New algorithm enables drone swarms to add members while reshaping mid-flight

Most multi-agent formation control systems assume a fixed set of agents, forcing teams to stop or restart when new members need to be added. This becomes a critical bottleneck in real-world scenarios like search-and-rescue or surveillance, where drones or robots must join an existing formation while it is already maneuvering through tight spaces. The new distributed control framework from researchers Tao He and Gangshan Jing solves this by enabling “non-uniform scaling” — allowing the formation to stretch or shrink by different ratios along different axes — while simultaneously incorporating new agents into the formation. The key innovation is that the algorithm preserves the spectral properties of the graph Laplacian, a mathematical structure that defines communication links between agents. This ensures that the entire system remains stable and coordinated as new nodes are added, even in arbitrary dimensions. The work has been accepted at IFAC 2026 and validated with simulation examples.

The practical implications are significant for industries relying on autonomous swarms. In precision agriculture, a team of drones could dynamically expand as more units are deployed to cover a field, while maintaining a flexible formation that adapts to obstacles. In warehouse logistics, robot fleets could merge separate groups into one formation during a coordinated transport task. The ability to scale non-uniformly adds another layer of adaptability: formations can squeeze through narrow corridors (scaling down along one axis) while stretching along another to maintain coverage. By removing the fixed-agent constraint, this research pushes multi-agent systems closer to truly autonomous, scalable operations in unstructured environments. The code and simulations have not been publicly released, but the paper provides a theoretical foundation that other engineers can build upon.