Adaptive Visual Servoing Control of Tilt-Rotor Quadrotors Based on Thrust-Loss Modeling

This paper introduces a novel robust visual servoing approach for tilt-rotor quadrotor systems designed to create airflow-free zones beneath the vehicle for precision agricultural spraying. We propose modeling the tilt-rotor dynamics as a thrust loss problem, where fixed rotor tilt angles directly reduce effective vertical thrust while complex aerodynamic couplings are treated as composite disturbances. Our integrated control framework combines a virtual camera-based robust observer with sliding mode control, adaptive thrust compensation, and image moment feedforward compensation to ensure stable visual servoing across various tilt angles. Theoretical stability analysis proves global asymptotic convergence of the closed-loop system using Lyapunov methods. Numerical simulations demonstrate effective control at tilt angles up to 40° with an optimal adaptation gain of γ b =0.2, achieving 96.8% RMSE improvement from 1.8746 m to 0.0602 m under 20° tilt conditions. ROS Gazebo SITL simulations validate the thrust-loss modeling approach by demonstrating comparable performance between physical tilt-rotor unmanned aerial vehicles (UAVs) and conventional UAVs experiencing equivalent thrust loss, confirming the feasibility of the simplified modeling hypothesis. This work establishes a foundation for developing practical tilt-rotor UAVs that can minimize pesticide drift through deliberate airflow manipulation, addressing a critical limitation in current agricultural drone technology.