Unmanned Aerial Vehicles Cooperative Tracking of Moving Ground Target in Urban Environments

The problem of autonomous tracking of a ground moving target in an urban terrain is studied. In the investigated scenario, the target is tracked from multiple unmanned aerial vehicles using gimballed or body-fixed sensors under the constraints of terrain occlusions and airspace limitations. Information regarding the occlusions may be available a priori from a database or may be provided to the system by the operator based on his understanding of the environment. To ensure flyable trajectories, the unmanned aerial vehicles' dynamic constrains must be taken into account. A methodology is proposed for solving in real time a general class of such problems by casting the tracking task as a cooperative motion planning problem. Because of the computational complexity of the problem, a stochastic search method (genetic algorithm) is proposed for finding in real time monotonically improving solutions. An important attribute of the proposed solution approach is its scalability and, consequently, applicability to large-sized problems. For testing the algorithm, it was implemented in a high-fidelity simulation test bed using a visual database of an actual city. The viability of using the algorithm is shown using a Monte Carlo study. It is envisioned that automating this part of a ground moving target tracking problem will considerably reduce operators' workload and dramatically improve mission performance in such real-life problems.