Uav swarms manage communication to improve search and rescue success
Managing Context and Communication in Distributed Agentic UAV Swarms
Multiagent SystemsArtificial IntelligenceMachine LearningNetworking and Internet ArchitectureRobotics
Summary
UAV swarms equipped with small language models can work together without a central controller, but this causes too much communication and confusion over time. The authors created a system where each drone keeps track of important information locally and only shares new, relevant facts with other drones. This selective communication helps the drone team finish search-and-rescue missions more reliably and efficiently than other methods. Their tests using ten drones showed better accuracy and less wasted communication.
What this means in practice
- •For drone fleet operators: Manage drone communication to improve mission success rates and reduce data exchange in multi-UAV search-and-rescue operations.
- •For emergency response coordinators: Deploy autonomous drone teams that coordinate efficiently in challenging environments without relying on centralized control infrastructure.
Authors
Andrea Iannoli, Ivan Zyrianoff, Angelo Trotta, Lorenzo Gigli, Marco Di Felice
Abstract
Unmanned aerial vehicle (UAV) swarms increasingly rely on language-model agents to provide adaptive mission-level reasoning in uncertain environments. Fully distributed control, in which each UAV hosts an independent Small Language Model (SLM), removes reliance on a centralized coordinator but introduces an information-management problem: long-running interaction histories can degrade the reasoning context, while indiscriminate information dissemination increases communication and inference overhead. We address these challenges with a distributed UAV-agent architecture that enables continuous local SLM control through an event-driven reason-act-observe lifecycle. Runtime knowledge is represented as structured atomic notes and organized into core, local, and peer-specific memory. A deterministic interest-aware gossip engine selectively disseminates these notes according to recipient-specific semantic novelty and recency. We evaluate the architecture using ten UAVs in a simulated search-and-rescue mission. Our approach completes all experimental runs, whereas unrestricted flooding messages completes only 70-85\%, and delegating forwarding decisions to the SLM prevents mission completion in every run. Compared with unrestricted flooding, our approach approximately halves inference-token consumption, reduces transmitted data, and achieves lower survivor-count error.