Neural network control of functional neuromuscular stimulation systems: computer simulation studies

Area of Science:

• Biomedical Engineering
• Control Systems
• Artificial Intelligence

Background:

• Functional Neuromuscular Stimulation (FNS) systems require sophisticated control for cyclic movements.
• Existing FNS control systems face challenges in individual customization, adaptation to physiological changes, and disturbance rejection.
• A robust and adaptive control system is crucial for effective FNS applications.

Purpose of the Study:

• To design and evaluate a neural network control system for FNS cyclic movements.
• To address customization, adaptation, and disturbance resistance in FNS control.
• To develop a novel learning algorithm for rapid parameter tuning.

Main Methods:

• Implementation of a two-stage neural network combining adaptive feedforward and feedback control.
• Development of a new learning algorithm for fast customization and adaptation.
• Evaluation using a computer-simulated musculoskeletal model with realistic muscle and skeletal dynamics.

Main Results:

• The neural network control system demonstrated automated customization of feedforward controller parameters.
• The system successfully adapted controller parameters online to account for musculoskeletal system changes.
• The control system exhibited resistance to mechanical disturbances in simulations.

Conclusions:

• The designed neural network control system effectively addresses key challenges in FNS control.
• The system's adaptive capabilities suggest suitability for controlling FNS and other dynamic systems.
• This approach offers a promising solution for improving FNS performance and usability.