Visual Predictive Control for Robotics with RBF-EKF Coupled State-Disturbance Estimation and Task-Oriented K-Means...

Area of Science:

• Robotics
• Control Systems
• Computer Vision

Background:

• Image-Based Visual Servoing (IBVS) systems face instability challenges from noise, modeling errors, and disturbances.
• Existing methods often struggle to simultaneously address these complex issues effectively.

Purpose of the Study:

• To develop a robust Visual Predictive Control (VPC) framework for enhancing IBVS stability and tracking accuracy.
• To integrate advanced estimation techniques for improved system performance under uncertainty.

Main Methods:

• A feedback linearization Model Predictive Control (MPC) law was designed to manage nonlinearities and constraints.
• A coupled state-disturbance estimation mechanism using Extended Kalman Filter (EKF) for noise suppression and Radial Basis Function (RBF) network for disturbance learning was implemented.
• Task-oriented K-means clustering was employed to optimize RBF center selection for improved efficiency.

Main Results:

• Lyapunov analysis confirmed the Uniformly Ultimately Bounded (UUB) stability of the proposed system.
• Simulations showed significant reductions in estimation errors and enhanced tracking accuracy compared to traditional methods.
• The integrated approach demonstrated superior robustness and practicality for robotic visual servoing.

Conclusions:

• The proposed Visual Predictive Control framework effectively enhances the robustness and performance of IBVS systems.
• The deep coordination of control and estimation strategies provides a practical solution for real-world robotic applications.
• This research contributes to advancing the field of robotic control through improved stability and accuracy.