TY  - JOUR
T1  - AI Fire Support Officer: Military Decision Support System  Based on Reward Adaptive Reinforcement Learning
AU  - Lee, Jaehwi 
AU  - Eom, Chanin 
AU  - Kim, Chan 
AU  - Kim, Kyeongsoo 
AU  - Lee, Hyeongdo 
AU  - Kang, Hyunsu 
AU  - Kwon, Minhae 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2026
DA  - 2026/1/1
DO  - 10.7840/kics.2026.51.1.209
KW  - Military decision-making
KW  - Command decision support
KW  - Weapon-target assignment
KW  - Reinforcement learning
AB  - Recent studies in military decision support have actively explored deep reinforcement learning (RL)
approaches to automate complex battlefield decision-making processes. This paper proposes a reward-adaptive
RL-based firepower operation system designed to support command decisions in dynamic combat environments.
The proposed system perceives battlefield situations through a perception module and makes decisions to
achieve the commander’s desired effects. The decision-making module integrates both pre-collected and online
interaction data while employing a reward-adaptive selective imitation mechanism to enhance sample efficiency
and stability simultaneously. Through simulated battlefield scenarios, the proposed system demonstrated an
average 29% improvement in mission achievement compared to conventional RL and heuristic-based methods,
while effectively satisfying given operational constraints.