Development of a Real-Time Neural Controller Using an Emgdriven Musculoskeletal Model.

Area of Science:

• Biomedical Engineering
• Robotics
• Neuroscience

Background:

• Electromyography (EMG) control systems face challenges in accuracy, latency, and robustness.
• Existing systems struggle with both isometric and non-isometric muscle contractions.
• Human motor control offers insights for adaptive and smooth robotic interactions.

Purpose of the Study:

• To develop a novel real-time neural controller for volitional robot and computer control.
• To improve EMG control by addressing accuracy, latency, and robustness.
• To enable seamless motion control during various muscle contraction types.

Main Methods:

• Developed an EMG-driven musculoskeletal model for neural command translation.
• Combined EMG signal processing, neural activation dynamics, and Hill-type muscle modeling.
• Integrated muscle activation dynamics with impedance control for adaptive interactions.

Main Results:

• Achieved high reference tracking performance in controlling a robotic actuator for lower-limb movements.
• Demonstrated state-of-the-art processing time of 2.9 ms for real-time embedded computing.
• Showcased robust control during static and dynamic movements at varying speeds.

Conclusions:

• The novel controller offers fast, accurate, and adaptable performance for EMG-based human-machine interfaces.
• This approach enhances robustness against electrode variability and signal noise.
• Lays the groundwork for next-generation neural-machine interfaces in diverse applications.