Empirical evaluation of multi UAV coverage path planning for aerial surveying

Unmanned Aerial Vehicles (UAVs) are being increasingly used for various tasks, including aerial surveys whereby a drone flies over a region of interest and takes photos. Often, these photos are then used to create one (3D) map, which can be used for later inspection. However, in some cases, this region of interest can be rather large, and performing the survey with a single UAV may take a long time. Besides the inconveniences, this can also worsen the quality of the (3D) map due to changing weather conditions over the day. Consequently, it is advantageous to parallelize this task by deploying multiple UAVs simultaneously. Nevertheless, when using multiple UAVs, the path-planning stage becomes more complicated. Despite multiple works concerning this multi-UAV coverage path planning (CPP) algorithms can be found in the literature, they fail to focus on the practical implications and outcomes of deploying multiple UAVs for survey missions. Hence, in this work, we address this research gap by performing real world tests, comparing the results between a single UAV coverage mission and a multi-UAV coverage mission. In order to do so, we first had to implement a multi-UAV CPP algorithm. To remain focused on the differences between a single and multi-UAV coverage mission, we implemented a simple CPP algorithm (i.e. the back-and-forth method). Then we performed multiple simulations to (i) verify our implementation, and (ii) to measure the difference in flight times (between multi and single UAV coverage missions). Once our implementation was verified in simulation, we performed real world tests. During the tests, the UAVs took images, which were used to create a (3D) map of the region. We then compared the differences between a map generated by only one UAV, and by multiple UAVs. From our experiments, we can conclude that we can reduce the total time to cover an area significantly without lowering the quality of the (3D) map..