Scholarlink

The Snake Game features simple rules, making AI for this game a fundamental problem that can be extended to other challenges. Currently, various learning algorithms have been developed for the Snake Game, raising a question about which algorithm yields the best results. This research compares the efficiency of reinforcement learning algorithms: Deep Q-Learning (DQN), Double Deep Q-Learning (DDQN), Proximal Policy Optimization (PPO), and Advantage Actor-Critic (A2C). The algorithms were tested under various conditions, using training step variations and analysis through Friedman Test and Nemenyi Test. Results showed the optimal training was determined at 5,000,000 steps. PPO achieved highest efficiency (0.9133), followed by A2C (0.7615), DQN (0.6644), and DDQN (0.4822). These findings offer insights for reinforcement learning research and lay a foundation for future research.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement Learning Evaluation for Solving Fundamental AI Problem

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Theenida Kunklon,
  • Thepparit Banditwattanawong

摘要

The Snake Game features simple rules, making AI for this game a fundamental problem that can be extended to other challenges. Currently, various learning algorithms have been developed for the Snake Game, raising a question about which algorithm yields the best results. This research compares the efficiency of reinforcement learning algorithms: Deep Q-Learning (DQN), Double Deep Q-Learning (DDQN), Proximal Policy Optimization (PPO), and Advantage Actor-Critic (A2C). The algorithms were tested under various conditions, using training step variations and analysis through Friedman Test and Nemenyi Test. Results showed the optimal training was determined at 5,000,000 steps. PPO achieved highest efficiency (0.9133), followed by A2C (0.7615), DQN (0.6644), and DDQN (0.4822). These findings offer insights for reinforcement learning research and lay a foundation for future research.