| The lower extremity exoskeleton robot system belongs to the category of human-computer integration.It has important research significance in the common key technologies such as intention perception and multi-sensor information fusion.Its achievements have typical application value in military,medical,logistics and other industries.However,the current research on exoskeleton robot technology is still lacking.The problems of low flexibility,serious motion interference and poor stability are still elusive for exoskeleton robots to impact the mass market demand.In order to create a low-power,high-flexibility and high-load characteristics for the integration of the lower extremity exoskeleton system,this paper focuses on the "man-machine interaction system’s motion/force multi-dimensional coupling action law","human motion behavior feature perception,recognition and Basic scientific issues such as classification mechanism,human-machine cooperation efficiency and reliability evaluation mechanism,and research on the original design of integrated lower extremity exoskeleton system,system modeling,gait feature classification and identification,stability evaluation mechanism,etc..Specifically,it includes the following aspects:First,the lower extremity exoskeleton kinematics modeling based on human physiological structural parameters.In order to meet the design requirements of safety and universality of lower extremity exoskeleton design,the physiological structure and joint motion parameters of lower limbs of adults were analyzed.The virtual prototype of lower extremity exoskeleton was designed by using three-dimensional software,and the lower extremity exoskeleton was established by DH parameter method.The kinematics model is simulated and verified by simulation software to provide theoretical basis for subsequent research;Secondly,based on support vector machine(SVM)gait recognition algorithm optimization research.In order to improve the efficiency and accuracy of gait recognition,an SVM multi-classification optimization algorithm is proposed.Through the analysis of human gait travel,the whole gait cycle is divided into six phases,and decomposition studies are carried out.Summarize the changes of plantar pressure and lower limb joint angle in each phase state,and extract the human motion characteristics in the relevant stage.By analyzing the SVM binary classification principle and the characteristics of four multi-class SVM models,a multi-classification algorithm based on directed acyclic graph multi-classification algorithm and binary tree multi-classification algorithm is constructed to avoid the large error of traditional algorithm and the imbalance of segmentation area.problem;Then,based on the theory of zero moment point(ZMP),the concept of virtual support point is proposed,and the optimization support polygon and dynamic stability criterion are extended.Study the dynamic stability of the exoskeleton robot.The traditional ZMP stability criterion is analyzed and the limitations of the criterion are pointed out.Then the dynamic phase motion characteristics of the human body are analyzed,and the relationship between the swinging foot and the ZMP position during the single-foot support period is compared.Point concept to expand dynamic object support polygons.The criterion can realize the dynamic behavior stability discrimination,make the ZMP stability criterion have the function of determining dynamic stability,break the fixed limitation of the traditional criterion structure,and improve the accuracy of stability discrimination.Finally,construct the force feedback system to verify the feasibility and effectiveness of the gait classification recognition algorithm.The foot pressure and joint angle signals of the experimenter were collected as data samples.After preprocessing and feature extraction,the classification model was trained and classified experiments.The algorithm was validated by comparing the classification results. |
PDF Full Text Request