Simulating prehensile pushing

This repository contains the MuJoCo simulation environment used for doing the prehensile pushing experiment in the paper Pushing Everything Everywhere All At Once: Probabilistic Prehensile Pushing .

T-shape-sim.mp4

Installation

conda create -f environment.yml
conda activate simulating_prehensile_pushing

Usage

Run

python simulate.py

If you want to generate new trajectories then please a look at the following code.

To see optional arguments run

python simulate.py -h

options:
  -h, --help                        Show this help message and exit
  --filename FILENAME               Path to the experiment file.
  --max_velocity MAX_VELOCITY       Maximum allowed joint velocity in radians/second.
  --no_control                      Run the simulation without any control.
  --damping DAMPING                 Damping term for the pseudoinverse. This is used to prevent joint velocities from becoming too large when the Jacobian is close to singular.
  --integration_dt INTEGRATION_DT   Integration time-step in seconds. This corresponds to the amount of time the joint velocities will be integrated for to obtain the desired joint positions.
  --no_gravity_compensation Disable gravity compensation.
  --dt DT                           Simulation time-step in seconds.

Acknowledgements

The controllers are borrowed from MuJoCo Controllers.

Citation

If you find this environment useful in your research, then please consider citing:

@ARTICLE{10930575,
  author={Perugini, Patrizio and Lundell, Jens and Friedl, Katharina and Kragic, Danica},
  journal={IEEE Robotics and Automation Letters}, 
  title={Pushing Everything Everywhere All At Once: Probabilistic Prehensile Pushing}, 
  year={2025},
  volume={},
  number={},
  pages={1-8},
  doi={10.1109/LRA.2025.3552267}}

License

This project is licensed under the MIT License. See LICENSE for more details.