This Repository is implementation of Relational Deep Reinforcement Learning to BoxWorld Environment.
The Reinforcement Learning Algorithm is A2C, but it's very easy to change the base algorithm.
- Python: 3.6.1
- Tensorflow: 1.11.0
- Tensorboard:1.11.0
- OpenAI Gym: 0.15.4
- stable-baselines, commit hash: 98257ef8c9bd23a24a330731ae54ed086d9ce4a7a1ab7a1c2903e7e1c38756d8cdf7a54a5fd5781e.
- Already exists in the project, but you need to install the dependencies of stable_baselines
The versions are just what I used and not necessarily strict requirements.
Go to the env/gym-box-world folder and run the command :
pip install -e .
This will install the box-world environment. Now, you can use this enviroment with the following:
import gym
import gym_boxworld
env_name = 'BoxRandWorld'
env_id = env_name + 'NoFrameskip-v4'
env = gym.make(env_id,level='easy')
All training code is contained within main.py. To view options simply run:
python main.py --help
An example:
python main.py BoxRandWorld -env_level easy RelationalPolicy
The following is the relation(attention) weight diagram of the agent
0:Dark 1:White
The greater the weight, the more the color tends to be white
