Real-time UAV swarm autonomy using edge-deployed small language models

Authors: NARASIMHAMURTHY, K - The University Of Texas At Dallas; SETHURANI, P - University Of Texas At Tyler; ANAND, A - University Of Texas At Tyler; SUNDARAVADIVEL, P - University Of Texas At Tyler; Torbert Iii, Henry; TAMIL, L - The University Of Texas At Dallas

Submitted to: Meeting Abstract
Publication Type: Proceedings
Publication Acceptance Date: 3/14/2026
Publication Date: 4/10/2026
Citation: Narasimhamurthy, K., Sethurani, P.M., Anand, A., Sundaravadivel, P., Torbert III, H.A., Tamil, L. 2026. Real-time UAV swarm autonomy using edge-deployed small language models. 19th IEEE Dallas Circuits and Systems Conference.

Interpretive Summary: This publication demonstrates an edge-enabled autonomous UAV swarm framework in which onboard intelligence enables real-time decision-making without reliance on cloud connectivity. The system highlights how lightweight language-based reasoning can be integrated directly into drone autonomy pipelines to support mission execution, coordination, and adaptive behavior. In the demonstrated setup, each UAV performs onboard inference to interpret mission context and execute autonomous actions while maintaining low latency and energy efficiency. The results highlight a viable pathway toward robust, scalable UAV swarm autonomy using embedded AI, supporting applications that require rapid decision-making in communication-limited environments.

Technical Abstract: This publication demonstrates an edge-enabled autonomous UAV swarm framework in which onboard intelligence enables real-time decision-making without reliance on cloud connectivity. The video demonstration showcases autonomous drone operation driven by a small language model (SLM) optimized for deployment across heterogeneous hardware platforms, including NVIDIA Jetson Orin, edge boards, and companion laptops. The system highlights how lightweight language-based reasoning can be integrated directly into drone autonomy pipelines to support mission execution, coordination, and adaptive behavior. In the demonstrated setup, each UAV performs onboard inference to interpret mission context and execute autonomous actions while maintaining low latency and energy efficiency. Hardware-aware optimization allows the same SLM architecture to run consistently across embedded and non-embedded devices, enabling seamless transitions between onboard autonomy and ground-based supervision. The demonstration illustrates realtime responsiveness, stable execution, and decentralized intelligence, which are critical for scalable UAV swarm operations. By validating autonomous behavior through an end-to-end video demonstration, this work emphasizes the practicality of deploying language-driven reasoning models directly at the edge. The results highlight a viable pathway toward robust, scalable UAV swarm autonomy using embedded AI, supporting applications that require rapid decision-making in communication-limited environments.