Visual Servoing Scheme and Nonlinear Model Predictive Control for Micro Quadrotor under Unknown Feature Depth

The visual servo for micro quadrotor is a complex nonlinear system with strong coupling, time-varying parameters, and underactuation. Traditional controllers are not suitable for such systems, especially in the presence of time-varying, underactuated, and state-coupled nonlinearities. The inadequacy of these controllers can result in the quadrotor moving too fast during tracking, causing motion blur to the camera, leading to tracking failure. To address these issues, we propose a robust nonlinear model predictive control (NMPC) scheme for micro quadrotor system that accounts for uncertain feature depth. First, we establish the system model using the virtual camera approach. We then design a feature depth estimator and high-gain observer to estimate the feature depth and linear velocity of the quadrotor. To ensure the system meets FOV, actuator saturation, and linear velocity constraints in the presence of external disturbances, we design tight state constraints based on the Lipschitz condition. Finally, we design an experimental platform to carry out the micro quadrotor tracking ground target experiment. The results shows that the proposed NMPC-based visual servoing scheme can achieve the stable tracking of micro quadrotor to ground target.