Swing Phase Control of Semi-Active Prosthetic Knee Using Neural Network Predictive Control With Particle Swarm...

Area of Science:

• Biomedical Engineering
• Robotics
• Rehabilitation Technology

Background:

• Semi-active prosthetic knees with magnetorheological dampers offer advantages but lack active force generation for natural gait.
• This limitation prevents them from matching the performance of active prosthetic devices.

Purpose of the Study:

• To propose a new control algorithm for semi-active prosthetic knees to improve gait performance during the swing phase.
• To reduce the performance gap between semi-active and active prosthetic knees.

Main Methods:

• Development of a novel control algorithm utilizing neural network predictive control and particle swarm optimization.
• Simulation of the prosthetic knee using a double pendulum model to analyze gait trajectory.
• Assessment of the algorithm's real-time computational feasibility for embedded systems.

Main Results:

• The proposed controller generated a knee trajectory more similar to normal human gait compared to previous open-loop controllers.
• The controller demonstrated effectiveness across various ambulation speeds.
• The algorithm was found to be calculable in real time, suitable for embedded implementation.

Conclusions:

• The novel control algorithm significantly enhances the performance of semi-active prosthetic knees.
• This advancement brings semi-active devices closer to the natural gait performance of active prosthetics.
• The algorithm's real-time capability facilitates practical implementation in wearable prosthetic devices.