Inverse Dynamics Modeling And Humanoid Control Of Lower Limb Rehabilitation Robot Based On Human Motion Mechanism

With the aging of society,the acceleration of urbanization,and the prevalence of unhealthy lifestyles,the risk factors for cerebrovascular diseases have also generally emerged.Cerebrovascular diseases are often accompanied by limb movement disorders,which cause great inconvenience to human life.Therefore,the recovery of walking ability is the main goal of the patient’s rehabilitation.Traditional rehabilitation therapy mainly relies on manual work or simple medical equipment,which is far from meeting the actual needs of patients.The use of rehabilitation robots can effectively alleviate the shortcomings and deficiencies of traditional treatment with significant advantages.Among them,lower limb rehabilitation robots have gradually become a current research hotspot.Humanoid controllers inspired by human lower limb characteristics play an important role in the research of lower limb rehabilitation robots.However,due to the nonlinear and strong coupling characteristics of the humanoid controller,the accuracy of the inverse dynamics model of the human lower limb rehabilitation robot is still a challenging problem.Therefore,this thesis research on inverse dynamics modeling and humanoid control for lower limb rehabilitation robots based on human motion mechanism,and the main work and research results include:(1)Based on the mechanism of human movement,the movement data of healthy human lower limbs are analyzed.First,through the laboratory’s 3D motion capture system and 3D force measurement platform,the gait data and plantar force data of the lower limb movement of the human body are measured.Then,the collected data are filtered and fitted,and processed into data that can be directly applied to the humanoid controller of the lower limb rehabilitation robot.Finally,a two-link model is established for the lower limb rehabilitation robot,and the dynamics of the lower limb rehabilitation robot is analyzed by the method of Lagrangian analysis.(2)In order to further improve the accuracy of the inverse dynamics model of the humanoid controller of the lower limb rehabilitation robot,a non-parametric modeling method is proposed to learn the inverse dynamics model.The main idea is to use the motion data of the main joints of the lower limbs of healthy people as input,and the corresponding joint torque as output,and learn the inverse dynamics model through the training of the neural network.In order to ensure that the learned model can be used on the humanoid controller of the lower limb rehabilitation robot,all the data collected in this thesis are real data of the lower limb movement of healthy people.In addition,since the data type of the collected data is based on time series,this thesis proposes to use the long-short-term memory network(LSTM)and the gated recurrent unit(GRU)network to learn the inverse dynamics model,and compares the learning effects of the two neural networks.The evaluation index of these two network models in this thesis is Root Mean Squared Error(RMSE).The experimental results show that both networks have a good learning effect,and the GRU network has a stronger learning ability than the LSTM network.(3)The humanoid control of lower limb rehabilitation robot is studied based on human motion mechanism.The main idea is based on the trajectory of the healthy person’s gait as the expected trajectory of the humanoid controller,and the joint torque that needs to be provided by the humanoid controller learned from the inverse dynamics model is used as input,and the PD controller is used to compensate and correct it.Finally,the tracking of the expected input trajectory of the lower limb rehabilitation robot is achieved.The humanoid controller is finally simulated and verified on the MATLAB simulink platform.