Research On Multi-Motion State Perception Technology Of Exoskeleton

Exoskeleton robot system has a broad prospect as a booster device integrating sensing,control,mechanical and energy technologies.As a key link,the sensing system acquires the current motion mode and gait phase through the sensor,judges the motion intention of human,and provides input to the control system.According to the input analysis and judgment signal,the control system completes the corresponding motion of the robot by controlling the motor.In this process,the perception system monitors the motion state of the exoskeleton robot in real time.As the feedback of the exoskeleton control system,the exoskeleton perception system needs to feedback the human body's various motion postures and motion intentions in real time to improve the coupling between the exoskeleton and the human body.This thesis mainly studies the attitude perception of exoskeleton in walking,running,uphill and downhill,squat and other motion modes.The main research contents are as follows:The first is the demand analysis of multiple motion states of the exoskeleton.The application environment of the exoskeleton studied in this paper is the daily motion states such as walking and running on the flat ground,going up and down stairs,going up and down hills,and squatting.Among them,walking on the ground and moving downhill can be divided into different motion phases,such as heel landing,supporting phase,toe landing and swinging with equal and different motion phases.Then the sensor selection layout and conditioning circuit design.According to the analysis of the motion state,the required sensors were determined,the pressure sensors with appropriate specifications and the inertial measurement unit were selected,and the relevant conditioning circuit was designed to ensure the accuracy of the initial input signal of the exoskeleton system.Then gait phase and motion pattern recognition,according to the above incoming multi-channel sensor signal conditioning circuit,the design is suitable for the exoskeleton perception algorithm,algorithm mainly includes:(1)phase of gait recognition,fuzzy algorithm and KNN algorithm to realize the heel strike and support phase,tiptoe on the ground,swing phase four gait phase partitioning.(2)motion pattern recognition: according to the multi-channel pressure and inertial sensor,design a multistage classifier,the realization of motion pattern recognition.Finally,the time series model is used for signal prediction.Due to the hysteresis of the signal transmission feedback of the exoskeleton system,a certain algorithm is needed to predict the sensor signal to realize the perception of motion intention.In this paper,ARMA model is mainly adopted to obtain the time series model suitable for the exoskeleton signal through comparison.