1. School of Management, Jinan University, Guangzhou 510632, China; 2. Hikvision Digital Technology Co. Ltd., Hangzhou 310051, China; 3. School of Public/Emergency Management, Jinan University, Guangzhou 510632, China; 4. The Laboratory for Military-civilian Integration Emergency Command and Control, Jinan University, Guangzhou 510632, China
Abstract：Modern battlefield informatization and multimodal features pose severe challenges to command and control（C2）systems. This paper proposes a software-defined C2 parallel testing method integrating multimodal perception, large language model（LLM）reasoning, dynamic knowledge graphs, and DDD-III simulation. The hierarchical architecture （perception, cognition, action layers） enables intelligent decision-making, while LLM-based multimodal agent models enhance situational understanding. Federated learning and game theory facilitate distributed collaboration. Experiments show that the method outperforms traditional systems in task completion rate and resource efficiency, offering a novel paradigm for C2 optimization in complex battlefields.
2025, Vol. 11 
