Interpretable Fuzzy Inference for UAV Target Tracking Using Bounding-Box Geometry

A recent study available on arXiv (2608.04121) introduces a fuzzy-inference framework designed for UAV target tracking using vision. This innovative system derives continuous yaw commands from low-dimensional features obtained from YOLO bounding boxes, including target centroid location, area, and aspect ratio. It circumvents the need for explicit geometric modeling by employing a Mamdani fuzzy system with a shoulder-triangle-shoulder input partition as a clear baseline, subsequently utilizing a first-order Takagi-Sugeno fuzzy system. This method effectively tackles issues related to sensing uncertainty, computational limitations, and the necessity for interpretable control in cooperative aerial-ground robotics, providing a clear alternative to deep-learning and geometric-reconstruction techniques that depend on extensive datasets or external localization. The paper was published on August 26, 2025, and is classified as a cross-type announcement.