DR-PETS (Distributionally-Robust PETS)

Implementation of DR-PETS
(paper) with:

• Probabilistic Ensembles with Trajectory Sampling (PETS)
• Wasserstein regularizer (ε) for distributional robustness
• Online dynamics + density retraining loop

Quick Start

git clone https://github.com/<YOU>/DR-PETS.git
cd DR-PETS
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Collect random data, then train+plan online
python main_train_and_plan.py --config configs/cartpole.yaml