A Nonlinear Adaptive Model-Predictive Approach for Visual Servoing of Unmanned Aerial Vehicles

Abstract

Aerial vehicles, due to the nature of their operations, are subject to many uncertainties. In particular, their positioning control problem is aggravated in GPS-denied environments. As a result, vision-based control (or visual servoing) solutions are sought to address their navigation control issue. In this work, visual servoing is formulated as a nonlinear optimization problem in the image plane. The proposed approach is based on a class of nonlinear model-predictive control that takes constraints into consideration. A 2-DOF model helicopter is considered for initial formulations and verification of the approach. The considered model helicopter has a coupled nonlinear pitch/yaw dynamics. It is also capable of a pitch over (i.e., hover at a non-zero pitch angle) around its pivotal axis of rotation due to the offset between the center of mass (CoM) of its fuselage and the rotation axis. This further makes the linearized model (calculated around any pitch angle). A linear parameter varying (LPV) system for an adaptive control is formulated under a model-predictive visual servoing framework. In this paper, a design guideline is provided under a model-predictive visual servoing paradigm that considers field of view (FOV) constraints, the constraint on the pitch angle, and also the maximum voltages that can be applied to independently controlled pitch/yaw motors. The control space is parameterized via Laguerre basis network functions that make the optimization scheme computationally less expensive, thus suitable for real-time applications.