| The human walking motion is achieved by highly coupled motion of bilateral lower limbs.However,lower limb dysfunction brings a great limitation to the normal activities of patients and a heavy burden of life and psychological pressure to patients and their families.The lower limb exoskeleton,as a wearable device,can help patients rebuild bilateral lower limb coupling motion ability and restore or even replace the motion function of the disabled limb.Aiming at the rehabilitation training needs of patients with unilateral lower limb movement disorders,the unilateral powered lower limb exoskeleton was used as the study object.The research goals were to achieve the synergistic movement between the lower limb exoskeleton and the human body and to realize the coupled movement of the healthy and affected lower limbs.The normal motion function of one limb of the patients was fully utilized.Based on the motion information and gait trajectory of the patient’s healthy lower limb,the exoskeleton drives the disabled limb to achieve coordinated following motion.The specific studies are as follows:(1)A method for characterizing motion intention based on relative changes in gait parameters of the healthy and affected lower limbs was proposed.The problem of motion intention characterization ambiguity of conventional gait parameters within the swing phase was solved.The intention recognition speed was improved.To address the problem that conventional gait parameters were difficult to accurately characterize motion intention in real time,the motion characteristics patterns and spatial position changes of bilateral lower limbs under five gait patterns were dissected.The characteristic parameters of relative changes between the healthy and affected lower limbs that characterize motion intention were proposed,and the initial set of gait parameters containing the characteristic parameters was constructed.The discrimination between different gait patterns was improved.The online accurate recognition of the motion intention of the healthy lower limb within swing phase was achieved.The recognition time was reduced by 3.46 s compared to the conventional gait parameters.This study could provide data support for the highly coupled motion between the healthy and affected lower limbs.(2)A spatio-temporally embedded convolutional long-short memory deep learning network model based on a generative adversarial training mechanism was developed.The problems of sequence data sparsity and poor data inference were solved.The intention recognition accuracy was improved.The traditional machine learning algorithms have difficulty in extracting features from nonlinear gait sequence data,which easily leads to the problem of low accuracy of intention recognition.Hence,the generative adversarial training mechanism was applied to the deep learning network architecture.A fusion model based on spatio-temporally embedded convolutional long-short memory deep learning network was constructed.The temporal sequence features implied in the gait sequence data were mined.The problem of difficulty in constructing correlations between gait sequence data and motion intention was solved.Compared with the traditional SVM algorithm,the average recognition accuracy was improved by 17.8%.This study provides technical support for highly coupled motion between the healthy and affected lower limbs.(3)A two-level gait planning method for the healthy and affected side coupling was proposed.The problems of low refinement and poor tracking accuracy in gait trajectory learning were solved.The motion synergy between the healthy and the affected side of the lower limb exoskeleton was improved.To address the problem of poor recovery of bilateral lower limb motion coordination for existing gait planning methods,a two-level gait planning method based on reinforcement learning of multiple coupled continuous dynamic primitives was proposed.At the first level,a four-point segmented gait trajectory generation strategy was proposed to describe the trajectory of the healthy side lower limb movement in five gait patterns;at the second level,the reward and punishment mechanism of path integral reinforcement learning was applied to the dynamic motion primitive shape parameter update rule.The lower limb exoskeleton was able to learn and output coordinated motion trajectories that match the healthy side lower limb.(4)A gait strategy multi-objective optimization method was proposed.The problems of low accuracy and slow convergence of traditional gait optimization algorithms were solved.The robustness of lower limb exoskeleton system was improved.To address the conflicting problems of achieving simultaneous optimization of walking stability and system energy consumption in complex gait patterns,based on dynamics modeling analysis,the underlying gait parameters were mined.A gait strategy multi-objective optimization function with ZMP stability margin and single-step drive energy consumption was established.Based on the theory of cooperative regulation of multi-mechanism optimization strategy,a gait strategy multi-objective optimization method was proposed to improve walking stability and reduce the energy consumption of the exoskeleton system.This study provides technical support for realizing human-machine coupled regulation.(5)A dual closed-loop human-machine collaborative controller was designed.The problems of poor motion synergy and poor human-machine synergy that existed in traditional impedance control methods were solved.The human-machine coupling of the lower limb exoskeleton was improved.To address the problem of poor dynamic adaptive capability of existing lower limb exoskeleton control methods,a dual closed-loop human-machine collaborative controller was designed,including an inner-loop fuzzy position control module and an outer-loop impedance control module.The inner-loop fuzzy position control module realized the tracking of the trajectory of healthy lower limb and the healthy-affected side correction of the joint motion error.The motion coordination of the healthy and affected lower limbs was improved.The outer loop impedance control module based on radial basis variable impedance control was designed to realize the dynamic adjustment of the interactive force and relative position between human and exoskeleton.The human-machine coupling of lower limb exoskeleton system was improved.The human-machine wear experiments were conducted.The feasibility of the humanmachine collaborative control method based on the healthy-affected side gait coupling was systematically verified in terms of motion synergy,walking stability,and system energy consumption.The experimental results show that: under the five gait patterns,compared with the traditional impedance control method,in terms of motion synergy,the average fit of the generated trajectories for the hip and knee joints of the exoskeleton was improved by 14.77%and 16.08%,respectively;in terms of walking stability,the average deviation values of ZMP in the sagittal and coronal planes of the human-machine system were reduced by 94.28 mm and 119.58 mm,respectively;in terms of system energy consumption,the average input current value of the lower limb exoskeleton system was reduced by 15.66%. |
PDF Full Text Request