No description, website, or topics provided.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
src
test
.gitignore
LICENSE
Project.toml
README.md

README.md

Environment

This package aims to provide a basic environment for experiments on reinforcement learning algorithms in Julia.

Supported Environments

Currently, only OpenAIGym is supported. More environments are coming soon.

TODO

  • ViZDoom

    Will be supported using CxxWrap.jl. Due to the grammar change in the latest CxxWrap.jl, the client of ViZDoom need to be updated.

  • bullet3

    Since there exists a PyBullet-gym repo, no need to add another wrapper. See more discussion here

  • Carla

    Carla is a very interesting project that I have been watched for a long time. The python client communicate with the server through socket. It seems possible to write a client in pure julia. I'll evaluate the workload later.

  • Support distributed environment so that we can take a batch of actions.

  • Multi-agent support.

  • Benchmark

Install

  1. Install OpenAIGym first.

  2. Add Environment.jl as a dependency.

    pkg> add https://github.com/findmyway/Environment.jl

Examples

Checkout gymenv.jl for usage.