Manned and Unmanned Aerial Vehicles Cooperative Combat Framework Based on Large Language Models

The collaborative combat of manned and unmanned aerial vehicles is one of the principal directions for the future development of aerial combat systems, facing the challenge of excessive decision-making and operational burdens on pilots during UAV control. This paper proposes a novel framework for the collaborative combat of manned and unmanned aerial vehicles, incorporating large language model(LLM) into the process of aircraft pilots commanding and controlling multiple unmanned aerial vehicles (UAVs). The framework utilizes LLM for complex semantic understanding and monitoring of task instruction execution. It allows pilots to issue task instructions to UAVs using non-standard natural language. The received natural language task instructions are matched with the preloaded policy library of the designed task executor in UAVs, and an appropriate policy is selected for execution. During task execution, UAVs provide feedback on the task execution status to manned aircraft at key nodes, and continue task execution upon confirmation by manned aircraft until task completion or receipt of new task instructions. The framework is tested in typical beyond-visual-range combat scenarios of manned and unmanned aerial vehicle collaboration. It exhibits good human-machine interaction, robustness, trustworthiness, explainability, and effectively reducing the decision-making and operational burdens on pilots. The research findings of this paper can be widely applied to various task scenarios where humans and robots collaborate to accomplish tasks, providing a feasible technical route for the collaborative combat of manned and unmanned aerial vehicles.