Flocking behavior for dynamic and complex swarm structures

TL;DR

This work addresses formation control for dynamic, complex UAV swarms by introducing FlockingBehavior, a virtual-centroid based algorithm that generates reference poses $P_i^W(t)$ from a single virtual-centroid trajectory $oldsymbol{ au}_{VC}^W(t)$ and a geometric formation $G_i^{VC}(t,i)$, enforcing Reynolds' cohesion, separation, and alignment. The method supports changing swarm size and formations in flight and relies on SE(3) transformations to couple 3D geometry with motion within a centralized offline control framework, while remaining compatible with other schemes. Validation includes extensive simulations across linear and curvilinear trajectories, dynamic reconfigurations, and scalability tests, as well as real-world experiments with Crazyflie drones, and the open-source ROS 2 modular code provides a practical platform for replication and extension.

Abstract

Maintaining the formation of complex structures with multiple UAVs and achieving complex trajectories remains a major challenge. This work presents an algorithm for implementing the flocking behavior of UAVs based on the concept of Virtual Centroid to easily develop a structure for the flock. The approach builds on the classical virtual-based behavior, providing a theoretical framework for incorporating enhancements to dynamically control both the number of agents and the formation of the structure. Simulation tests and real-world experiments were conducted, demonstrating its simplicity even with complex formations and complex trajectories.

Figures (12)

• Figure 1: Taxonomy of flocking behavior in multi-agent systems.
• Figure 2: Coordinated frames of the inertial frame W, the virtual centroid VC and each agent of the swarm and their relationships.
• Figure 3: Overview of the setup phase. On the left, it is pictured time $t_0$ prior to the formation to be shaped. On the right, UAVs occupy their position ready to start the collective movement at $t_s$. Red dot represent the virtual centroid, while green arrow points to its front orientation.
• Figure 4: Zenithal view with the linear trajectory according to the individual velocity of each drone. Color map on the right represents velocities between 0.0 and 0.5 $m/s$. The colored triangle represents the formation of the swarm at the beginning, middle, and end of the trajectory.
• Figure 5: Speed for the linear trajectory experiment at $0.5m/s$. Solid curves represent each drone speed $||\dot{P}_i^{W}(t)||$, while dotted curves stand for their relative speeds to the centroid $||\dot{P}_i^{VC}(t)||$. Yellow curve shows the speed of the centroid $||\dot{P}_{VC}^{W}(t)||$.