Code for CEM-GD: Cross-Entropy Method with Gradient Descent Planner for Model-Based Reinforcement Learning.
Implementation for the CEM-GD planner can be found in mbrl/planning/gradient_optimizer.py.
pets-experiments.py is a notebook for running the end-to-end MBRL algorithm based on PETS + CEM-GD and the experiments in the paper.
After running experiments, figures can be generated using figure_generator.ipynb
Portions of this code, including the implementation of PETS, is adapted from mbrl-lib under the MIT license.