Dynamic reinforcement learning benchmarks

This repository contains three open source reinforcement learning environments that require the agent to adapt its behavior to or make use of dynamic elements in the environment in order to solve the task. The environments follow OpenAI's gym interface.

Installation

With python3.7 or higher run

pip install dyn_rl_benchmarks

Usage

After importing the package dyn_rl_benchmarks the environments

• Platforms-v1
• Drawbridge-v1
• Tennis2D-v1

are registered and can be instantiated via gym.make.

The following example runs Platforms-v1 with randomly sampled actions:

import gym

import dyn_rl_benchmarks

env = gym.make("Platforms-v1")

obs = env.reset()
done = False
while not done:
  action = env.action_space.sample()
  obs, rew, done, info = env.step(action)
  env.render()

How to cite

@article{gurtler2021hierarchical,
  title={Hierarchical Reinforcement Learning with Timed Subgoals},
  author={G{\"u}rtler, Nico and B{\"u}chler, Dieter and Martius, Georg},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  year={2021}
}