Identification Of Exoskeleton Dynamics Model And Gait Analysis Based On Gaussian Process

As a special type of robot that needs close cooperation between human body and machinery,lower limb exoskeleton has a broad application prospect in industrial,medical,military and other fields.However,due to the diversity and complexity of its structure,the design of control strategy is extremely difficult.In order to accurately control the exoskeleton to complete various human-machine coupling cooperative tasks,the study of exoskeleton dynamics model and human gait is very important.The main research contents in this paper are as follows:1.Firstly,the mechanical structure and working principle of the lower limb exoskeleton prototype were theoretically analyzed,and the mechanical structure of the exoskeleton was simplified into a two-link structure,and the Lagrange method was adopted for its dynamic modeling.For unknown parameters and modeling errors in the model,Particle Swarm Optimization & Sparse Gaussian Processes Based on Input/Output Residuals,(PSO&RIO-SGP)exoskeleton dynamics model identification method,used to identify the unknown parameters of the theoretical model and to analyze the lumped error modeling.2.Aiming at the "human-host auxiliary" control mode in human-machine coupling tasks,taking into account the safety and coordination of human body itself and the stability of exoskeleton,the coupling relationship between human lower limb double joints and legs is deeply analyzed,and a probabilistic coupling model of lower limb joints based on deep Gaussian process regression method is proposed.3.Aiming at the prediction of exoskeleton joint trajectory under the active control mode,considering the problems of gait information acquisition equipment such as sampling delay and the large resource consumption and the increase in computing time caused by the inapplicability of the prediction model under different terrain and the need for real-time update,an online gait trajectory prediction method of exoskeleton based on the iterative Gaussian process regression model was proposed.