Sparse Identification Of Motor Learning Using Proxy Process Models.

Area of Science:

• Neuroscience
• Robotics
• Motor Control

Background:

• Effective neurorehabilitation requires computational models to track motor learning for personalized training.
• Error feedback is vital for skill acquisition and internal model formation in goal-directed movements.

Purpose of the Study:

• To investigate motor learning changes using intermittent "catch" (no-vision) trials as a proxy for internal model formation.
• To develop and validate computational models for assessing motor learning under sparse visual feedback conditions.

Main Methods:

• Trained 15 healthy subjects on visuomotor tasks with intermittent no-vision trials.
• Applied linear, affine, and second-order proxy process models to analyze learning data.
• Validated models using trial-to-trial and sample-to-sample updates, cross-validating across subjects and tasks.

Main Results:

• The second-order model with trial-to-trial updates demonstrated the best prediction of visuomotor learning.
• This model effectively captured the dynamics of internal model formation during intermittent feedback.

Conclusions:

• A second-order, trial-to-trial updating model is optimal for assessing motor learning in neurorehabilitation settings.
• This framework can enhance robotic and virtual reality systems by optimizing training parameters based on real-time learning assessment.