Design And Control Of Fiber-reinforced Humanoid Soft Robotic Manipulator

As an important branch of robotics,soft robots play an important role and advantage over rigid robots.Soft robots can withstand greater impact and deformation which are more suitable in unstructured environments.In addition,soft robots have better robustness and safety.This paper designs and produces a fiber reinforced humanoid robot with the soft material.According to statistics,there are nearly 400,000 people with upper limbs disabilities in China.What's more,the current commercial prostheses are rigid hands and the functions are limited.Therefore,in order to improve the living standards of people with disabilities in the upper limbs,the soft manipulator is used as a flexible prosthesis.This paper proposes to combine the soft robotic hand and the surface electromyographic signal(sEMG)signal,and use sEMG signal to control the movement of the soft robotic hand.The sEMG is a kind of physiological signal of the human body and can more directly and effectively reflect the purpose of human activities.First,a large amount of literature data were collected and read,the research and development of soft manipulators and myoelectric prostheses were summarized and analyzed,and the significance of this study and the overall technical route which is the main content were proposed.Second,based on the structure of the human hand,a fiber-reinforced with three-chamber soft finger structure was proposed and designed.Three theoretical models of soft manipulators were established:bending-angle and pressure model,fiber-reinforced torsion and pressure model,soft finger force output and pressure model.Abaqus simulation software was used to simulate the flexural motion and bending-twisting combined motion of the soft finger.According to the simulation results,the theoretical model is optimized.Based on the mechanical design of the mold,Ecoflex 00-50 and Dragon skin 30 were mixed and poured with the ratio of 1:1.Thirdly,under the conditions of different grip strengths,three-channel sEMG signals are collected near the muscles of the human forearm,such as the shallow flexor muscles.The sEMG signals are preprocessed,detected in the active segment,and extracted the 48 eigenvalues among the time domain,frequency domain,and time-frequency domains.Then select the appropriate dimensions and reduce dimensionality reduction.Finally,based on the obtained 10dimensional eigenvalues,the SVM with linear kernel function was selected as the classifier.The classification model was obtained and the classification accuracy rate is 94.2%.However,since the force output is a continuous output,the obtained classification model is expanded,and a force output control strategy for the softhand is established based on the extended model.Fourth,design soft hardware and software control platform.The hardware platform mainly includes pressure control manufacture and experimental platform manufacture.The experiments include the bending angle acquisition experiment,the twist angle acquisition experiment,the force output measurement experiment,the manipulator envelopment grab experiment,etc.The experimental results were compared with the theoretical model,and a modified theoretical model was obtained.In addition,using the sEMG model and the soft robotic hand,four different softness items were selected to grab,and finally get a good results.Software control design includes interface design and pneumatic control program design.The interface can realize the real-time monitoring of the EMG signal and air pressure inside the soft finger,the predicted value of the EMG grip model,and other configuration parameters.