Design of AI-CGF for Analysis and Training in Warship Combat M&S

Hyun-Geun Kim, Jae-Woo Kim, Jin-Won Kim, Bum-Joon Park, Kang-Moon Park, Sung-do Chi

Abstract: This study proposes a solution to the resource-intensive issue and environmental pollution associated with military training. By utilizing evolutionary and deep learning techniques within naval analyzing and training Modeling and Simulation (M&S) environments, this study shows significant progress in both the Analyzing and Training M&S. In Analyzing M&S, optimal tactical scenarios tailored to specific situations were automatically generated through evolutionary, supervised, and reinforcement learning. The traditional role of human operators was replaced with AI-based models. In Training M&S, reinforcement learning was used with optimization functions such as A3C, PPO, SAC, and DAC. AI-CGF (Artificial Intelligence-based Computer-Generated Force) was developed to replace human controllers in both Analyzing and Training M&S, which significantly reduces the requirement for human operators and empowers the generation of new Red Force scenarios. The research team achieved a 95% efficiency level in red-force scenarios effectively countering blue-force scenarios. These results are expected to enhance military training effectiveness and alleviate environmental impact, with potential real-world applications in the realm of naval M&S.

External IDs:dblp:journals/access/KimKKPPC25