This repo hosts the source code for the paper Efficient Object Manipulation Planning with Monte Carlo Tree Search
- Pinocchio for rigid body dynamics
- Bullet Utils for interfacing PyBullet with Pinocchio
- Robot Properties NYU Finger for NYU finger URDFs and configuration files
- Other standard dependencies can be found in
requirements.txt
.
We prepared a few notebooks
/demos
mcts[miqp]_playground.ipynb
shows step-by-step how to generate a contact plan for a given motion with force plots and plan visualization.mcts[miqp]_bullet_simulation.ipynb
simulates the contact plan with an impedance controller in PyBullet./experiments
contains notebooks to reproduce the respective experiments in the paper./train
contains notebooks to generate the training data and to train the neural networks.
@article{zhu2022efficient,
title={Efficient Object Manipulation Planning with Monte Carlo Tree Search},
author={Zhu, Huaijiang and Meduri, Avadesh and Righetti, Ludovic},
journal={arXiv preprint arXiv:2206.09023},
year={2022}
}
- Huaijiang Zhu
Copyright(c) 2023 New York University
BSD 3-Clause License