Hierarchical Image-Based Visual-Servoing Control for Stewart Manipulators via Data-Driven Adaptation and Incremental Prediction

This work presents a hierarchical image-based visual servoing (IBVS) framework for a Stewart manipulator, integrating target recognition, tracking, and adaptive control. To mitigate calibration uncertainties and external disturbances inherent in IBVS systems, a data-driven adaptive incremental predictive control scheme is developed by constructing a linear image-plane model from visual feedback. A structured incremental prediction model and a data-driven adaptive law are introduced to avoid updating the full image Jacobian, a process subject to pose-observation difficulty, delay-induced instability, and parameter divergence in eye-in-hand (EIH) Stewart manipulators. The proposed controller employs an online saturated recursive least-squares algorithm with a forgetting factor to optimize a multi-objective cost function that incorporates image-pixel errors, base disturbances, and observed manipulator poses. In addition, a Newton-Raphson iteration with adaptive termination thresholds is used to ensure efficient and accurate kinematic computation. Comparative experiments against feature-space control and proportional-derivative predictive control validate that the proposed method significantly reduces pixel errors and enhances tracking robustness, demonstrating the effectiveness for dynamic target tracking of Stewart manipulators.