MiniMBRL is a minimal implementation of model-based reinforcement learning algorithms. The goal is to implement the most basic version of the algorithms to understand the core concepts and to build on top of them.
- Install and activate a new python3.8 virtualenv:
virtualenv mbrl_venv --python=python3.8
source mbrl_venv/bin/activate- Install requirements:
pip install "gymnasium[all]"
pip install mujocoBelow is a structured list of algorithms that will be implemented, organized by category:
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Dyna: An Integrated Architecture for Learning, Planning, and Reacting | Link | Dyna Notes | Done |
| World Models | Link | - | Ongoing |
| PlaNet: Learning Latent Dynamics for Planning from Pixels | Link | - | Planned |
| Dream to Control: Learning Behaviors by Latent Imagination (Dreamer V1) | Link | - | Planned |
| Mastering Atari with Discrete World Models (Dreamer V2) | Link | - | Planned |
| DayDreamer: World Models for Physical Robot Learning | Link | - | Planned |
| Planning to Explore via Self-Supervised World Models | Link | - | Planned |
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Neural Network Dynamics for Model-Based Deep Reinforcement Learning | Link | - | Planned |
| Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models | Link | - | Planned |
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Transformers are Sample-Efficient World Models | Link | - | Planned |
| The Benefits of Model-Based Generalization in Reinforcement Learning (ICML 2024) | Link | - | Planned |
| Facing Off World Model Backbones: RNNs, Transformers, and S4 (NeurIPS 2023) | Link | - | Planned |
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Curiosity-driven Exploration by Self-supervised Prediction | Link | - | Planned |
| Curious exploration via structured world models yields zero-shot object manipulation | Link | - | Planned |
| SENSEI: Semantic Exploration Guided by Foundation Models to Learn Versatile World Models | Link | - | Planned |
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Learning to Model the World with Language | Link | - | Planned |
| Language-Guided World Models: A Model-Based Approach to AI Control | Link | - | Planned |
| Paper | Link | Personal Notes | Status |
|---|---|---|---|
| Contrastive Learning of Structured World Models | Link | - | Planned |