Visual feature prediction based interaction control of aerial manipulator under transient visual disappearance

Visual target disappearance within the camera's field of view (FOV), often caused by environmental variability, presents significant challenges for outdoor aerial manipulation tasks. Focusing on aerial contact operations, this article presents a visual feature prediction based interaction control strategy to enhance task continuity when visual feedback is lost. First, an adaptive visual servoing impedance control method is designed to improve the force tracking performance, in which the impedance stiffness is updated in real-time to meet the requirements of diverse contact tasks. Second, a confidence-weighted visual feature prediction method is presented to estimate visual states, enabling stable visual servoing position control even in the presence of visual feedback loss. Subsequently, a continuous characteristic model (CCM) based control method is developed to govern the attitude loop, which requires no accurate dynamics and enables a decoupled controller design by lumping coupling effects into characteristic parameters and bounded modeling errors. The Lyapunov criterion proves the effectiveness of the proposed strategy theoretically. Finally, comparative simulation and physical experimental results demonstrate the validity of the proposed scheme in performing both static and sliding contact operations under the temporary disappearance of the visual target.