# Competition_Olympics-Wrestling
## Environment
<img src=https://jidi-images.oss-cn-beijing.aliyuncs.com/jidi/env74.gif width=600>
Check details in Jidi Competition [RLChina 智能体挑战赛 - 壬寅年秋赛季](http://www.jidiai.cn/compete_detail?compete=23)
### Olympics-Integrated
<b>Tags: </b>Partial Observation; Continuous Action Space; Continuous Observation Space
<b>Introduction: </b>Agents participate in the Olympic Games. In this series of competitions, two agents participate in the challenge arena sumo competition. The goal is to avoid falling off the challenge arena and try to knock each other off the challenge arena.
<b>Environment Rules:</b>
1. This game has two sides and both sides control an elastic ball agent with the same mass and radius, which is located at both ends of a circular playing field.
2. At the beginning of the competition, agent of both sides can move freely within the circular field, but they can't touch the boundary of the field, otherwise they will be regarded as falling off the challenge arena and out of bounds.
3. Agents can collide with each other, but they will lose a certain speed according to the friction coefficient of the sphere.
4. The agent has its own energy, and the energy consumed in each step is directly proportional to the applied driving force and displacement.
5. The energy of the agent recovers at a fixed rate at the same time.
6. If the energy decays to zero, the agent will be tired, resulting in failure to apply force.
7. When one side falls into the challenge arena or the environment reaches the maximum number of steps of 500, the environment ends.
<b>Action Space: </b>Continuous, a matrix with shape 2*1, representing applied force and steering angle respectively.
<b>Observation: </b>A dictionary with keys 'obs' and 'controlled_player_index'. The value of 'obs' contains a 2D matrix with shape of 40x40 and other game-releated infomation. The 2D matrix records the view of agent along his current direction. Agent can see walls, marking lines, opponents and other game object within the vision area. The value of 'controlled_player_index' is the player id of the game.
<b>Reward: </b>Each team obtains a +100 reward when scoring the ball into the opponent's goal, otherwise 0 point.
<b>Environment ends condition: </b>The game ends when reaching maximum number of 500 steps or one side falls into the challenge arena.
<b>Registration: </b>Go to (http://www.jidiai.cn/compete_detail?compete=23).
---
## Dependency
>conda create -n olympics python=3.8.5
>conda activate olympics
>pip install -r requirements.txt
---
## Run a game
>python olympics_engine/main.py --map football
## How to test submission
You can locally test your submission. At Jidi platform, we evaluate your submission as same as *run_log.py*
For example,
>python run_log.py --my_ai "random" --opponent "random"
in which you are controlling agent 1 which is green.
---
## Ready to submit
Random policy --> *agents/random/submission.py*
没有合适的资源?快使用搜索试试~ 我知道了~
AI-奥林匹克·相扑比赛高分强化学习智能体实现代码
共553个文件
json:178个
pyc:151个
py:132个
需积分: 1 2 下载量 120 浏览量
2022-10-16
17:32:57
上传
评论 1
收藏 7.81MB ZIP 举报
温馨提示
AI-奥林匹克·相扑比赛基于规则的智能体和基于强化学习的智能体解决方案,该方案训练出的智能体能进入对战排行榜前十。基于代码实现的规则,能进入排行榜前三。 1 手动提取特征(manual feature extraction) 2 重新封装环境 3 训练表现
资源详情
资源评论
资源推荐
收起资源包目录
AI-奥林匹克·相扑比赛高分强化学习智能体实现代码 (553个子文件)
events.out.tfevents.1664037970.DESKTOP-5TCVNMF.20592.0 432B
events.out.tfevents.1664037828.DESKTOP-5TCVNMF.7232.0 383B
events.out.tfevents.1664038397.DESKTOP-5TCVNMF.16792.0 40B
events.out.tfevents.1664037828.DESKTOP-5TCVNMF.7232.1 14KB
events.out.tfevents.1664038397.DESKTOP-5TCVNMF.16792.1 40B
