Abstract:Adaptation to the operator's intents and the external physical uncertainties is critical to the performances of the human-robot collaboration. In this study, we proposed a control interface that combined the electrical impedance tomography (EIT) based sensing approach with the robotic controllers to produce proper assistance with external uncertainties in the collaborative task. The interface first estimates the continuous forearm muscle contractions (represented by the grasp forces) by the optimized EIT features with an easily worn fabric band. The recognition decisions then serve as the control inputs to adjust the state transitions and the desired interaction forces in real time. We evaluated the interface in the tasks of grasp force estimation and human-robot sawing by recruiting more than 20 subjects in total. For grasp force estimation, the interface produced an average of R-2 >= 0.9 in both offline and online validations with the feature optimization procedure and a sigmoid regression function. The interface was robust to external disturbances without retraining or manual calibrations. The average R-2 = 0.86 with the untrained dynamic postures, and the average R-2 values ranged from 0.85 to 0.88 in the tests with redonning the front-end in interday and intraday uses. For human-robot sawing, the interface accomplished the tasks with a high success rate in controlling the states (>96%) and intuitive adjusting of the sawing forces, being combined with the designed hybrid admittance/position controller. It was also adaptive to the sawing frequency changes and the sawing directions according to the operator's intents. The interface's performances are comparable, if not better, to the state of the art on both biological signal based grasp force estimation and human-robot sawing. Future efforts are worth being paid in this new direction to get more promising outcomes.

Continuous Multi-DoF Wrist Kinematics Estimation Based on a Human-Machine Interface with Electrical-Impedance-Tomography

Wrist Angle Estimation with a Musculoskeletal Model Driven by Electrical Impedance Tomography Signals

THE INFLUENCE OF TORQUE AND ELBOW POSITION ON WRIST ROTATE ANGLE RECOGNITION BASED SURFACE ELECTROMYOGRAM

Real-Time Wrist Motion Decoding with High Framerate Electrical Impedance Tomography (EIT)

Toward a Human-Machine Interface Based on Electrical Impedance Tomography for Robotic Manipulator Control.

Simultaneous and Continuous Estimation of Shoulder and Elbow Kinematics from Surface EMG Signals

Simultaneous estimation of 2-DOF wrist movements based on constrained non-negative matrix factorization and Hadamard product

Mapping Individual Motor Unit Activity to Continuous Three-DoF Wrist Torques: Perspectives for Myoelectric Control

Non-Invasive Analysis of Motor Unit Activation During Simultaneous and Continuous Wrist Movements

Continuous Kalman Estimation Method for Finger Kinematics Tracking from Surface Electromyography

A Multi-DoF Prosthetic Hand Finger Joint Controller for Wearable sEMG Sensors by Nonlinear Autoregressive Exogenous Model

An EMG-Based Deep Learning Approach for Multi-DOF Wrist Movement Decoding

Training Wrist Extensor Function and Detecting Unwanted Movement Strategies in an EMG-controlled Visuomotor Task.

Accurate Continuous Prediction of 14 Degrees of Freedom of the Hand from Myoelectrical Signals through Convolutive Deep Learning

Simultaneous and Proportional Estimation of Finger Joint Angles from Surface EMG Signals during Mirrored Bilateral Movements

An Electrical Impedance Tomography Based Interface for Human–Robot Collaboration

Long Short Term Memory Model based Continuous Estimation of Human Finger Joint Angles

Estimating continuous data of wrist joint angles using ultrasound images

Continuous Prediction of Wrist Joint Kinematics Using Surface Electromyography from the Perspective of Muscle Anatomy and Muscle Synergy Feature Extraction

Motion estimation of elbow joint from sEMG using continuous wavelet transform and back propagation neural networks

Simultaneous Estimation of Digit Tip Forces and Hand Postures in a Simulated Real-Life Condition With High-Density Electromyography and Deep Learning