环境配置

建议在python3下运行

Install dependencies

需要另外安装的第三方库主要有：theano keras gym matplotlib

可以参考以下代码进行安装 pip3 install theano keras gym matplotlib

使用前

1. 需将环境文件ChannelEnv.pv放到gym的安装目录下。如/usr/local/lib/python3.6/site-packages/gym/envs/classic_control/
2. 打开1中目录中的__init__.py文件，加入语句： from gym.envs.classic_control.ChannelEnv import ChannelEnv
3. 进入gym安装目录的gym/gym/envs/打开__init__.py，添加代码：

register( id='Channel-v0',
entry_point='gym.envs.classic_control:ChannelEnv',
max_episode_steps=200, reward_threshold=100.0, )

完成上述操作后

使用python3运行channel1.py

Set up environment using Anaconda

Install anaconda

see Conda documentation#Installation and choose anaconda3

Create Environments

1. Create a conda environment with python 3.6: conda create --name gymtestbed python=3.6
2. Switch to created environment: source <anaconda path>/bin/activate gymtestbed. For example, source ~anaconda3/bin/activate gymtestbed. see also Conda documentation#Managing Environments

Install dependencies

• See Conda documentation#Managing packages to install theano, keras and matplotlib
• Install gym using:

git clone https://github.com/openai/gym.git
cd gym
pip install -e '.[classic_control]'

see also the github page of gym.

Add channel environment

1. Open <where you clone the gym respo>gym/gym/envs/__init__.py, add:

register( id='Channel-v0',
entry_point='gym.envs.classic_control:ChannelEnv',
max_episode_steps=200, reward_threshold=100.0, )

2. Open <where you clone the gym respo>/gym/gym/envs/classic_control/__init__.py and add: from gym.envs.classic_control.ChannelEnv import ChannelEnv

Run the program

1. Use git clone to clone this respo.
2. cd channel_dqn and open run.sh and change the line 3 and 4 of code.

#!/bin/bash
# Update the environment, then run the agent
cp ChannelEnv.py <where you clone the gym respo>/gym/gym/envs/classic_control/ChannelEnv.py
cp channelConfig.py <where you clone the gym respo>/gym/gym/envs/classic_control/channelConfig.py
python ./channel1.py

3. Use chmod +x ./run.sh to give execute permission.
4. Run the program.

./run.sh

The run.sh file will copy the ChannelEnv.py into the gym and execute program channel1.py