A Learning-Based Hierarchical Control Scheme for an Exoskeleton Robot in Human-Robot Cooperative Manipulation

Area of Science:

• Robotics
• Human-Robot Interaction
• Control Systems

Background:

• Exoskeleton robots offer potential for assisting humans in daily activities.
• Effective human-robot cooperation requires sophisticated control strategies for seamless interaction.
• Existing methods may lack adaptability to dynamic environmental changes during cooperative tasks.

Purpose of the Study:

• To present a hierarchical control scheme for cooperative manipulation between humans and exoskeleton robots.
• To develop an adaptive neural network controller for back-drivable upper limb exoskeletons.
• To implement a skill learning strategy for adaptive human assistance during manipulation tasks.

Main Methods:

• A two-layer hierarchical control scheme was designed.
• Low-level control utilized an asymmetric barrier Lyapunov function-based adaptive neural network controller for admittance control.
• High-level interaction employed Gaussian mixture models for learning human skills from demonstration.

Main Results:

• The proposed admittance control scheme enabled the exoskeleton robot to be back-drivable.
• The skill learning strategy successfully extracted human skills for adaptive assistance.
• Experimental results demonstrated that the control scheme provided needed assistance during cooperative manipulation tasks.

Conclusions:

• The developed hierarchical control scheme effectively enables cooperative manipulation between humans and exoskeleton robots.
• The adaptive neural network controller ensures safe and compliant robot behavior.
• Skill learning from demonstration allows for personalized and context-aware robotic assistance.