Abstract
Using swarms consisting of UAV for surveillance, mapping, or search-and-rescue missions has been of great interest in recent years. To be able to perform the same tasks in an indoor environment their trajectories have to be planned precisely, i.e., their dynamics have to be taken into account. They do not only need to keep safety distances to fixed obstacles such as walls or furniture but also to moving obstacles, e.g., moving machine parts or a person. Additionally, the UAVs in the swarm should perform the mission cooperatively without staying closely together covering only a small part of the area. Therefore, path or trajectory planning is of great interest. Path planning for multiple agents of a swarm is still a very challenging task. Especially when a priori unknown obstacles, moving obstacles, and realistic dynamics are taken into account, the problem becomes NP-hard. In this paper, we introduce advancements to a promising path planning algorithm based on model predictive control (MPC). The algorithm is extended by a method to assign waypoints only to certain agents, a closest waypoint search, and an energy consumption model leading to more realistic trajectories. Since it allows to efficiently use the limited energy, longer missions can be carried out. Additionally, the model enables to initiate the return of agents running low on energy on time and safely return them to the starting location. The proposed strategies are tested in an indoor scenario showing that different rooms can be assigned to individual agents of the swarm and a safe return can be combined with still performing some of the mission objectives.

DOI: https://doi.org/10.1109/ICNS60906.2024.10550797