The Research On Some Key Technologies Of Hand Rehabilitation Robot System

With the increasing number of the hemiplegia patients, attention has been drawn to the rehabilitation therapy of the movement functional disorder. Rehabilitation of the movement function has influenced on patient's daily life greatly. In nowadays, this topic is one of the most difficult problems in rehabilitation therapy and has become a hotspot of modern rehabilitation and medical studies. Rehabilitation robot research was born as a new study area due to the great demand in rehabilitation of the movement function. It involves robotics, rehabilitation medicine, ergonomics, mechanical design, control theory, computer science and other fields.Under the support grants from National 863 project and National Natural Science Foundation, the research work in this dissertation focus on investigation some key technologies of wearable robot for rehabilitation of hand function. Our study includes mechanical and control system design of rehabilitation robot, modeling and control of pneumatic muscle actuator, identification of hand's surface electromyography (sEMG) signal, and so on. The main contributions of this dissertation focus on the following aspects:First, the existing rehabilitation equipments for hand function are analyzed and compared. The structure, function and motion constraints of human hand are analyzed and biomechanical model of representative index finger is also investigated. Based on the investigations, the mechanical design of the hand rehabilitation robot is studied. Considering the needs of clinical rehabilitation, the wearing comfort, as well as safety and flexibility of motion, we design a mechanism which can assist fingers to complete multi-joints functional rehabilitation movements. The exoskeleton structure is adopted and the flexible PMs are selected as its actuator. A control system platform of the hand rehabilitation robot is built and the basic control experiments are also implemented to verify the effectiveness of the mechanical and control system. Furthermore, the improved mechnical design is also proposed, which lays a good foundation for realizing the clinical use of the hand rehabilitation robot.Two different structural actuators consisting of the pneumatic muscle (PM) which is work as rehabilitation robot's actuator are analyzed and compared. The work indicates the proposed actuator configuration consisting of the PM and the torsion spring is an effective and promising method for rehabilitation robot application. The complex nonlinear dynamics and time-varying parameters of the PM make it difficult to control. Thus, the modeling and control of PM are also investigated in this dissertation. We identify three-element model of the PM by experiments. Based on this model, the sliding mode controller is designed. Considering the unmodeled dynamics and uncertain disturbance in the system, the disturbance observer is adopted to compensate for these uncertain nonlinear disturbances. The simulation and experimental results demonstrated that the model and the controller achieved the desired performance in tracking a desired trajectory within guaranteed accuracy regardless of modeling uncertainties and perturbation. The PM made by different workmanship and material has different dynamics. To get more convenient modeling and control method, a new type of recurrent neural networks-Echo Neural Network (ESN) is proposed to model the PMs. The adaptive controller is developed based on ESN. The effectiveness of the proposed method is validated by simulation and experiments.The other important part of the rehabilitation robot is the sEMG signal feedback, which can be used to evaluate the effectiveness of the rehabilitation and recognize the motion intension. In order to fulfill the sEMG signal recognizing task, sEMG signal of elbow flexion, elbow extension, wrist pronation, wrist supination, palm extension and fist are collected. Then discrete wavelet transform is used to extract the signal's features. The Support Vector Machines (SVM) is used to classify these features. Compared with the neural network classifier, the experimental results show that SVM is more effective for small-scale sample classification. Besides having the advantages of neural network, which can extract the classification information and select features automatically, SVM also has better generalization ability and is easy to apply and control. Furthermore, it overcomes the disadvantages of artificial neural network such as overlearning and partially leading to minimum.Finally, the summary of this dissertation and the directions of future work are presented.Some key issues on hand rehabilitation robot system are investigated deeply in this thesis. These works provide the necessary theoretical basis, experimental data and valuable research experience for development of the clinic-oriented hand rehabilitation robot. With improvement of the further research work, some products on hand rehabilitation robot will be realized. These products can provide more patients effective rehabilitation training and improve the quality of rehabilitation. This is a research work of positive academic and practical significance.