This study addresses the real-time visual tracking task in edge environments by proposing a robust visual servoing control system based on a higher-order sliding mode observer, enabling a quadrotor UAV to autonomously track a moving soccer ball during outdoor sports broadcasts while relying solely on a monocular camera and an inertial measurement unit, thereby eliminating any dependency on external positioning or velocity sensors such as GPS. The system adopts a hierarchical control architecture in which the observer plays a central role: operating on resource-constrained edge devices, it leverages only visual information to estimate unknown external disturbances and target motion states in real time, significantly reducing both hardware requirements and computational overhead to meet the lightweight deployment demands of edge nodes while simultaneously providing the continuous disturbance estimation necessary to sustain a low-latency control loop. Experimental validation was conducted under simulated real soccer match scenarios featuring wind disturbances and abrupt trajectory changes, and the results demonstrate that the system restores the ball to a stable position near the center of the camera’s field of view within seconds of a perturbation, thereby satisfying the stringent stability and real-time performance requirements of live broadcasting. This work confirms the feasibility of delivering low-latency, highly reliable visual intelligence services on resource-constrained edge platforms through advanced observer design, offering a technical reference for the convergence of real-time perception and control within the edge Internet of Things and providing valuable insights for future research on large-scale edge collaboration and secure, sustainable service architectures.
Long et al. (Thu,) studied this question.