Automated Curriculum Design for High-Dimensional Human Motor Learning.

Area of Science:

• Robotics
• Motor Learning
• Control Systems

Background:

• Designing effective practice schedules for complex motor learning is challenging, particularly when internal skill states are hidden.
• Task performance may not accurately represent true learning progress in high-dimensional tasks.

Purpose of the Study:

• To develop an automated framework for designing personalized practice curricula in novel motor learning tasks.
• To improve the efficiency of skill acquisition using a model-based approach.

Main Methods:

• Combined a human motor learning model with real-time skill estimation.
• Utilized Stochastic Nonlinear Model Predictive Control for adaptive curriculum generation.
• Validated the framework through simulations and human-subject studies (N=36) with a hand exoskeleton.

Main Results:

• The proposed framework accelerated skill acquisition by approximately 23% compared to a random curriculum.
• Achieved a ~17% improvement in learning efficiency over a performance heuristics-based curriculum.
• Demonstrated significant gains in learning efficiency for novel motor learning paradigms.

Conclusions:

• Model-based, individualized curricula can significantly enhance motor learning efficiency.
• The developed framework shows potential for motor rehabilitation and complex skill training.
• Automated curriculum design offers a promising direction for optimizing motor skill acquisition.