EFLUX: Elastic Multi-Robot Formation Navigation and Adaptation with Agentic LLMs

Jul 15, 2026

EFLUX is a geometry-grounded framework that leverages large language models (LLMs) for adaptive, elastic multi-robot formation navigation. The system enables robot teams to autonomously deform and reconfigure their formations in cluttered environments by reasoning over both deformation and reconfiguration actions. Simulation and hardware experiments demonstrate that EFLUX reduces deadlock and navigation failures compared to baseline methods, while maintaining coordinated team behavior.

Why it matters: This work shows that LLMs can enable more adaptive and robust coordination in multi-robot systems, advancing autonomous navigation in complex, real-world environments.

Full story at: arXiv Robotics ↗