Nick Heppert, Toki Migimatsu, Brent Yi, Claire Chen, Jeannette Bohg
arXiv | website | Presented at IROS 20222, Kyoto, Japan
Robots deployed in human-centric environments may need to manipulate a diverse range of articulated objects, such as doors, dishwashers, and cabinets. Articulated objects often come with unexpected articulation mechanisms that are inconsistent with categorical priors: for example, a drawer might rotate about a hinge joint instead of sliding open. We propose a category-independent framework for predicting the articulation models of unknown objects from sequences of RGB-D images. The prediction is performed by a two-step process: first, a visual perception module tracks object part poses from raw images, and second, a factor graph takes these poses and infers the articulation model including the current configuration between the parts as a 6D twist. We also propose a manipulation-oriented metric to evaluate predicted joint twists in terms of how well a compliant robot controller would be able to manipulate the articulated object given the predicted twist. We demonstrate that our visual perception and factor graph modules outperform baselines on simulated data and show the applicability of our factor graph on real world data.
This repository is the official implementation
The code was developed and tested under Ubuntu 20.04 and python version 3.8.
Install all dependencies through
conda create --name fg python=3.8
conda activate fg
sudo apt-get install -y libsuitesparse-dev libboost-all-dev libcholmod3
pip install -r requirements.txt
We also highly recommend, if possible, to use JAX with GPU/TPU-support. See the official Documentation for install instructions
In minimal_example.py we provide a small (non-) working example on how to use the factor graph as a stand-alone method to integrate in your workflow.
We provide a simple example on how to use the factor graph with pre-computed poses in the only_poses-directory. To execute the example run
python -m only_poses.main \
--experiment-root-path ./experiments/only_poses \
--experiment-name <my_experiment> \
--motion-type <TRANS or ROT> \ # Which motion type we generate examples for
--stddev-pos <Noise on pose position in meters> \
--stddev-ori <Noise on pose orientation in radians> \
--sample-length <Length of the sample/trajectory, for TRANS: in meters, for ROT in radians> \
--observation-amount <Observations per sample/trajectory> \
--number-samples <Number of samples created>
The script first checks if there are already pre-generated samples available (overwrite with --create-samples to always create new samples) and then runs the inference. Deactivate methods with
--no-use-sturm
--no-use-sturm-original
--no-use-fg
--no-use-fg-gt
An example could be
python -m only_poses.main \
--experiment-root-path ./experiments/only_poses \
--experiment-name testing \
--motion-type TRANS \
--stddev-pos 0.03 \
--stddev-ori 0.01 \
--sample-length 0.5 \
--observation-amount 10 \
--number-samples 10
For evaluation open the notebook evaluation.ipynb to analyze the results. In the first cell update
os.chdir("/root/of/the/repo/")
to point to the root of the repository. The notebook automatically gathers all experiments in the given root-path.
Auto-generate commands for sample settings used in the paper through running
python -m only_poses.run_all
and prints them to the console.
Available upon request.
For academic usage, the code is released under the GPLv3 license. For any commercial purpose, please contact the authors.
Toyota Research Institute (TRI) provided funds to assist the authors with their research but this article solely reflects the opinions and conclusions of its authors and not TRI or any other Toyota entity.
If you find our work useful, please consider citing our paper:
@inproceedings{DBLP:conf/iros/HeppertMYCB22,
author = {Nick Heppert and
Toki Migimatsu and
Brent Yi and
Claire Chen and
Jeannette Bohg},
title = {Category-Independent Articulated Object Tracking with Factor Graphs},
booktitle = {{IEEE/RSJ} International Conference on Intelligent Robots and Systems,
{IROS} 2022, Kyoto, Japan, October 23-27, 2022},
pages = {3800--3807},
publisher = {{IEEE}},
year = {2022},
doi = {10.1109/IROS47612.2022.9982029},
}