Research On The Control Methods For Rehabilitation Assistance With Wearable Lower Limb Exoskeletons

With the background of increasing aging population and injured patients,there is great thirst for novel technical solutions for reducing the burden of the society.As a type of intelligent mechatronical device,the wearable lower limb exoskeletons have shown great potentials in gait rehabilitation and walking assistance,which lead to an attractive research field.The wearable lower limb exoskeletons integrate the techniques of biomechanics,ergonomics,mechanism,electronics,sensors,control and etc.As a kind of wearable device,real-time physical interaction exists between the exoskeleton and the wearer.This leads to strong coupling,high nonlinearity and large uncertainties.Therefore,the stability,accuracy,rationality and effectiveness become crucial issues,which are the main difficulties and challenges in the community of exoskeleton research.This thesis mainly focuses on the control methods and passive/active assistance strategy design for gait rehabilitation with lower limb exoskeletons.The research includes following 4 parts:1.In order to design an applicable lower limb exoskeleton,necessary analysis of ergonomics is conducted,which involves the characteristics of joints and human walking gait.A virtual prototype of wearable lower limb exoskeleton is designed with Solidworks software.The virtual prototype is further modified with some practical considerations during applications.In addition,according to the design and physical configurations,the kinematic model and dynamic model of the exoskeleton are further established,and the inverse kinematics is verified with simulation in MATLAB software.The work of this part provides the basis for further study on trajectory tracking and assistance strategy.2.For the trajectory tracking control of lower limb exoskeleton,the coupling humanexoskeleton dynamics is firstly analysed according to the configurations of the virtual prototype.By utilizing the acquired model information from 3D design software Solidworks,the computed torque control(CTC)is adopted for achieving trajectory tracking.In order to improve the adaptability and robustness of the controller,time-delay estimation technique(TDE)and Radial Basis Function(RBF)neural networks(RBFNN)theory are integrated into the CTC structure,forming a final TDE and RBFNN-based CTC method(RBF-TDE-CTC).Furthermore,for getting rid of the dependency on dynamic modeling,a model-free controller(MFC)is developed based on the "ultra-local model" concept.By ultra-local modeling,the original second-order human-exoskeleton dynamics is reconstructed to be a first-order form.The TDE is used to estimate the unknown lumped total disturbance in the reduced-order dynamics.Thus a TDE-based intelligent PI controller(iPI)can be realized.By combining terminal sliding mode theory(TSMC)to the iPI control structure,an adaptive model-free controller(aTSMC-iPI)is finally constructed and applied to the trajectory control of lower limb exoskeleton.The tracking errors can theoretically converge to zero in finite time.Both of the trajectory control methods above are verified by comparative co-simulation in Solidworks-MATLAB/Simulink.The simulation results demonstrate that both controllers can achieve stable and accurate tracking control.3.For passive rehabilitation assistance,a ZMP trajectory tracking control-based method is proposed.Based on the dynamic analysis and ultra-local modeling,a digital model-free trajectory controller called intelligent PD controller(iPD)is developed.For the realization of real-time digital control,the human-exoskeleton dynamics is firstly re-formulated with discrete-time "ultra-local model".The unknown lumped total disturbance is estimated by a linear discrete-time extended state observer(LDESO),which can release the requirement of high derivatives in TDE method.With the compensation of LDESO and zero-pole placement method,the LDESO-iPD controller is finally constructed.The control performance is firstly verified with simulation study.With a hardware-in-loop control platform dSPACE supporting the real-time computing and control,an assistive lower limb exoskeleton is established according to previous design.By applying a group of well-planned rehabilitation gait trajectories,trajectory control and gait rehabilitation experiments are conducted.The tracking accuracy and stability of the proposed LDESO-iPD is verified.4.For active rehabilitation assistance,a torque control-based assistance method for correcting asymmetric gait is developed with a torque-controllable nonlinear Series Elastic Actuator(nSEA).The target task is specified as correcting the gait asymmetry.In order to achieve accurate and stable torque control,a TDE and sliding mode control(SMC)based model-free torque controller(SMC-TDE-iPI)is developed with "ultra-local model"concept.Then for the correction training of asymmetric gait,the assistance torque is generated obeying an optimal principle.The optimization procedure is operated according to the walking performance of previous gait.Therefore,the torque optimization-based assistance(TOA)method can be realized.The torque control performance of the SMCTDE-iPI is firstly evaluated by comparative experiments both in grounded condition and wearing condition.The effectiveness and online adaption of the TOA method is further verified by assistance experiments.