Variable Admittance Control Based on Fuzzy Reinforcement Learning for Minimally Invasive Surgery Manipulator

Area of Science:

• Robotics
• Artificial Intelligence
• Medical Engineering

Background:

• Minimally invasive surgery (MIS) requires precise control of robotic manipulators.
• Enhancing natural and intuitive physical interaction is crucial for surgeon comfort and task efficiency in MIS.

Purpose of the Study:

• To propose a novel hybrid variable admittance model for pose adjustment in MIS manipulators.
• To improve the naturalness and intuitiveness of human-robot interaction during surgical tasks.

Main Methods:

• Developed a hybrid variable admittance model integrating Fuzzy Sarsa(λ)-learning.
• Implemented continuous variable virtual damping adjusted dynamically based on operator state and robot dynamics.
• Utilized fuzzy partitioning of the state space to capture operator characteristics in physical human-robot interaction.

Main Results:

• The proposed model dynamically modifies virtual damping to enhance operator comfort.
• Fuzzy Sarsa(λ)-learning enables the model to adapt to human intentions and changing robot dynamics.
• Comparative experiments demonstrated the model's effectiveness in joint space control for MIS manipulators.

Conclusions:

• The hybrid variable admittance model significantly improves intuitive physical interaction in MIS robotic surgery.
• Dynamic adjustment of virtual damping enhances surgeon comfort and control precision.
• Fuzzy Sarsa(λ)-learning provides an effective strategy for adaptive control in human-robot interaction for surgical applications.