Visibility-Constrained Vision-Based Air-to-Air Tracking and Control for Noncooperative UAVs With Unknown Maneuvers

This article proposes a visibility-constrained vision-based air-to-air tracking system for the pursuit of noncooperative unmanned aerial vehicles (UAVs) with complex and unknown maneuvers, which comprises a target motion estimator and a position-based visual servo (PBVS) controller. First, the target motion estimator integrates pseudolinear Kalman filter (PLKF) with interactive multimodel (IMM) filters, which can effectively cope with the high uncertainty of target motion in air-to-air scenarios, overcoming the limitations of traditional single-model estimation in complex maneuvering trajectories. Moreover, the proposed estimator does not require the observer to perform higher order motion than the target to recover the target state. Second, the designed PBVS controller is based on a virtual coordinate frame to decouple the translation and rotation control, thereby simplifying the controller design. In order to ensure the target visibility throughout the entire tracking process, the control barrier function (CBF) is incorporated to dynamically adjust the controller input by imposing constraints on the actual image plane, which enhances the reliability and safety of the system. Finally, simulation and real-world experiment results have verified the effectiveness of the proposed method.