Gait Planning And Control Method Of Lower Extremity Exoskeletal Robot

Lower extremity exoskeleton and the human body form a complex human-machine coupling system,the technology involves gait perception and estimation,motion mode determination,as well as control.The gait perception,acquisition and analysis can provide theoretical support for the mechanical design;human motion intention estimation and motion pattern discrimination can effectively excavate human locomotion information.Robotic control strategy is the key to realize the function of lower extremity exoskeleton,which include robot kinematics and dynamics,human-machine interaction,impedance control and so on.In view of the above problems,this disstrtation will expend in three aspects,which include gait acquisition and analysis,locomotion prediction and gait phase separation,lower limb exoskeletal robot kinematics,dynamics and control.Gait is the behavioral characteristics of human walking.Gait acquisition and analysis are prerequisite of exoskeletal robot walking.Human motion capture methods include visual method,footprints,optical signals,ultrasonic signals,pressure signals and so on.At present,the most popular mothod is the optical method which using optical camera and the force-measuring board.However,the optical method has some disadvanges,which include expensive device,limited using place and complicated data processing.In view of those shortcomings,a mechanical light-weight human gait collecting rack is designed,which using angle sensors and the film pressure sensors to obtain human gait motion information and plantar pressure information.On this basis,Lab VIEW is used to design the PC interface of the human gait acquisition system,and the database is constructed by MySQL to store the gait data.Further,the functional standard gait is obtained by analyzing the gait data and successfully applied in the fixed gait walking.Estimation of human intent refers to the prediction of the body's next movement,which can be realize by joint angle prediction.According to the periodicity and rhythmicity of human motion,the Takens nonlinear time series prediction algorithm is applied to human gait prediction.In view of the problem that the algorithm predicts incorrectly with a few history data or changing locomotion pattern,this dissertation improved the Takens' prediction algorithm and get a better result.Because of the poor behaver of short-term prediction of Takens' algorithm,Newton's prediction method is improved,which has a better performance in short-term prediction.Gait phase recognition is important in human locomation.The gait phase usually can be descriminated by methods such as fuzzy logic and the hidden Markov model.The fuzzy logic strategy is greatly affected by the parameters threshold,while the hidden Markov model requires the tagged data to be trained,however the tagged data usually is difficult to get.Therefore,a strategy of unsupervised classification,mixed Gaussian model,is proposed.Facing with the difficulty of physical meaning confusion and parameter sensitivity,The article improved the algorithm and peform a better phase separation accuracy and robustness.In aspect of control strategy,the kinematics model and dynamic model of exoskeleton robot are analyzed,the parameters of dynamic model are identified,the control of lower exoskeleton robot is completed by master-slave control strategy on the end,which uses different control methods according to gait phase.Experiments show that the control strategy can realize human-machine systerm walking on the ground below 2 km/h,going up and down stairs,squatting,sitting down and crossing the obstacle.