This repository contains a collection of model based reinforcement learning algorithms implemented in PyTorch. The goal is to provide a simple and easy to understand implementation of the algorithms, so that they can be used as a reference for future projects. The algorithms are implemented in a modular way, so that they can be easily extended and modified.
The repository is still under development, so some features may not be implemented yet. If you find any bugs or have any suggestions, please let me know.
The repository is mainly for educational purposes and does not provide an extension of the research in the respective paper.
For now, I am implementing the following papers:
- Learning Latent Dynamics for Planning from Pixels
- World Models
- Dream to Control: Learning Behaviors by Latent Imagination
The repository is structured as follows:
models/: Contains the neural network models used in the algorithms.envs/: Contains the environment wrappers used in the algorithms.train: Contains the training scripts for the algorithms.
Corresponding blog post where all the math is derived can be found here: https://medium.com/@lukasbierling/training-agents-to-plan-in-latent-space-a-technical-overview-f4380a94ec88
source: https://arxiv.org/pdf/1912.01603