Learning Control Based On RTO For Lower Limb Rehabilitation Robot

With the improvement of human living standards and medical conditions in modern society,the increasing number of people with lower limb movement disorders due to the aging of the population has posed a great challenge to the social medical security system.Compared with traditional rehabilitation treatment methods,lower limb rehabilitation robots can effectively solve the problems of lack of medical resources,low treatment efficiency and lack of quality assurance.At the practical level,the ability to efficiently control the lower limb rehabilitation robot is the basis for its large-scale promotion.However,during the research process of the robot control method,the uncertainty presented in the established model does not perfectly match the actual robot system,which leads to the problem that the controller designed for the model may have unsatisfactory control effect or even the controller is not available at all when applied to the actual robot system.To address the uncertainty problem in the robot control process,this topic is developed from the following aspects.(1)The open source biomechanical simulation software Open Sim is used to analyze and summarize the human lower limb gait pattern,which are used to fit and obtain the normal human gait tracking trajectory.(2)The kinematic model of the lower limb rehabilitation robot is established for the right leg of the human swing phase in the sagittal plane based on the gait pattern using the D-H modeling method.On the basis of kinematics,the Lagrange method is used to establish the dynamic model of the lower limb rehabilitation robot with uncertainty,which is used as a "black box" system to replace the actual robot system to provide the actual operational data for the control process.(3)Considering the highly repetitive characteristics of the rehabilitation training process,the control problem for the lower limb rehabilitation robot is described as a form of RTO(Real-Time Optimization)problem,and the adaptive RTO control method based on the modifier adaptation and the RTO control method enhancing SCFO(Sufficient Conditions for Feasibility and Optimality)are proposed.The modifier adaptation is obtained by finite difference method and BP neural network,which is used to drive the model optimization results to the corresponding optimal set point of the actual system by modifier adaptation RTO method.And the latter directly optimizes the control of the actual robot system by means of its actual operation data,and complements the feasibility and optimality of the traditional RTO control process by adding SCFO theoretical guarantees.(4)MATLAB software is used as the simulation test platform to verify the control effects of the above two control methods.The simulation results show that both control methods can achieve the purpose of accurate tracking of the desired trajectory of the lower limb rehabilitation robot under uncertainty,and have strong robustness and anti-interference capability.Compared with other control methods,they have the advantages of simple controller structure and controller parameter self-tuning,which are more suitable for use in practical context.