events.out.tfevents.1664037970.DESKTOP-5TCVNMF.20592.1 40B
.gitignore 193B
.gitignore 87B
Competition_Olympics-Wrestling-main.iml 1KB
maps.json 80KB
scenario.json 44KB
202209291847_wrestling.json 27KB
202209282202_wrestling.json 27KB
202209291504_wrestling.json 27KB
202209291856_wrestling.json 27KB
202209261834_wrestling.json 26KB
202209262145_wrestling.json 26KB
202209291906_wrestling.json 25KB
202209291904_wrestling.json 25KB
202209291900_wrestling.json 25KB
202209272152_wrestling.json 23KB
202209292337_wrestling.json 23KB
202209221556_wrestling.json 22KB
202209232035_wrestling.json 21KB
202210071251_wrestling.json 20KB
202210052352_wrestling.json 20KB
202209292246_wrestling.json 20KB
202209282326_wrestling.json 19KB
202209232015_wrestling.json 19KB
202209230005_wrestling.json 18KB
202210071327_wrestling.json 17KB
202209232043_wrestling.json 17KB
202210071345_wrestling.json 16KB
202210071320_wrestling.json 16KB
202209230031_wrestling.json 16KB
202209230013_wrestling.json 15KB
202210071335_wrestling.json 15KB
202209230016_wrestling.json 13KB
202209292245_wrestling.json 13KB
202210071401_wrestling.json 12KB
202209230030_wrestling.json 12KB
202210071332_wrestling.json 12KB
202209230037_wrestling.json 12KB
202210071425_wrestling.json 11KB
202209282201_wrestling.json 11KB
202209221604_wrestling.json 11KB
202209221617_wrestling.json 11KB
202210071404_wrestling.json 11KB
202209230018_wrestling.json 11KB
202209292230_wrestling.json 11KB
202209221606_wrestling.json 10KB
202209232033_wrestling.json 10KB
202210071344_wrestling.json 10KB
202210071347_wrestling.json 10KB
202210071426_wrestling.json 9KB
202209232029_wrestling.json 9KB
202210071421_wrestling.json 9KB
202210071330_wrestling.json 9KB
202209221602_wrestling.json 8KB
202209232018_wrestling.json 8KB
202209232011_wrestling.json 8KB
202209212110_wrestling.json 7KB
202209232013_wrestling.json 7KB
202210071405_wrestling.json 7KB
202209232017_wrestling.json 7KB
202209232034_wrestling.json 7KB
202209232012_wrestling.json 7KB
202209230029_wrestling.json 6KB
202210071321_wrestling.json 6KB
202210071348_wrestling.json 6KB
202209230012_wrestling.json 6KB
202209221611_wrestling.json 6KB
202209232004_wrestling.json 6KB
202209221609_wrestling.json 6KB
202209230022_wrestling.json 6KB
202209232032_wrestling.json 6KB
202210071331_wrestling.json 6KB
202209230044_wrestling.json 6KB
202209221605_wrestling.json 6KB
202210071420_wrestling.json 6KB
202209232014_wrestling.json 6KB
202209230021_wrestling.json 6KB
202209230023_wrestling.json 6KB
202209221610_wrestling.json 5KB
202210071424_wrestling.json 5KB
202209230019_wrestling.json 5KB
202209221608_wrestling.json 5KB
202209230050_wrestling.json 5KB
202209230014_wrestling.json 5KB
202209230036_wrestling.json 5KB
202210071334_wrestling.json 5KB
202210071423_wrestling.json 5KB
202210071407_wrestling.json 5KB
202209230035_wrestling.json 5KB
202210071328_wrestling.json 5KB
202209221616_wrestling.json 5KB
202209281253_wrestling.json 5KB
202209292249_wrestling.json 4KB
202209292248_wrestling.json 4KB
202210071403_wrestling.json 4KB
共 553 条
- 1
- 2
- 3
- 4
- 5
- 6
二向箔不会思考
- 粉丝: 5328
- 资源: 23
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- BLOCK_TYPE_HEARTBEAT_D70A3465D4EE4E9_046141_dump_1st.dmp
- 项目方法测试调用接口工具
- studyupdate
- 基于西瓜数据集的决策树实现.zip
- 60套HTML网站源码-响应式-涵盖(简历&作品展示&商业&科技&培训&商城&课设等)-适配移动设备-解压即用.zip
- 贪心算法要点和难点实例代码解析
- 65套HTML网站源码-响应式-涵盖(简历&作品展示&商业&科技&培训&商城&课设等)-适配移动设备-解压即用.zip
- 多因素决策树的Python实现.zip
- 使用Python在莺尾花数据集上实现了决策树算法,文件里有数据集.zip
- python实现决策树.zip
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0