Active Compliance Control Of A Lower-limb Robotic Exoskeleton Through A Human-machine Force Interactive Interface Based On An Oscillator Model Of EMG

At present,the prevailing control strategies for exoskeletons（force control,position control or force-position hybrid control）suffer from motion lag by theory,making them hard to achieve the true integration of human and machine.Therefore,in order to solve the lag problem,academia has attempted to introduce electromyography（EMG）signal into exoskeleton system as the control signal.Although some progresses have been made,it remains unclear about how to select proper EMG-based features,and how to introduce these features into the control system reasonably.Aiming at solving this difficulty,this study is dedicated to first making clear the biomechanical mechanism of muscular contraction and joint movements of human body.Next,it obtains the corresponding cybernetic model serving real-time engineering application.Finally,the active compliance control of lower limb exoskeleton is implemented using knee joint as the object.In order to try to efficiently extract correct activation degree of muscle from EMG,one needs to start from the origin of EMG,i.e.,the action potential（AP）fired by motoneuron,thus studying first the cybernetic mechanism regarding how action potential regulates and drives muscular contraction.Based on that,the dissertation discusses how to establish a new biomechanical model geared to the needs of real-time prediction of muscle force.Further,it makes clear the cybernetic connotation concerning the process from EMG features to force generation,as well as the input-output relations among all the physical modelling quantities of the model.Secondly,starting from the biomechanical principles,the dissertation attempts to develop a novel EMG feature extraction method based on a physical model of EMG,thus enabling it to be used in real-time characterization of muscle’s activation level.Finally,based on the new biomechanical model of skeletal muscle,the coupling dynamics between human and exoskeleton,and the production mechanism of human-machine interactive force are analyzed.Then,by applying online intelligent control algorithms,an adaptive human-machine force interactive interface is built,and the real-time and accurate prediction of human knee joint and the active compliant control of exoskeleton are implemented.To conclude,the main research contents and achievements of this dissertation are listed as follows:1)The bio-electrochemical variable-frequency regulation mechanism of muscular contraction.Taking excitation-contraction coupling（ECC）as the object,a bioelectrochemical model is presented for the activation of action potentials on sarcolemma and variation of myoplasmic Ca²⁺concentration([Ca²⁺]).The control mechanism of muscular contraction is elucidated from the perspective of variable-frequency regulation,and action potential with variable frequency is proposed as the control signal to directly regulate(Ca²⁺)and indirectly control muscle force.Finally,the experimental research verifies initially the proposed variable-frequency control mechanism based on the relation between the characteristic frequency of EMG and the dominating firing rate of relevant motor units.2)The semiphenomenological model for muscular contraction.Focusing on the challenges in implementing real-time predictions for contraction statuses,this thesis proposed a decoupled scheme of the links involved in the working process of a sarcomere and established a semi-phenomenological model integrating both linear and nonlinear frames.In order to facilitate engineering application and cybernetics,the proposed model contains a reduced number of parameters and no partial differential equation,making it highly concise and computationally efficient.Through the simulations of various contraction modes including isometric and isotonic contractions,etc.,the effectiveness of the model is validated.3)The oscillator model of EMG signal and the energy kernel characterization method for muscle’s excitation level.This thesis presents a new method for estimating the excitation level and the characterization of muscle’s intrinsic property.The method,called the energy kernel method,starts with converting the EMG signal into planar phase portraits,on which the elliptic distribution of the state points is named as the energy kernel.Based on such stochastic features of the phase portraits,it approximates the EMG signal within a rectangular window as a harmonic oscillator（EMG oscillator）.The study establishes initially the relationship between the control signal（EMG）and output signal（force/power）from the perspective of biomechanics,and a characteristic energy is proposed to estimate the muscle force.4)The generation mechanism of human-machine interactive force and the active compliant control of lower limb exoskeleton.Aiming at late-phase rehabilitation when“patient-in-charge”mode is expected,this dissertation proposed an iterative prediction-compensation motion control scheme for an exoskeleton knee joint.Based on the analysis of human-machine interactive mechanism（HMIM）and the semiphenomenological biomechanical model of muscle,an online adaptive human-machine force interactive interface（predictor for limb motion）is designed using a focused time-delay neural network（FTDNN）with the inputs of electromyography（EMG）,position and interactive force,where the activation level of muscle is estimated from EMG using the energy kernel method.The compensating controller is designed using the normal force-position control paradigm.Initial experiments validated the effectiveness of the proposed control scheme on improving the compliance of the human-machine integrated knee system.