Control Approaches For Five Degree Of Freedom Upper-Limb Rehabilitation Training Robotic Systems

With the development of modern rehabilitation medicine, it has been an attractive research focus on applying the robotics-aided training system in nerve centers rehabil-itation. This dissertation attempts to improve the existing upper-limb rehabilitation ap-proaches in clinical treatment on the following aspects:enhancing the rehabilitation value of the training programm, providing more man-machine interaction information, guaran-teeing the stability against disturbance during passive training, improving the comfort and the effectiveness during active training, performing detection and protection for spasms. Based on the rehabilitation medicine principle and advanced control theory, and combin-ing the physician experience with ergonomic design, combining information collection method with man-machine interaction technique, combing the off-line programming and on-line intelligence learning, this dissertation deals with the following problems:the op-timal design of rehabilitation training trajectory, the estimation of toque parameters, the intelligence learning in passive/active training, the detection and emergence control for spasms. A group of robotic aided approaches are to be discussed on upper-limb rehabili-tation treatment.The main research contents are outlined as follows:Based on summarizing the rehabilitation medicine and ergonomic principle, a series of systemic indexes are proposed by integrating the movement complexity, the training strength and the gentle and smooth degree. Then the problem of rehabilitation training trajectory programming is transformed into a problem of nonlinear multi-objective opti-mization, which is solved by a DNA-based genetic algorithm with a strictly satisfactory rehabilitation quality.Considering that the existing impedance toque modeling methods can only provide part information of the end joint, an impedance model is first proposed with three joint five degree of freedom. Based on an integral transformation-based nonlinear observer method, a convergence condition are presented by the quantitative relationship between estimation error with observer parameters, which can guarantees an on-line quick (exponentially convergent) estimation according to a specified accuracy.Consider the unknown disturbance and uncertainties of muscle strength, a logic switching-based self-learning compensation strategy is proposed to guarantee the smooth tracking of the upper arm movement to a pre-given training trajectory. To cope with the output feedback control problem of nonlinear man-machine interaction system in the presence of the.uncertainties, a self-learning mechanism is proposed to ensure the global stability and a high accuracy reliable tracking.To satisfy the requirement for improving the effectiveness by the modification of rehabilitation trajectory, a DNA-based genetic algorithm is proposed self-modification mechanism with an optimal approximation to the given rehabilitation trajectory and a compliance of the upper-arm traction. By on-line information collection of the move-ment parameters and the upper-limb toque characteristics and judging the rehabilitation state, the servo-trajectory is modified through balancing the effectiveness and comfort-ness, which can guarantee the safe accomplishment and optimal modification of the train treatment.To cope with the spasms during training procedure, this dissertation proposes a ex-tended observer-based spasms detection and emergence protection strategy with the im-provement in detection accuracy, detection sensitiveness and protection quickness. By constructing the extended observer, the present method can exactly estimate the perturba-tion of arm torque caused by the muscle tension changes and perform the protection with appropriate torque compensation.