Research On SEMG-based Prediction Of Arm And Hand Stiffness And Humanoid Hand Variable Impedance Control

With the advent of eras of human-machine cooperation,it has become a research focus at home and abroad that robots can accomplish more practical tasks with the help of human beings.Supported by "973" project,the acquisition and application of mechanical and stiffness properties of human body is studied in this paper.According to the complex and flexible structure,human arm and hand system has become the subject of this paper due to its characteristics of high dexterity,strong robustness and good safety.Surface electromyography has become the main application tool due to its rich information.Firstly,the prediction method of the human arm and hand stiffness is studied.From the nerve point of view,the motion control principle of the joint is explored,and the rationality of s EMG-based stiffness prediction is shown.The muscle distribution that affected the dynamic characteristics of the arm and hand is analyzed,and the acquisition position of the s EMG signal is determined.The s EMG-based mathematical model for endpoint force prediction and stiffness prediction is established with different features of s EMG.Among them,different types of regression models(BLR and GEP)are used to establish the relationship between s EMG and stiffness.Secondly,an experimental system based on multi-modal sensing information is set up and experimental analysis is performed on the stiffness prediction based on s EMG.The experimental system includes hardware structure and software programs,which can simultaneously collect force,displacement and s EMG signals.A linear mapping relationship between different features of s EMG and force is established.According to root mean square error and correlation coefficient,the prediction accuracy of each feature is compared.After extracting better features,the linear and nonlinear mapping relationships between the stiffness of arm and hand endpoint and s EMG are established.The two models are analyzed and compared.Then their advantages and disadvantages are discussed.Then,two kinds of humanoid hand variable impedance control strategies applied to different scenarios are proposed.As the basic strategy,a location-based joint impedance control strategy is established and simulated.According to the human hand motion control mechanism,an adaptive humanoid hand variable impedance control strategy is proposed.The dependent factor is designed to modify the impedance update law according to the influence of joint independence.And simulation is conducted.It can be used in the flexible interaction between the bionic device and the environment.In order to apply to human-machine collaboration scenarios,a humanoid hand variable impedance control strategy based on s EMG stiffness prediction model is proposed.Finally,an experimental platform for multi-fingered dexterous hand is built and different impedance control strategies are studied experimentally.The dexterous hand finger moves in contract with different hardness obstacles and the characteristics of each motion parameter under different impedance control strategies are analyzed.The evaluation index is put forward by combining cumulative error and energy consumption.The comparison results verified the proposed control strategies.