Many reinforcement learning environments in the Python world are not available in Julia yet. Though we can leverage PyCall.jl to interact with them, the overheads usually make this approach unacceptable. Following are some experiments we'd like to have, either by wrapping the underlying C/C++ libraries or rewriting in Julia from scratch.
Beside writing environments, it would be great if a unified wrapper is also provided in ReinforcementLearningEnvironments.jl.
Note that each algorithm is suggested to provide at least one reproducible experiment.
Twin Delayed DDPG