Dual-EKF System Identification and Model Predictive Path Integral Control of a Retinal Microsurgical Robot

Area of Science:

• Robotics
• Biomedical Engineering
• Control Systems

Background:

• Retinal microsurgery demands high precision and dexterity.
• Existing robotic systems face challenges in controlling complex, coupled movements.
• The Improved Integrated Robotic Intraocular Snake (I²RIS) is a novel snake-like robot designed for intraocular procedures.

Purpose of the Study:

• To develop and validate a computationally efficient modeling and control framework for the I²RIS robot.
• To address the coupled pitch and yaw dynamics inherent in snake-like robotic systems.
• To compare the performance of an optimal stochastic controller against a traditional controller for real-time trajectory tracking.

Main Methods:

• A multi-input multi-output (MIMO) mass-spring-damping (MSD) model was formulated to capture coupled dynamics.
• A dual-extended Kalman filter (dual-EKF) was employed for joint parameter and state estimation.
• The Model Predictive Path Integral (MPPI) controller was designed and compared with a Linear Quadratic Regulator (LQR).

Main Results:

• The coupled MSD model accurately represents the interdependence between pitch and yaw dynamics.
• The MPPI controller demonstrated superior trajectory-tracking performance compared to the LQR.
• The proposed framework enables computationally efficient, real-time control with minimal training data.

Conclusions:

• The developed modeling and control framework offers a robust and efficient solution for snake-like robotic systems.
• MPPI provides enhanced performance and robustness for controlling the coupled dynamics of the I²RIS robot.
• This approach is promising for advancing precision in retinal microsurgery and other delicate robotic applications.