Brain-Inspired Multi-Pathway Motion Decision-Making for Obstacle Avoidance of Humanoid Arms

Area of Science:

• Robotics
• Artificial Intelligence
• Neuroscience

Background:

• Robotic obstacle avoidance in dynamic environments is challenging.
• Humanoid arms require enhanced adaptability and flexibility.
• Existing methods struggle with complex, real-time scenarios.

Purpose of the Study:

• To propose a brain-inspired multi-pathway motion decision-making method for humanoid arm obstacle avoidance.
• To improve the adaptability and flexibility of humanoid arms.
• To integrate rational planning and habitual actions for enhanced performance.

Main Methods:

• A novel framework with slow (rational planning) and fast (habitual action) pathways.
• The slow pathway uses an improved Real-Time Rapidly exploring Random Tree Star (RT-RRT*) algorithm.
• The fast pathway employs a Soft Actor-Critic trained model, guided by an emotional neural network for pathway modulation.

Main Results:

• The proposed multi-pathway framework significantly improves obstacle-avoidance success rates compared to existing methods.
• Motion characteristics generated by the framework align with human obstacle-avoidance behaviors.
• The emotional neural network effectively modulates pathway integration for smooth transitions.

Conclusions:

• The brain-inspired multi-pathway approach enhances humanoid arm obstacle avoidance.
• This method offers a more adaptable and flexible solution for complex robotic tasks.
• The integration of rational and habitual actions provides a promising direction for future robotic control systems.