Environments for non-stationary meta-RL problems
This repository contains the environments investigated in the masterthesis Meta-Reinforcement Learning in Non-Stationary and Dynamic Environments
.
Its current main usage is supporting the implementation of the Continuous Environment Meta-Reinforcement Learning (CEMRL)
algorithm (see CEMRL).
This repository is inspired by rand_param_envs.
Follow the installation instructions from the CEMRL repository.
- Most environments are based on parent classes in
meta_rand_envs/base.py
- Environments not actively used in the CEMRL implementation are deprecated and might not work.