This package aims to provide a basic environment for experiments on reinforcement learning algorithms in Julia.
Currently, only OpenAIGym is supported. More environments are coming soon.
[ ] 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.
[ ] 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
Environment.jl as a dependency.
pkg> add https://github.com/findmyway/Environment.jl
Checkout gymenv.jl for usage.
over 1 year ago