게임캐릭터 및 객체 배치를 위한 AI기반 다중 에이전트 협업 시스템 연구

초록

This study proposes a novel collaborative model based on Multi-Agent Reinforcement Learning (MARL) to address the automation of object and character placement, a critical challenge in game, animation, and VR/AR production. Conventional clustering or rule-based placement methods suffer from several limitations, including static layouts, heavy reliance on manual design work, and insufficient modeling of interactions among objects. In the proposed approach, each character and object is defined as an independent agent that autonomously adapts to environmental changes through reward-driven learning. This enables the dynamic generation of realistic spatial arrangements, thereby improving production efficiency and providing immersive spatial configurations that closely resemble real-world environments.

키워드