Research On Human-in-the-loop Assistance Optimization Of Lower Limb Exoskeletons During Human Walking

Lower limb exoskeletons can be used to improve walking efficiency of abled people and to assist the elderly and patients with motor dysfunction in daily activities or rehabilitation training.These devices have shown a great potential in improving people’s health and alleviating the crisis of population aging and have been attracting more and more attentions in the academic and industrial world.Assisting human walking has been one of the most important applications in lower limb exoskeletons.How to improve the walking efficiency for the wearers with exoskeleton assistance has always been a great challenge in this field no matter for frontiers of science and technology and people’s life and health.In recent years,human-in-the-loop optimization(HILO)has been proved to be an effective approach to address the aforementioned challenges which used human physiological signals as feedback information to continuously adjust exoskeleton assistance patterns.Although human-in-the-loop optimization has shown the ability to effectively improve the exoskeletons performance,there is still has some shortcomings,which leads to inefficient optimization process and keeps it from being widely applied in practical daily life.To address the issue of the disadvantages of the existing human-in-the-loop assistance optimization,this dissertation studied the human-in-the-loop optimization of ankle exoskeleton based on human physiology and biomechanics.The main contributions of this dissertation are summarized as follows:1)Design and implementation of a cable-driven ankle exoskeleton system for human walking assistance and assistive torque trajectory planning and controlling.First,a light-weight cable-driven ankle exoskeleton system which was used as the main experimental set-up for the following research contents was designed and developed.Then,the exoskeleton assistive torque trajectory was designed according to biological ankle moment.To achieve accurate torque control,a algorithm combining proportional control,damping injection and adaptive iterative learning was proposed and compared with existing prominent algorithms.Results showed that the proposed control method can achieve higher accurate torque tracking with a short settling time.2)The wearers’ physiological and biomechanical responses to different exoskeleton assistance patterns were studied.The initial parameter value of human-in-the-loop optimization was derived based on the wearers’ responses.We designed a series of assistance patters within the parameter space of human-in-the-loop optimization as typical assistance patterns which combined multiple peak torque magnitudes and peak torque time values.Then we carried out experiments to evaluate the wearers’ metabolic energy costs,lower limb muscle activation levels,lower limb joints’ kinematics and kinetics,and lower limb muscle coordination patterns under different assistance patterns.This study provided preliminary understanding of human-exoskeleton interaction and guidance to the choice of initial parameter values in the optimization.Results showed that under one certain assistance pattern,the wearers’ metabolic cost,lower limb muscle activation level and biological ankle moment was decreased the most among the investigated patterns which suggested that this pattern can serve as a fairly decent starting point for the optimization procedure.3)We proposed two kinds of physiology information use as human-in-the-loop optimization cost functions and evaluated the feasibility of them in this dissertation.Aiming at improving the efficiency of exoskeleton assistance pattern optimization in a certain gait,we analyzed the correlation between metabolic cost and physiology and biomechanics signals during human walking.The we proposed an optimization objective function based on metabolic cost estimation.We estimated human metabolic cost during human walking from heart rate and lower limb muscle activation levels and used estimated cost as the optimization objective function.Next we proposed an optimization objective function based on a weighted sum of lower limb muscle activation levels.The weight vector was created such that the sum was most correlated to the metabolic cost.We carried-out experiments to valid this two methods and results showed that both can effectively decreased metabolic cost during walking.4)For exoskeleton assistance pattern optimization in multi-gait condition,an objective function based on lower limb muscle activation levels was proposed to replace metabolic cost to improve efficiency of the optimization.This dissertation focused on different walking speeds and slopes and carried-out experiments to evaluate lower limb muscle activation levels under multiple assistance patterns in multi-gait walking condition.Then we constructed a cost function based on optimizing the difference between objective functions under different assistance patterns by using artificial bee colony algorithm.The performance of the proposed method was validated by experiments.Results showed that the proposed objective function can effectively guide the humanin-the-loop optimization in multi-gait walking condition.The wearer’s metabolic costs and lower limb muscle activation levels were both decreased under slow walking,normal walking,fast walking and incline walking conditions.To sum up,this dissertation aiming to address the issue of the disadvantages of the existing human-in-the-loop assistance optimization.Firstly,we designed and implemented a light-weight cable-driven ankle exoskeleton system for the following research contents.Then we carried out our research from three aspects: the wearers’ physiological and biomechanical responses to different exoskeleton assistance patterns,human-in-the-loop optimization based on human physiological information in a certain gait,and human-in-the-loop optimization based on lower limb muscle activities in multi-gaits.Experiments validation show that our research can provided comprehensive understanding of human-exoskeleton interaction,and can improve the efficiency of exoskeleton assistance pattern optimization in both one gait and multi-gaits conditions.