Motor synergy generalization framework for new targets in multi-planar and multi-directional reaching task

Area of Science:

• Neuroscience
• Robotics
• Motor Control

Background:

• Human movements exhibit adaptability despite numerous degrees of freedom.
• Motor synergies, low-dimensional control signals, explain human movement patterns.
• Existing research suggests motor synergy repertoires are task-specific.

Purpose of the Study:

• To investigate if combinations of motor synergy repertoires can be systematically reused for new targets.
• To demonstrate motor synergy generalization beyond the capabilities of individual repertoires.
• To explore synergistic control in a realistic 7-degree-of-freedom arm model.

Main Methods:

• Utilized a multi-directional reaching task.
• Trained multiple policies with reinforcement learning for limited target ranges.
• Extracted motor synergies and optimized their activation patterns.
• Employed a full 7-degree-of-freedom arm model.

Main Results:

• Demonstrated that combined motor synergy repertoires can reach new targets.
• Showed that these combined repertoires outperform original policies and single repertoires.
• Achieved generalization to new targets in new planes.

Conclusions:

• The combination of motor synergies offers a powerful mechanism for generalizing motor control to novel tasks and targets.
• This approach enables adaptation in complex, high-dimensional systems like the human arm.
• Paves the way for more adaptable and versatile robotic control systems.