The aim of this paper is to present a novel approach to swarm control of small fixed-wing UAVs, which combines only two flocking behaviours with a leadership feature. In the presented approach, two fundamental rules of Reynolds flocking are applied, i.e., cohesion and repulsion, as the base of a decentralized control of self-organization of the flock. These rules are combined with a leadership feature, which is responsible for a global behaviour of guidance, as in the case of animals. Such a bio-inspired combination allows the achievement of a coherent collective flight of a flock of fixed-wing UAVs without applying formal behaviours of migration and alignment. This highly simplifies an implementation of the algorithm. The presented results include both numerical simulations and experimental flights, which validate the hardware implementation of the approach.

Implementations of multi-UAV systems can be divided mainly into two different approaches, centralised system that synchronises positions of each vehicle by a ground station and an autonomous system based on decentralised control, which offers more flexibility and independence. Decentralisation of multi-UAV control entails the need for information sharing between all vehicles, what in some cases could be problematic due to a significant amount of data to be sent over the wireless network. To improve the reliability and the throughput of information sharing inside the formation of UAVs, this paper proposes an approach that combines virtual structure with a leader and two flocking behaviours. Each UAV has assigned different virtual migration point referenced to the leader's position which is simultaneously the origin of a formation reference frame. All migration points create together a virtual rigid structure. Each vehicle uses local behaviours of cohesion and repulsion respectively, to track its own assigned point in the structure and to avoid a collision with the previous UAV in the structure. To calculate parameters of local behaviours, each UAV should know position and attitude of the leader to define the formation reference frame and also the actual position of the previous UAV in the structure. Hence, information sharing can be based on a chain of local peer-to-peer communication between two consecutive vehicles in the structure. In such solution, the information about the leader could be sequentially transmitted from one UAV to another. Numerical simulations were prepared and carried out to verify the effectiveness of the presented approach. Trajectories recorded during those simulations show collective, coherence and collision-free flights of the formation created with five UAVs.