Autonomous UAV Surveillance

The problem of surveillance using unmanned aerial vehicles (UAVs) has been largely researched by many different groups and organization. It features a real life problem of monitoring ground targets using UAVs, and presents several different challenges, such as route optimization, flight planning and clustering.

In this work we aimed to suggest a possible solution to the Multi-Target-Multi-UAVs problem, meaning surveillance of multiple targets using multiple UAVs, by utilizing a previous solution to the Multi-Target-Single-UAV problem. Providing a reliable, efficient solution will greatly lower the costs of reconnaissance missions, as well as improve their effectiveness, via the use of an optimal number of UAVs, as well as optimal flight routes. These two might also lead to reduced pollution, resulting from efficient energy consumption.

We have based our work primarily on the work of Jevtic et al, Unmanned Aerial Vehicle Route Optimization Using Ant System Algorithm. This work suggested a swarm intelligence based solution to the Multi-Target-Single-UAV problem. Their work showed the effectiveness of the Ant Colony Optimization algorithm for the problem.

In Our work we reproduced their results and suggested an alteration to the algorithm in order to achieve a solution to the problem of clustering targets. We then combined the two, together with a loss function we defined, in order to achieve a solution for the Multi-Target-Multi-UAV problem by presenting it as multiple Multi-Target-Single-UAV problems.

Our results show great promise for the swarm intelligence approach, revealing the ant colony approach to be highly powerful and adaptable at solving different types of problems.

A full review of the work [in Hebrew] can be found here