Paper | Twitter

Weikang Wan*, Ziyu Wang*, Yufei Wang*, Zackory Erickson, David Held

This repository contains the official implementation of DiffTORI, which utilizes Differentiable Trajectory Optimization as the policy representation to generate actions for deep Reinforcement and Imitation learning.

This project is organized into two main components:

• mbrl
  This folder contains the code for model-based reinforcement learning using DiffTORI. For detailed instructions on installation, training, and troubleshooting, please refer to the mbrl/README.md.

• DiffTORI_IL_Metaworld
  This folder provides the implementation of DiffTORI for imitation learning for Metaworld tasks. Detailed guidance on data generation, training, and evaluation can be found in the DiffTORI_IL_Metaworld/README.md.

For further details on the methodology, please refer to our paper:
DiffTORI: Differentiable Trajectory Optimization for Deep Reinforcement and Imitation Learning

If you find our work useful, please consider citing:

@article{wan2025difftori,
  title={DiffTORI: Differentiable Trajectory Optimization for Deep Reinforcement and Imitation Learning},
  author={Wan, Weikang and Wang, Ziyu and Wang, Yufei and Erickson, Zackory and Held, David},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={109430--109459},
  year={2025}
}