A Robust Adaptive Kalman Filter Based Visual Servoing Control for an Inertial Stabilization Platform

This article presents a visual servoing scheme designed for inertial stabilization platforms, addressing the impact of visual measurement delays, sensor noise, and model uncertainties on high-precision tracking. The proposed framework integrates a robust adaptive Kalman filter (RAKF) and an improved integral model predictive controller (IMPC). The filter fuses measurements from visual and inertial sensors to estimate the long-term kinematic behavior of the target, mitigating the effects of sensor noise and time delays. Furthermore, the filtered predictions are directly incorporated into the iterative optimization of the model predictive controller, ensuring tight integration between perception and control. The robust IMPC combines a predictive control framework with smooth integral state resetting and a finite-time discrete-time disturbance observer, effectively eliminating steady-state errors while enhancing dynamic tracking performance. A comprehensive analysis of the controller’s feasibility and stability is provided. Experimental results demonstrate a 69.8 % improvement in tracking accuracy compared to conventional model predictive control methods.