Building an internal model of a myoelectric prosthesis via closed-loop control for consistent and routine grasping

Area of Science:

• Neuroscience
• Rehabilitation Engineering
• Human-Computer Interaction

Background:

• Myoelectric prostheses lack somatosensory feedback, hindering intuitive control.
• The precise benefits and role of feedback in prosthesis use remain unclear.
• Understanding human control adaptation is crucial for advanced prosthetic design.

Purpose of the Study:

• Investigate the impact of somatosensory feedback on learning and maintaining feedforward control for grasping.
• Examine how feedback influences performance across different control interfaces (real vs. virtual, joystick vs. myocontrol).
• Determine the necessity of feedback for sustaining learned control models.

Main Methods:

• Nine able-bodied subjects performed repetitive grasping tasks with varying feedback conditions (feedback vs. no-feedback).
• Control configurations included virtual/real prosthetic hands and joystick/myocontrol interfaces.
• Outcome measures focused on relative force errors (median and dispersion).

Main Results:

• Somatosensory feedback significantly reduced grasping variability caused by system or interface uncertainties.
• Subjects successfully learned and utilized internal feedforward control models, maintaining performance even after feedback removal.
• Learned models exhibited instability over time, particularly with myocontrol interfaces.

Conclusions:

• Somatosensory feedback facilitates learning of myoelectric prosthesis control systems.
• Feedback is vital for maintaining learned control models, suggesting intermittent delivery is a viable strategy.
• The practical utility of learned control depends on the consistency of the prosthesis control interface.