mbrl

Model-based reinforcement learning in TensorFlow

Robotics: Science and Systems (RSS) 2025 | Action Flow Matching for Continual Robot Learning | Online and Non-episodic Robot Dynamics Model Refinement.

A gym-like classical control benchmark for evaluating the robustnesses of control and reinforcement learning algorithms.

Study Model-Based Policy Optimization by varying the model estimator classes (e.g Decision Trees vs MLP)

UrbanWorldModel, UrbanSim WM is a DreamerV3-style world model for simulating city-scale air quality, energy, and mobility under policy interventions. It provides a Dockerized FastAPI + Next.js app with WAQI ETL integration, real-time PM2.5 inference, interactive charts, and a training pipeline in PyTorch.

A reinforcement learning project exploring different RL algorithms. Namely: QLearning, DQN, PPO, TreeQN, SAVE,

Minimal model-based RL algorithm implementations

Production-grade Model-Based Reinforcement Learning system with learned world models (VAE + RSSM) and planning (MCTS, CEM) for sample-efficient decision making in continuous control environments.