Skip to content

Latest commit

 

History

History
48 lines (42 loc) · 1.67 KB

README.md

File metadata and controls

48 lines (42 loc) · 1.67 KB

Force-Aware ProDMP

This repository provides a utility for fast implementation of the Force-Aware ProDMPs. It provides the following 3 commands:

  • get_FAProDMP to generate a FA-ProDMP from a given set of demonstrations
  • condition_FAProDMP_on_force to condition the FA-ProDMP
  • blend_trajectories to blend between 2 trajectories

Getting Started

Ensure that all submodules are initialized:

git submodule update --init

Install Conda and run the following command:

conda env create -n <env_name> -f conda_env.yml

After this, you can run the demonstration notebook in the demo folder.

Expected Data Format

The utility expects each demonstration in the form of a Pandas DataFrame. Each time step should contain positional and force information. Additionally, we assume that the DataFrame is indexed on the time information.

Dependency Requirements

This utility depends on the following packages:

  • pandas
  • numpy
  • torch
  • MP_PyTorch (preferably using the submodule provided)
    • matplotlib

We recommend usage of Python 3.9.0. A conda configuration is provided in conda_env.yml.

Citation

If you interest this project and use it in a scientific publication, we would appreciate citations to the following information:

@misc{lödige2024useforcebot,
      title={Use the Force, Bot! -- Force-Aware ProDMP with Event-Based Replanning}, 
      author={Paul Werner Lödige and Maximilian Xiling Li and Rudolf Lioutikov},
      year={2024},
      eprint={2409.11144},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2409.11144}, 
}