Visual-based Online Control of Gimbal on the UAV for Target Tracking

This paper investigates how to keep tracking of a ground target with a camera-equipped unmanned aerial vehicle (UAV) by autonomously steering the pan-tilt movements of the gimbal. In particular, we design and compare two different approaches, the position-based visual servoing (PBVS) and the image-based visual-servoing (IBVS). They mainly differ in whether the UAV attitudes are explicitly reflected in the control law. More specifically, PBVS utilizes the attitude measurements for target localization and compensates for the attitude changes. In contrast, IBVS only uses image tracking errors as feedbacks inputs. A detailed analysis is provided in the main text to reveal as well as interpret the underlying differences of their accuracy and robustness. To put a deeper sight into the tracking performance of the proposed controllers, a high-fidelity semiphysical simulation system has been constructed and a batch of numerical examples have been conducted. As expect, the results demonstrate that the two approaches outperform each other subjecting to different experimental settings. IBVS has a higher tracking accuracy for a moving target, while PBVS is more robust against wind disturbance. Therefore, our work provides potential guidance for practical applications.