Neural Network Based Lower Limb Prosthetic Control Using Super Twisting Sliding Mode Control
No Thumbnail Available
Date
2025-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
A prosthetic is an artificial limb employed to substitute lost anatomical structures due to
disease, injury, or trauma. In this paper, an approach for controlling the prosthetic leg
using a surface Electromyography (sEMG) signal, an artificial neural network (ANN),
and super twisting sliding mode control (ST-SMC) is presented. The triggering signal is
extracted from surface electromyographic (sEMG) signals recorded from the nine muscles
at the lower limb, which are grouped into three movement groups. Intense signal analysis
that includes filtering, rectification, and linearization is done to obtain the reference sEMG
to the NN from the raw sEMG because the raw sEMG data can’t be used as training data
for NN. An artificial neural network (ANN) predicts the joint angle for walking, upstairs
and downstairs using the processed sEMG signals of the muscles. The super-twisting
sliding mode control (ST-SMC) is used to regulate the motion of the prosthetic joints in
accordance with the specified reference trajectories. The kinematic model is formulated
using forward and inverse kinematics principles, which determine the position and orientation
of the prosthetic leg. The dynamic model is based on a fixed coordinate system
of human lower leg modeling, formulated by the Euler-Lagrange principle. Rather than
using a simple triple pendulum model, the proposed approach presents a more realistic
model of the human lower leg. Matlab software’s signal analyzer, neural network fitting
packages, and MATLAB/Simulink are used for signal analysis, neural network training,
and model simulation. The dynamic system modeling and ST-SMC controller design are
implemented. Parameter change analysis and disturbance analysis were done to show that
the controller is robust against internal parameter changes and external environmental
changes. From the simulation results, it is observed that training the neural network with
processed data and implementing the ST-SMC results in increased regression value and
decreased mean squared error (MSE) in trajectory tracking instead of using SMC.
Description
Keywords
Prosthesis, surface Electromyography (sEMG), Artificial neural network (ANN), Super twisting Sliding Mode Control (ST-SMC).