Development On Multi-terrain Adaptive Assistant Strategy Of Lower Limb Exoskeleton

Exoskeleton robots have been widely used in military,medical and lifestyle applications to enhance human locomotion and reduce exercise exertion.This thesis addresses the three core problems of lower limb-assisted exoskeletons used in urban multi-terrain environments: 1)traditional terrain recognition algorithms are susceptible to external interference and have low portability; 2)gait phase recognition in multi-terrain environments has low accuracy; 3)the assertion strategy does not take into account terrain adaptability,and proposes a lower limb-assisted exoskeleton assertion strategy for urban multi-terrain environments.The aim is to design a lower limb-assisted exoskeleton system that can adaptively adjust the assist output according to different terrain and slope.The main work of this paper includes:A terrain slope estimation algorithm based on inertial measurement units is proposed to address the problem that traditional terrain recognition algorithms are susceptible to external interference and have low portability.Using the basic human body parameters and the lower limb joint motion posture measured by the inertial measurement unit integrated into the lower limb-assisted exoskeleton as the input and the terrain slope as the output,data from 10 subjects were collected as the training set and a neural network regressionbased terrain slope estimation model was established.The model was tested with an exoskeleton wearing experiment,and the results showed that the average slope estimation error of the model was 3.73°.To address the problem of low accuracy of traditional gait phase recognition algorithm in multiple terrains,a multi-terrain gait phase recognition algorithm based on adaptive oscillator is proposed.The adaptive oscillator is used to process and learn the human knee curve in multiple terrains,and the adaptive oscillator is optimised to solve the ”antiphase” problem.Simulations using an open-source database show that the adaptive oscillator can accurately and efficiently output continuous gait phase over multiple terrains,and that it can still follow the learning steadily when the speed changes.In the simulations,the average reset delay was 2.853% and 4.752% for slope and stair environments respectively.In the exoskeleton wear experiment,the average reset delay was 4.15%,outperforming the conventional multi-terrain gait phase recognition algorithm.To address the problem of adaptive boosting in multi-terrain environments of exoskeleton robots,a multi-terrain boosting algorithm based on parametric dynamic motion primitives is proposed.According to the difference of human knee joint motion moment curves in different terrains,three parametric dynamic motion primitives are used to construct boosting sub-models for up-slope,down-slope and stair terrains by combining the self-built stair moment database and the open source slope moment database.Polynomial fitting is used to obtain the mapping relationship between sub-model style parameters and terrain slope,so as to build a multi-terrain booster algorithm model based on parametric dynamic motion primitives.The generalisation of the moment model was verified using knee moment curves for ±10° slopes and ±30° stairs,and the experimental results showed that the peak error in moment prediction was less than 5%.A lower limb assisted exoskeleton system was built to test the algorithm in indoor and outdoor multi-terrain experiments.Indoor experiments show that the system can effectively reduce the activation of the human rectus femoris muscle,while outdoor experiments show that the system can reduce human exercise exertion by approximately 5%.