Root / CYBERNETICS AND PHYSICS / Volume 15, 2026, Number 2 / Accelerated formation-recovery layer for swarm control with a single global command

Accelerated formation-recovery layer for swarm control with a single global command

Konstantin Amelin, Andrey Chernov, Oleg Granichin

The paper proposes an accelerated local formation-recovery layer for robot swarms driven by a single global command. The global command specifies the motion of the group, while each robot uses only neighbor-to-neighbor communication to restore formation consistency. The local correction problem is formulated as a regularized strongly convex consensus objective and solved using the Accelerated Local Voting Protocol (ALVP), an accelerated variant of the standard Local Voting Protocol (LVP). In the considered aerial-swarm implementation, the same recovery principle is applied to heading, altitude, and horizontal position alignment. The approach is evaluated in a high-fidelity multirotor software-in-the-loop (SITL) environment with ArduPilot SITL, emulated inter-robot communication, packet losses, and bounded control inputs. In parameter-wise recovery experiments, ALVP reduces the mean convergence time by 37.2%, 39.8%, and 38.6% for heading, altitude, and horizontal position, respectively, compared with LVP. In stationary formation assembly, ALVP remains faster for all tested swarm sizes and shows lower variance for larger swarms. Commanded-motion experiments further indicate that local recovery is necessary under global motion and that ALVP improves post-maneuver recovery in turning and circular trajectories. These results suggest that accelerated local voting is a promising mechanism for improving transient formation recovery in single-interface swarm control.
CYBERNETICS AND PHYSICS, VOL. 15, NO. 2, 2026, 119–128
https://doi.org/10.35470/2226-4116-2026-15-2-119-128

File: download