Study On The Design Of Human-machine Interaction Force Sensors And Assosciated Control Method For Lower Extremity Assist Exoskeleton

Exoskeleton robot is a kind of intelligent wearable device that can detect wearer's movement and help the wearer to coordinate its force output.It can be widely used in industrial,military,civil and other cases.The lower limb assisting exoskeleton robot,which was studied in this research subject,is mainly used in helping wearer to carry heavy load.Based on the existing platform,this research project has completed the design of the exoskeleton's man-machine interaction system and the research of the cooperative control method,which improves exoskeleton's abilities on signal perceiving and decision making.Whether exoskeleton can acquire comprehensive and rich interaction force information or not is matter for detecting human moving intention.This subject has completed the design and performance test of the man-machine interaction force sensors which are located on the exoskeleton's foot,shank,thigh and the back-binding part.The test results show that the sensors designed in this subject has good precision and usability,which can provide rich interactive information for exoskeleton robot system.The building of dynamic model of exoskeleton is basic for exoskeleton coordination control.After analyzing the shortcomings of the traditional modeling method of the exoskeleton dynamics which is based on the vertical and inverted pendulum,a new unified exoskeleton dynamic modeling method based on the calculation of the foot pressure is designed.In this method,the dynamic modeling of support phase and swing phase of the exoskeleton are no longer divided,so that the discontinuity of the joint output force caused by the switching of the support and the swing phase is avoided,and the smooth transition of the foot pressure can also be realized through this method.In this paper,a parameter identification method based on the gradient descent method is designed for recognizing the parameters of the dynamic model of the exoskeleton.The accurate identification of the parameters of the external skeletal dynamics is realized and the precision of the control is improved.The understanding of the movement intention of the human is the basis for improving the intelligence of exoskeleton.In this paper,an exoskeleton dynamic parameter identification method based on LSTM network is designed.LSTM is a classic RNN network model,which can predict the next system state according to the input sequence of the long or short term.Based on the LSTM model,this paper realized the prediction of exoskeleton's next movement state by analyzing the periodic motion state in the past period.This method greatly improves the speed and intelligence of the systemThe decision-making ability after obtaining rich information is the embodiment of the high-level intelligence of exoskeleton robots.Based on the rich exoskeleton man-machine interaction information,this paper designs a closed loop control algorithm aiming to the minimizing of multiple interaction force,which realizes the decrease of the interaction force on the operator's back,shank and thigh by changing the output torque of exoskeleton,which can also enhance the comfort of wearing exoskeleton robot.Based on a large number of experiments,the assisting effect,speed of movement and the adaptability of the environment of the exoskeleton were tested.The results show that the exoskeleton robot has a fine assisting ability,following ability and environmental adaptability.