Evolutionary Mapping with Multiple Unmanned Aerial Vehicles

Unmanned aerial vehicles (UAVs) have been very successful in many civilian and commercial applications, including disaster relief, search and rescue, precision farming, archaeology, cargo transport, and surveillance. In most of these applications, mapping is a required phase that needs to be performed as an initial step. While mapping has attracted much attention in the last decades, much of the works rely on single drones. In this context, we propose a multiple UAV system for efficient mapping, minimizing mission time and cost. The system includes offline and online planning, and a good balance between both to reduce on-board processing. Offline planning includes area decomposition, take-off location finding, and path planning. Online planning will then be used to react to any unforeseen event that might occur during the plan execution. These incidents include a sudden change in weather conditions, communication loss or drone malfunction, and the presence of a nearby flying obstacle. Each of the main offline and online planning tasks are formalized as a Multi-Objective optimization (MOO) problem where requirements need to be met while objectives have to be optimized. In this regard, we consider several evolutionary techniques to tackle these MOO problems. To assess the performance of these techniques, we conducted several experiments and reported the related results. One finding is that MOEA/D outperforms NSGA2, while the latter requires less processing time.