[53]T. W. S. Chow and Y. Fang, “A recurrent neural-network-based real-time learning control strategy applying to nonlinear systems with unknown dynamics,” IEEE Trans.on Industrial Electronics, vol. 45, no. 1, pp. 151-161, 1998.
[54]C. J. Lin and C. H. Chen, “Identification and prediction using recurrent compensatory neuro-fuzzy systems,” Fuzzy Sets and Systems, vol. 150, no. 2, pp. 307-330, 2005.
[55]F. J Lin and P.H Shen, “Adaptive Fuzzy-Neural-Network Control for a DSP-Based Permanent Magnet Linear Synchronous Motor Servo Drive,” IEEE Trans. on Fuzzy Sets and Systems, vol. 14,no. 4, pp.481-495, August 2006
[56]R. J. Wai, Y.C. Huang, Z.W. Yang, C.Y. Shih, “Adaptive fuzzy-neural-network velocity sensorless control for robot manipulator position tracking,” IET Control Theory Appl., vol. 4, no. 6, pp. 1079–1093, June 2010.
[57]C. H. Lin “Adaptive Backstepping FNN Control for a Linear Synchronous Motor Drive,” IEEE Trans. on Fuzzy Sets and Systems, vol. 14, no. 4, pp.494-499, August 2009
[58]S. Ahmad, M.O. Tokhi, S.F. Toha, “Genetic Algorithm Optimisation for Fuzzy Control of Wheelchair Lifting and Balancing,” Third UKSim European Symposium on Computer Modeling and Simulation, pp.97-101, June 2009.
[59] S. Ahmad and M. O. Tokhi, “Fuzzy Logic Control of a Wheelchair on Two Wheels,"presented at 27th IASTED International Conference on Modeling, Identification and Control (MIC 08), Innsbruck, February 2008.
[60] S. Ahmad and M. O. Tokhi, “Modelling and Control of a Wheelchair on Two Wheels,” presented at 2nd Asia International Conference on, IEEE computer Society, pp. 579-584, 2008 Modelling & Simulation (AMS 08), Kuala Lumpur, 2008.
[61] D. A. Winter, Biomechanics and Motor Control of Human Movement, New York:Wiley-Interscience, 1990.
[62] S. Oh, N. Hata, and Y. Hori, “Control Development for Wheelchairs in Slope Environments,” 2005 American Control Conference, Portland, OR, USA, 2005.
[63] M. Ohkita, R. Tamanaha, M. Okugumo, J. Tanaka, M. Ohki, and A. Kimura,“Traveling control of the autonomous mobile wheel-chair DREAM-3 considering correction of the initial position,” The 2004 47th Midwest Symposium on Circuits and Systems, 2004.
[64] M. J. Lawn and T. Ishimatsu, “Modeling of a stairclimbing wheelchair mechanism with high single-step capability,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 11, pp. 323-332, 2003.
[65]http://www.engineersaustralia.org.au/sites/default/files/shado/Learned%20Groups/National%20Committees%20and%20Panels/Engineering%20Design/Part%206.pdf