The study introduces MAIDCRL (Multi-Agent Dense-CNN Reinforcement Learning), a reinforcement learning algorithm that incorporates convolutional layers into the deep model architecture for better multi-agent control in StarCraft Multi-Agent Challenge scenarios. The CNN-enabled MAIDCRL has shown significantly improved learning performance, faster learning rate, and better stability and robustness compared to the existing MAIDRL, particularly in more complex scenarios. The algorithm also enhances the fine-grained decision-making process of artificial intelligence systems.
Publication date: 13 Feb 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2402.07890
