Study On Control Strategy Of Upper Limb Rehabilitation Robot In Active And Passive Training Mode

With the aggravation of population aging in China,there are more and more stroke patients.Upper limb rehabilitation robot can provide scientific and effective rehabilitation training for stroke patients.In the research of many upper limb rehabilitation robots,control strategy is the core technology,and its scientificity and man-machine integration affect the rehabilitation effect.In the face of patients in different rehabilitation treatment periods,the design of corresponding control strategies is the main work of this thesis.Aiming at the problem of involuntary muscle tremor and spasm during passive rehabilitation training,a passive strategy of Sliding Mode Control Based on Nonlinear Disturbance Observer(NDO-SMC)is proposed.Aiming at the inaccurate modeling of nominal model during active rehabilitation training,an Adaptive Active Control Based on Radial Basis Function Neural Network(RBFNN-AAC)active strategy is proposed.Aiming at the two problems mentioned above,a comprehensive strategy of Radial Basis Function Neural Network Sliding Mode Control Based on Nonlinear Disturbance Observer(NDO-RBFNNSMC)in active and passive training mode applicable to both active and passive rehabilitation training is proposed.Finally,through simulation verification,the upper limb rehabilitation robot can track the desired training trajectory stably,quickly and accurately.The specific work is as follows:(1)In order to ensure the control and execution of rehabilitation movement,robot kinematics analysis and dynamic modeling analysis are the basis.The kinematics part introduces the rigid body attitude,coordinate system description and homogeneous coordinate transformation in space,and establishes the forward kinematics equation of the upper limb rehabilitation robot by Denavit and Hartenberg(D-H)method.The dynamics part is solved by Lagrange method to obtain the robot dynamics equation that can represent any n joints.In order to understand the upper limb rehabilitation robot more intuitively and vividly,a 7 Degree of Freedom(DOF)upper limb exoskeleton rehabilitation robot is designed through SolidWorks software.(2)Aiming at the problem of introducing system chattering by increasing the switching gain in sliding mode control to eliminate external interference due to spasm and tremor of patients during passive rehabilitation training,a passive control strategy based on NDO-SMC is proposed.A nonlinear disturbance observer is designed to estimate the compound disturbance on-line,and its estimated value is fed back to the torque controller to offset the error value of on-line observation.In the MATLAB/Simulink simulation environment,six groups of comparative simulation experiments are designed.The results show that the proposed passive control strategy based on NDO-SMC has a good observation effect on the compound interference such as muscle tremor and spasm.It not only ensures the on-line estimation of the disturbance value and improves the trajectory tracking accuracy,but also indirectly decreases the switching gain in the sliding mode control and effectively reduces the chattering of the system.(3)Aiming at the problem of model error in using nominal model instead of accurate mathematical model during active rehabilitation training,an active control strategy based on RBFNN-AAC is proposed.Because RBF neural network has the ability to approximate any nonlinear uncertain function and adaptive control,it has the ability of neural network weight convergence.RBF neural network approximates the modeling error online,and the adaptive algorithm adjusts the weight of the hidden layer.Through simulation verification and comparison of the tracking effects under two different control algorithms,it is concluded that the active control strategy based on RBFNN-AAC can better track the desired trajectory,which is significantly better than the uncompensated nominal model control algorithm in tracking accuracy,speed and stability,so as to ensure the effect and efficiency of patients during active rehabilitation training.(4)Aiming at the problems raised during active rehabilitation training and passive rehabilitation training,a comprehensive control strategy based on NDO-RBFNNSMC in active and passive training mode is proposed.In the sliding mode controller,the saturation function sat is used to replace the symbolic function sgn and the acceleration term without θ is designed to ensure that the acceleration signal can be effectively obtained through the differential signal in practical engineering.The stability of the designed controller is proved by Lyapunov stability method.The simulation results show that the proposed algorithm not only has a good observation effect on the interference,and can eliminate the chattering problem caused by the symbolic function in the sliding mode control,but also estimates and compensates the inaccurate modeling of the nominal model,so as to ensure the efficiency and safety of patients and equipment in the process of active training and passive training.