/DualSMC

Primary LanguagePython

DualSMC

This is the PyTorch code for the following paper to be presented at IJCAI-20, in which we proposed a deep model for POMDPs planning.

DualSMC: Tunneling Differentiable Filtering and Planning under Continuous POMDPs by Yunbo Wang*, Bo Liu*, Jiajun Wu, Yuke Zhu, Simon S. Du, Li Fei-Fei, and Joshua B. Tenenbaum. (* indicates equal contributions)

[Project Page] [Paper]

We currently release the code for the first experimental domain, the floor positioning. The rest of the code for the 3D light-dark navigation domain and the modified Reacher domain will be released later. All experiments are reproducible.

Citation

Please remember to cite our paper if you use the code.

@inproceedings{wang2019dualsmc,
  title     = {{DualSMC}: Tunneling Differentiable Filtering and Planning under Continuous {POMDPs}},
  author    = {Wang, Yunbo and Liu, Bo and Wu, Jiajun and Zhu, Yuke and Du, Simon S. and Fei-Fei, Li and Tenenbaum, Joshua B.},
  booktitle = {IJCAI},
  pages     = {4190--4198},
  year      = {2020},
}