Switch to use RLCore

[findmyway]

Motivation

To help people write different kinds of algorithms more easily and increase cooperation. I decided to split the original https://github.com/JuliaReinforcementLearning/ReinforcementLearning.jl into separate packages.

• Some basic interfaces and default implementations will go to https://github.com/JuliaReinforcementLearning/ReinforcementLearningBase.jl
• Some reusable core components are moved into https://github.com/JuliaReinforcementLearning/ReinforcementLearningCore.jl
• Originally many tabular reinforcement learning algorithms are in ReinforcementLearning.jl. Considering that maybe they are not that reusable, I'll move them here instead.

Current Status

• ReinforcementLearningBase.jl is registered. But it's not that stable yet. I just made a separate branch https://github.com/JuliaReinforcementLearning/ReinforcementLearningBase.jl/pull/16/files to track the latest progress.
• ReinforcementLearningCore.jl is also registered. Similar to RLBase, it is not very stable. Many reusable components are not ported yet. Dev ReinforcementLearningCore.jl#18 will be used to track the latest progress.

Following is the current progress:

• Update original environments in this repo to use the new API defined in RLBase
• Chapter01
• Chapter02
• Chapter03
• Chapter04
• Chapter05
• Chapter06
• Chapter07
• Chapter08
• Chapter09
• Chapter10
• Chapter11
• Chapter12
• Chapter13

If anyone else would like to contribute, just reply to this thread and mark the chapters. Generally speaking, the development steps are:

1. Clone this repo
2. Clone https://github.com/JuliaReinforcementLearning/ReinforcementLearningBase.jl and checkout the dev branch.
3. Clone https://github.com/JuliaReinforcementLearning/ReinforcementLearningCore.jl/pull/18 and checkout the dev branch.
4. cd /path/to/ReinforcementLearningAnIntroduction.jl && julia --project
5. ] dev /path/to/ReinforcementLearningBase.jl
6. ] dev /path/to/ReinforcementLearningCore.jl
7. ] instantiate
8. Open the notebooks in the notebooks folder.
9. Make necessary changes in each repo and then make pull requests.

The Next Step

1. Once all changes in this repo are merged, a new release of each package will be created. ReinforcementLearningBase.jl and ReinforcementLearningCore.jl will be reexported in ReinforcementLearning.jl.
2. Update & register OpenSpiel.jl to use the [email protected] once add OpenSpiel JuliaPackaging/Yggdrasil#457 (comment) gets merged
3. Update https://github.com/JuliaReinforcementLearning/ReinforcementLearningEnvironments.jl to use the new API use RLBase instead ReinforcementLearningEnvironments.jl#29
4. Create ReinforcementLearningZoo to hold DRL related algorithms
5. Adapt AlphaZero.jl to use OpenSpiel.jl
6. Design MultiAgentReinforcementLearning
7. Design ModelBasedReinforcementLearning

---

cc: @sriram13k, @jonathan-laurent, @Martyn-R, @rezabayani

Comments are welcomed.