Study On Variable Weight Control Algorithm For Lower Limb Powered Exoskeleton

The lower limb assisted exoskeleton is a wearable mechanical device that can support human movement and assist the wearer to carry heavy objects.At the same time,the lower limb assisted exoskeleton can help reduce the fatigue when the body is carrying heavy things and improve the ability of people to carry loads.It has very broad application prospects in many fields such as military and civil.However,the current exoskeleton system has the following problems: auxiliary torque fixation cannot change with the change of load,and it is difficult to meet the requirements of different load bearing scenarios.When the load size changes,it is often necessary to manually adjust the torque value,which is lack of adaptive ability and affects the system's power assist efficiency,so the practicability is not high.Therefore,this paper proposes a variable load control strategy applied to the assisted exoskeleton to solve this problem.In different load scenarios,the torque prediction model is used to provide appropriate auxiliary torque,which can improve the system's adaptability.The work of this paper is summarized as follows:First,the experiment of load-bearing walking of human body was conducted to divide the gait cycle of human body and analyze the changes of the Angle of each joint under different loads.Then,the OpenSim simulation software was used to establish the loadbearing model of human body to obtain the torque data of lower limb joints and provide reference tracks for the torque prediction model.Secondly,a torque prediction model based on dynamic motion primitive(DMP)is proposed.The torque prediction model based on DMP learns the acquired torque curve and USES the nonlinear forced term function to reproduce the reference trajectory,and then establishes the regression mapping relationship between the load parameters of human body and the nonlinear function,so as to be able to predict the torque curve of joints under different load without manual adjustment.Then a model of torque prediction based on style parameterization(SDMP)is proposed to solve the problem of large prediction error of DMP torque prediction model.Unlike the DMP moment prediction model,which can only learn one moment curve,the SDMP moment prediction model can learn several moment curves at a time.The SDMP torque prediction model USES singular value decomposition and local weighted linear regression method to extract the characteristic parameters of nonlinear functions of several torque curves,and then calculates the weight of gaussian kernel function,and then establishes the regression mapping model between the load-bearing parameters of human body and the characteristic parameters.The method can better predict the torque curve of joints under different loads,the prediction error is smaller and the algorithm is more practical.Finally,the effectiveness of the control algorithm is verified by system experiments.A lower extremity assisted exoskeleton system hualex-3.2 was built,which was composed of bionic mechanical structure,hydraulic power unit,sensing hardware unit and algorithmic control unit,and the effectiveness of the control strategy proposed in this paper in solving the problem of system adaptability and improving the system assisted efficiency was tested by the exoskeleton system wearing experiment.