Adaptive neural network control of cyclic movements using functional neuromuscular stimulation

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Functional neuromuscular stimulation (FNS) systems require customized patterns for effective muscle activation.
• Previous controllers like pattern generator/pattern shaper (PG/PS) showed promise in isometric conditions.
• Adapting stimulation patterns to dynamic isotonic conditions presents unique challenges.

Purpose of the Study:

• To evaluate an adaptive feedforward controller (PG/PS) for automatically developing FNS stimulation patterns.
• To compare the PG/PS controller's performance against a proportional-derivative (PD) feedback controller under isotonic conditions.
• To assess the controller's ability to account for complex musculoskeletal dynamics.

Main Methods:

• Computer simulations were used to optimize controller parameters for various musculoskeletal models.
• The PG/PS and PD controllers were experimentally tested on individuals with spinal cord injury.
• Performance was evaluated based on tracking accuracy during cyclic movements.

Main Results:

• The PG/PS controller demonstrated superior tracking performance compared to the PD controller.
• The adaptive controller successfully adjusted stimulation patterns for isotonic contractions, considering muscle and limb dynamics.
• Simulations provided effective parameters for both controllers across diverse systems.

Conclusions:

• The PG/PS adaptive control system offers an effective method for automatic FNS stimulation pattern customization.
• This approach may enhance the efficacy and usability of FNS systems for individuals with neuromuscular impairments.
• The controller's ability to adapt to dynamic conditions is crucial for FNS applications.