Drone Coverage Control for Moving Target Tracking
Adapting a drone's altitude and its object-detection model in real time, so detection accuracy and field of view stay matched to how the target is moving.
In collaboration with Fujitsu
Tracking a moving target from the air forces a trade-off. Fly low and the object detector is accurate but the field of view is narrow; fly high and the drone sees everything but detects little. This project, carried out in collaboration with Fujitsu, enhances drone coverage control for dynamic target tracking by adapting both the drone’s altitude and the detection model online.
The controller treats altitude as a decision variable alongside horizontal position, and switches detection models to match the resulting ground sample distance. The method was implemented in ROS with TensorFlow-based detectors and presented at the SICE Annual Conference 2024 in Kochi, Japan.