Simulation Realization Of Stable Gait Parameters Learning For Paraplegic Lower Limb Walking Assistance Exoskeleton

In the traditional gait planning method of the lower limb exoskeleton,the following problems exist: different gait models need to be generated for different users;the gait model is only applicable to the common flat environment,and is not suitable for the uphill environment;it is impossible to learn and update the parameters of gait model according to the comfort and stability of the wearer during walking.In order to solve the problems of traditional gait planning methods,this paper proposes a unified gait model for different users,and extends the application range of gait planning methods from common flat environment to uphill environment.At the same time,reinforcement learning is used to update the gait model parameters according to the stability and comfort indicators of the wearer during walking.The main contents are as follows:1)Gait analysis of the exoskeleton of the lower limb: First,the gait of the humanexoskeleton system is analyzed.Then we analyzes the kinematics of the lower limb exoskeleton,the relationship between the gait trajectory curve and the angle of each joint is obtained.At the same time,the mapping of the trajectory curve from the end of the joint to the joint angle is realized,which is the basis of the gait planning.2)Design of gait learning methods of exoskeleton in flat environment: Firstly,the gait trajectory curve of the pre-trained human body is learned by the Dynamic Movement Primitives(DMP)gait modeling method.At the same time,the anthropomorphic gait trajectory curve is generated.Secondly,the gait planning method based on the centroidbased step size model is obtained according to the static stability criterion of the robot,and the optimal parameters of the system model are obtained by the reinforcement learning method.Finally,the obtained optimal model parameters are used to plan the gait of each step of the exoskeleton,so that the stability and comfort of the user during walking are improved.3)Design of exoskeleton gait learning method in uphill environment: Firstly,DMP is used to learn the anthropomorphic gait trajectory curve according to the human gait trajectory curve obtained by prior training.Secondly,the gait learning method in the flat environment is extended to the uphill environment.The gait model based on the centroid is proposed according to the particularity of the uphill environment,and the optimal parameters of the model are obtained by the method of reinforcement learning.Finally,the gait of each step of the exoskeleton is planned by obtaining the optimal model parameters,so that the stability and comfort of the user during the ascending process are improved.4)The verification of the gait learning method of the flat environment and the uphill environment: In order to verify the gait learning method proposed in this paper,the lower limb exoskeleton simulation platform was built.At the same time,several comparative experiments of gait learning planning methods and the predefined fixed gait planning method was designed and proposed in this paper.By comparing and analyzing the experimental results of the gait learning method and the predefined fixed gait planning method in the flat environment,the effectiveness of the gait learning method in the flat environment is obtained.By comparing the experimental results of the predefined fixed gait learning method and the gait learning method in the uphill environment,the effectiveness of the gait learning method in the uphill environment is obtained.The gait learning method proposed in this paper realizes the unified expression of the gait model for different users and improves the generalization of the gait model.The gait learning method is suitable for both flat and uphill environments,and can optimize the gait according to the stability and comfort index of the wearer during walking.It enlarges the application scope of the gait planning method and improves the stability and comfort of the wearer during walking.