Aerial Camera Network for Observing Moving Targets

This paper considers a network of camera-equipped quad-rotors (aerial cameras) as aerial camera network (ACN) where each aerial camera is smart in the sense that it is mobile and can process, store, and wirelessly communicate information. This paper then presents an algorithm for decentralized and cooperative motion planning of these aerial cameras in 3-D environment to observe multiple targets moving on the ground. The proposed algorithm enables the ACN to observe targets from different elevations with the objective of jointly maximizing duration and quality (resolution) of observation for each target. This multi-scale observation depends on the field-of-view (FOV) of aerial camera and the quality of observation depends on the elevation and/or zoom level of aerial camera. The algorithm uses a quad-tree data structure to model not only the discrete movement decisions of aerial cameras in a 3-D environment but also the variations in the FOVs and qualities of observations. The use of artificial forces in the algorithm allows the aerial cameras to avoid collisions and to share the workload of assigning locations and targets to themselves. Simulation results present the effectiveness and performance evaluation of the proposed algorithm by comparison with the existing approaches of multi-target cooperative observation.