Depth-Independent Augmented Dynamics Visual Servoing of Multirotors
Archit Krishna Kamath, Zewei Zheng, Mir Feroskhan
Abstract
This article introduces a depth-independent augmented dynamics visual servoing technique for multirotors. Not only does this method eliminate the necessity for inverse Jacobian computations, but it also effectively addresses a limitation of monocular camera-based visual servoing by eradicating its reliance on depth information. The proposed visual servoing strategy establishes a direct correlation between pixel variations and the multirotor's thrust and torque commands. This correlation empowers the multirotor's outer loop controller to proficiently manage image noise within the spectrum of system uncertainty, consequently bolstering the system's resilience to image noise and obviating the extensive need for image filtration. To regulate the augmented dynamics of the multirotor, a super-twisting fast-terminal sliding mode controller is formulated, offering both finite-time error convergence and minimal chattering in control efforts. Tis article showcases the efficacy of the proposed approach through numerical simulations and real-time experimental tests. In addition, the method's performance is evaluated against established position and image-based visual servoing techniques.