Robust Online Predictive Visual Servoing for Autonomous Landing of a Rotor UAV

This paper presents a novel approach that integrates image-based visual servoing (IBVS) with online model predictive control (MPC) for landing a rotor Unmanned Aerial Vehicle (UAV) on maritime vessels, which ensures long-term visibility of the landing tag and enhances robustness against external disturbance. In contrast to existing studies that overlook the influence of UAV attitude changes on target visibility, this work distinguishes itself by comprehensively considering the attitude changes and the visibility constraints in the optimization framework, thereby effectively averting target loss. Furthermore, this paper contributes to the field by analyzing the input-to-state stability of the proposed approach under bounded disturbances, where the upper bound of disturbances is provided when the convergence domain is specified. To address the specific challenge of landing from high altitudes, a two-stage control strategy leveraging a self-designed nested landing tag is introduced. This strategy enables fast and accurate landings with smooth velocity transitions. Additionally, a soft constraint method is introduced to prioritize solution feasibility in online optimization. Extensive numerical simulations, software-in-the-loop (SITL) simulations, and field experiments validate the feasibility and superiority of our proposed method, underscoring its potential for practical implementation.