Research On Assistance Optimization Of Lumbar Exoskeleton Under Lifting Task

Thanks to its potential in enhancing exercise ability and reducing human consumption,exoskeleton has drawn increasing research interest and gradually applied in various fields,such as medical rehabilitation training.Workers who have been engaged in heavy load lifting for a long time are prone to lumbar muscle strain and occupational diseases,which have great impacts on personal physical and mental wellbeing and family happiness.How to effectively assist human in exercise,such that protecting human from injury,is of great significance in the field of lumbar assisted exoskeleton.Developing effective assist strategy is one key to realize human-exoskeleton cooperation.However,due to the lack of accurate modeling of cooperative objects and tasks,it still remains challenging to obtain promising cooperation performance between exoskeleton and human.In order to deal with this challenge,this thesis first establishes the model regarding to human lifting motion.Subsequently,followed after the mechanical structure design of a lumbar assist exoskeleton,an optimization framework based on Bayesian algorithm is proposed to optimize the model parameters of the exoskeleton system.The main works of this thesis can be summarized as follows:(1)Establishment of the torque model with respect to human joint under lifting motion and design of the exoskeleton control system.Firstly,human joint torque model is established based on the force analysis of human hip joint under lifting motion.Then,the motion characteristics of human body under lifting task are analyzed and the exoskeleton driving mode is determined to obtain better human-exoskeleton interaction control effect.Next,followed after the mechanical structure design of a lumbar assist exoskeleton,this thesis completes the system design for the developed device based on torque control.Finally,the efficacy of the developed controller is evaluated via our established hardware platform over different subjects.(2)Characterization of muscle fatigue of main muscle groups through the combination of Electromyography and muscle synergy.Based on characteristics of s EMG signal,a s EMG signal acquisition system is built.The time and frequency domain indexes of s EMG signals are used to characterize single muscle fatigue.Muscle synergy is used to analyze the cooperative work of muscles in the process of lifting,which is then combined with single muscle fatigue to characterize human muscle fatigue of main muscle groups.(3)Design of a Bayesian-based "human-in-the-loop"(HIL)framework for the model parameter optimization of the exoskeleton.In the developed HIL framework,the human cost evaluation indicator is first established according to the muscle fatigue of the wearer.Afterwards,the Bayesian optimization algorithm is used to optimize the air pressure parameters of pneumatic artificial muscle.Moreover,the experimental platform of wearing exoskeleton to assist human body in heavy object lifting task is built.Lastly,different comparative experimental tests are conducted over different subjects to verify the effectiveness of the proposed method.The experimental results show that the waist assisted exoskeleton has few effects on muscle synergy,while can reduce the activation degree of human waist and back muscles as well as the fatigue of human waist and back muscles in the case where the device is used to assist human lifting motion.Moreover,the developed Bayesian-based HIL framework can identify the pneumatic artificial muscle pressure values for different subjects,such that the activation degrees of upper limb muscles and fatigue speed of each subject can be reduced.Thus,the proposed method in this thesis could be regarded as an important alternative in the field of human-exoskeleton cooperation control,which may also suggest valuable clues for the optimization control of different wearable devices.