In context
In mid-2025, low-altitude UAV operations were expanding rapidly across logistics, inspection, and surveillance, yet complex obstacle environments continued to cause target loss due to occlusion. Single-UAV systems struggled with limited fields of view and slow angle adjustments, highlighting the need for robust perception and multi-agent coordination.
What was reported
Researchers at Beihang University proposed a unified framework for occluded target detection and multi-UAV cooperative tracking. For perception, they developed a lightweight Transformer-based tracker that removes the decoder and uses a pure encoder with a RepVGG backbone, reducing computational load for onboard real-time operation.
In fixed-view simulations, the method stably detected targets with occlusion rates exceeding 90%. Compared to DiMP18 and E.T.Tracker, it achieved roughly twice the frame rate with comparable accuracy. Outdoor flights over dense jungle validated real-time onboard detection and stable tracking.
For tracking control, a multi-UAV strategy dynamically plans trajectories to cover the target's visible area from multiple angles. Using ray intersection and B-spline prediction, three UAVs coordinated in simulations and flights, improving tracking stability over a single UAV in multi-obstacle environments.
Why it mattered
This work demonstrated a practical path to reliable autonomous tracking in cluttered low-altitude airspace, combining efficient onboard perception with scalable multi-UAV coordination—key for applications like search-and-rescue, security patrol, and traffic monitoring.
“The proposed framework solves the problems of detection failure and tracking loss caused by target occlusion, demonstrating significant superiority in both real-time performance and tracking stability.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-06-09 · “无人机遮挡目标检测与协同跟踪方法”