Model and experiments to optimize co-adaptation in a simplified myoelectric control system.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Myoelectric prostheses use surface electromyography (EMG) to control artificial limbs.
• Current myoelectric controls require significant user adaptation and differ from natural movement.
• Brain-machine interface co-adaptation techniques offer potential improvements.

Purpose of the Study:

• To investigate concurrent machine co-adaptation for myoelectric control.
• To explore how machine co-adaptation influences human adaptation in a simplified myoelectric context.
• To develop and test adaptive algorithms for enhanced prosthesis control.

Main Methods:

• A simplified myoelectric control system with perturbed muscle pulling vectors was used.
• Simulations modeled human adaptation to directional errors under co-adaptation.
• Experimental verification on human subjects tested variable gain co-adaptation strategies.

Main Results:

• Simulations showed low machine co-adaptation gain leads to slow adaptation, while high gain increases errors.
• A variable gain approach, implemented with directionally tuned neurons, improved adaptation rate and error management.
• Machine co-adaptation was shown to locally enhance myoelectric control and handle perturbations.

Conclusions:

• A simplified model facilitated the exploration of co-adaptation settings.
• Variable gain, locally encoded, is crucial for effective machine co-adaptation in myoelectric control.
• The study highlights considerations for applying these techniques to more complex prostheses.