SEMG-Based Simultaneous Estimation Of Motion And Force Of Human Upper Limb And Its Application In Teleoperation

Human have excellent motor capabilities and dexterous manipulation skills.In manipulation tasks such as screwing,human interact with the environment where motion is accompanied with force exertion.How to estimate a continuous profile of motion and force exerted by the upper limb during constrained motion is a challenging task.This paper focuses on simultaneous decoding of human motion and force intention from the surface electromyography(sEMG).The human manipulation skills are transferred to the robot through the proposed sEMG interface and is used to control a remote robot.The main researches of this thesis include:(1)Design and build a sEMG-based simultaneous estimation method of human motion and force intention in non-ideal conditions.As the arm position varies,the activation pattern of muscles of the upper limb will changes,and the accuracy of sEMG-based estimation model will decrease significantly.Considering the effect of arm position on sEMG-based estimation,the method based on hierarchical clustering decomposes sEMG signals into the motor units action potential(MUAP).The long short-term memory neural network is used to build the mapping model between MUAP features and joint angles,interaction forces,and include the data from different positions during training LSTM model.The root mean square error(RMSE)of joint angle estimation is(14.71±3.60)°,and the RMSE of interaction force estimation is(5.07±1.29)N.The method simultaneously decouples human motion and force intention from sEMG with high accuracy in different arm positions of daily scene.(2)Proposed a force control framework that can regulate the robot contact force in a desired force trajectory in human-robot interaction(HRI)scene with unknown environmental parameters.The proposed force control framework consists of two control loops.The inner loop is an admittance control and prescribes the transient response of the robot towards the environment by specifying the impedance parameters.The outer loop is a force tracking controller using a feedback linearization technique and control the robot to follow the desired force trajectory.Gazebo simulation verifies the feasibility of the control framework.The RMSE of the contact force tracking experiment on three different objects is(0.897±0.255)N,which verifies that the control framework has high control accuracy and good robustness in practical applications.(3)Design a teleoperation system which the operator can control the motion and force of robot through the proposed sEMG interface.The designed system decouples motion and force intention from the acquired sEMG in real time,and control the robot to track the estimated joint angle and interaction force trajectory.The sEMG interface transfers human manipulation skills to robot and enable operator to control the motion and force of robot through with the natural motion of arm.The operator controls the robot to complete the HRI task in only 10 s each time,and does not need to reprogram when the task changes,which verifies the feasibility and superiority of the EMG interface.