https://github.com/CWHer/CS410-AI-project

Top 1 solution for CS410: Artificial Intelligence snakes 3v3 :snake: game.
https://github.com/CWHer/CS410-AI-project

reinforcement-learning

Last synced: about 1 month ago
JSON representation

Top 1 solution for CS410: Artificial Intelligence snakes 3v3 :snake: game.

• Host: GitHub
• URL: https://github.com/CWHer/CS410-AI-project
• Owner: CWHer
• License: mit
• Created: 2021-11-14T05:33:55.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2022-06-12T07:57:15.000Z (over 2 years ago)
• Last Synced: 2024-06-11T07:33:33.330Z (6 months ago)
• Topics: reinforcement-learning
• Language: Python
• Homepage:
• Size: 3.92 MB
• Stars: 4
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-cs - @CWHer @yuri-you, 2021 Fall

README

        
# CS410-AI-project

本分支用于本地evaluate:thinking:

目前包括: random, rl(DDPG), greedy, heuristic, dqn(mlp), rot_dqn, IL, defense_dqn, defense_IL

训练好的模型可以在[Link](https://drive.google.com/drive/folders/1vOMKE5JC1PCZ6HCpZIJ9L2SpiNcOY5yG?usp=sharing)找到，其中包含DQN模型（用于rot_dqn和defense_dqn）和IL模型（用于IL和defense_IL），将model文件移入对应的`agent/xxx`文件夹，并修改`network.py`中加载模型的代码即可运行。

注：`dqn_models`包含若干效果较好的DQN模型。`il`仅包含一个效果较好的IL模型以及其拆分后的两个part（测试平台有文件大小限制:weary:）。

## 如何运行

```python

python evaluation_local.py --my_ai rl --opponent random

python run_log.py --my_ai "random" --opponent "rl"

```

## Description of branches

| Branch Name | Content                    |

| ----------- | -------------------------- |

| greedy      | greedy snake               |

| AlphaZero   | Alpha Zero snake           |

| D3QN        | (deserted)                 |

| single-D3QN | DQN without rotation trick |

| rot-D3QN    | DQN with rotation trick    |

| IL          | Imitaion Learning          |

Note: defense module is in `main`.

## Final Result

![](img/result.png)