TY  - JOUR
T1  - Design and Implementation of a Simulator for the Application of Cooperative Multi-Agent Systems in Naval Battle Environments
AU  - Kim, Kyungbeom 
AU  - Moon, Sujeong 
AU  - Jin, Woobeen 
AU  - Lee, Hyukjoon 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2024
DA  - 2024/1/1
DO  - 10.7840/kics.2024.49.1.165
KW  - Multi-Agent
KW  - cluster control system
KW  - reinforcement learning
KW  - MA-POCA
KW  - USV
KW  - Naval weapon system
AB  - With the recent changes in the technology and industrial system of the 4th Industrial Revolution, research on innovative future unmanned systems continues to increase in the marine defense industry sector. There is a great interest in multi-agent cluster control systems in unmanned systems. In the case of multi-agent cluster control systems, there are usually many complex and dynamic problems, so reinforcement learning algorithms are effective to solve these problems.
However, there is a deficiency in simulators for verifying multi-agent algorithms in a realistic naval battle environment.
Therefore, this study proposes a simulator based on cooperative scenarios of Unmanned Surface Vehicles (USVs) utilizing the MA-POCA (Multi-Agent Posthumous Credit Assignment) reinforcement learning algorithm. Implemented simulator is utilized to compare the learning performance of various multi-agent algorithms through combat tests with simulated adversaries, specifically rule based USVs.