Human Motion Prediction And Human-robot Coordination Control For Lower Extremity Exoskeleton

A wearable lower extremity exoskeleton is designed to assist the human body's waking,enhance the human user's walking ability,reduce the energy consumption and extend the ability of carrying payloads.The wearbale exoskeleton is a typical highly human-robot coupling system,which needs human-robot coordination movement to improve the wearing comfort and enhance the ergonomics.The research of the lower extremity exoskeleton has several challenges,i.e.,human gait phase estimation,human gait mode identification,human gait trajectory prediction,human-exoskeleton interaction acquriation and human-exoskeleton coordination control.In this paper,we focus on the following aspects,i.e.,human gait phase and gait mode identification,human gait trajectory prediction,human exoskeleton coordination control strategy design and analysis,and ergonomics performance evaluation.Based on anatomy and kinematics analysis of human body,the features of human motion are obtained.Based on the Lagrange method,the exoskeleton's dynamics model is established.Based on the human-robot coupling characteristic,the human-robot interaction model is obtained.Based on Matlab and Adams,the simulation of the sliding mode control with CMAC neural network compensation is realized.In simulations,the proposed control algorithm is verified.Human gait phases are identified based on multi-modal sensor information.With the fuzzy logic system with particle swarm optimization(PSO),four sub-phases are identified online,and the average identification accuracy is over 95%.Based on the ground reaction force and the posture of the foot and the shank,several typical gait patterns(level-ground walking,stair ascent,stair descent,and ramp ascent and ramp descent)are identified by using PSO-based support vector machine.Experimental results show that the accuracy of gait pattern identification can reach 98.35%±1.65%.Experimental results show that the proposed method can achieve smooth and fast trasition for gait mode identification.Based on the physical human-robot interaction(pHRI)information,the human joint trajectory is estimated and predicted.Since the pHRI measurement lags behind the real human movement intention,Kalman filter is used to deal with pHRI,eliminating the noise and compensating the delay.Since the relationship between the pHRI and the human gait is highly nonlinear,it is difficult to obtain the relationship by using mathematical model.Based on the pHRI measurement,online sparse Gaussian process regression algorithm is proposed to estimate the human movement.Specific experiments are performed and experimental results show that the proposed method is effective and valid.To achieve human-robot coordination movement,the model-free adaptive pHRI minimization control is studied,where the parameters of the controller are adjusted by the fixed penalty function.The gravity compensation control is able of reducing the effects on the human user body.Based on the identified walking phase,the output of the controller is adjusted by different ways,where the level-ground walking experiment is carried out.In order to realize the human robot integration control better,an model-based adaptive impedance control strategy is proposed,where the parameters of impedance model are adjusted online according to the target state information and gait tracking error.Since the proposed method is model-based approach,the dynamic model can be obtained by RBF neural network in real time.To avoid the un-modeled dynamics and the uncertainties,which can be regarded as disturbances in the system,the extended state observer is designed to estimate the disturbances.The estimated disturbances can be eliminated by the designed control law to improve the control performance.Experimental results show that the proposed control strategy can achive the human exoskeleton coordination movement.The performance evaluation index is the main aspect of the exoskeleton research.The comprehensive evaluation based on the subjective indexes and the objective indexes can be achived,which mainly compares the subjective indexes and the objective indexes before and after wearing.Experiments are carried out on various terrians,including walking on the stairs,walking on the grassy land and walking on the complex terrain.In each kind of experiment,each human subject gives their scores on four items,i.e.,wearing comfort,assistance,wearing convenience and movement smoothness,by filling in the questionnaire.Objective performance indexes include ground reaction force,pressure on th shoulder and heart rate,which are measured by force sensors and Polar heart rate measurement device respectively.The objective performance indexes measurement is performed on the treadmill.The speed of treadmill is 5km/h and each test time is 5 minutes.Under the conditions with payloads and without payloads,the mean plantar pressure difference is within 5N and the increment of the shoulder pressure increased is 10.85 N.The average heart rate of the load decreased by 9.6% compared to the loads on human body directly.Compared to the exoskeleton wearing without payloads,the heart rate increased by only 3.45%.Experimental results show that the load can be transfered through the mechanism to the ground by wearing the exoskeleton so that the wearer is not subject to the heavy oppression.The exoskeleton can help human users to bear heavey payloads and reduce the burden on the human body.