Proportional and simultaneous myoelectric control of a robotic arm

Abstract

In the past years, machine and robot technologies have progressed rapidly, improving human life and making tasks much easier to execute. In a world that caters to the able-bodied, someone who has lost his/her upper limb often becomes dependent on other people in order to perform the simplest of tasks. Human-machine interfaces (HMI) are used to help these people perform speci c limb movements. HMIs provide control of a device using only signals produced by the body which are called biosignals. This project's aim is to develop an HMI for the control of a robotic arm manipulator through the use of non-invasive surface electromyography (EMG) signals. These signals can provide information on the limb movement being performed, such as its position in 3D space. Current research is focused on developing a reliable prosthetic device which incorporates simultaneous control of di erent joints. In this project, a model which converts EMG signals from multiple muscles to elbow and shoulder angles for simultaneous and proportional control is developed and tested in real-time. Furthermore an application to this project is demonstrated where a robotic arm replicates the user's wrist position in real-time as the user moves in the horizontal plane, stops and points towards an object. This system's performance was assessed using the correlation coe cient (CC) measure which was found to be 0.908 0.043 and 0.951 0.015 for the x and y directions respectively. This project focuses on developing a system which caters for di erent speeds and exible variety of movements. In this project seven di erent sequential and simultaneous movements in di erent planes were successfully modelled. The average root mean square error (RMSE) for sequential movements varied from 5.43° to 12.34° while the best performing simultaneous joint model resulted with an average cross-validated RMSE of 9.03° 0.33° for the elbow angle estimation and 7.30° 0.85° for the shoulder angle estimation. Incorporating these models into a single model would be the nal step in creating a fully functional prosthetic device which could replace the functions of the elbow and shoulder.