Redefining Human Command of sUAS Swarms Leveraging GenAI for Dynamic Battlefield Adaptation

Petra Dataĺs project introduces an innovative approach to controlling small Unmanned Aerial System (sUAS) swarms, leveraging Generative AI for real-time adaptability in dynamic battlefield environments. The solution applies MITĺs LLaMAR framework for multi-agent, long-horizon planning, alongside advanced vision-language models like Googleĺs PaliGemma, to enable autonomous communication and cooperation among drones. Addressing the limitations of current systems that depend on human intervention, this technology allows drones to interact independently, interpreting mission requirements and coordinating actions. The project will validate the feasibility of deploying edge vision-language models on a small aerial network, demonstrating peer-to-peer coordination among drones in a simulated search-and-rescue scenario. This adaptive swarm intelligence reduces cognitive load on operators and enhances mission success through resilient, autonomous decision-making. Designed for scalability, the system aligns with DoDĺs strategic goals, offering a reliable pathway for integrating AI-driven swarm intelligence into diverse, real-time operational settings.