Reconfigurable Robot Control Using Flexible Coupling Mechanisms

Sha Yi, Katia Sycara, and Zeynep Temel

If you find our work helpful, please consider citing our paper:

@inproceedings{yi2023reconfiguration,
  title={Reconfigurable Robot Control Using Flexible Coupling Mechanisms},
  author={Yi, Sha and Sycara, Katia and Temel, Zeynep},
  booktitle={Robotics science and systems},
  year={2023}
}

Installation

1. Create conda environment with

conda env create -f puzzle_env.yml

This will create a environment named puzzle. You will need to activate it with

conda activate puzzle

2. Install three other dependencies with pip in the puzzle environment.

pip install casadi
pip install polytope
pip install pybullet

3. The package is originally in the ROS catkin source workspace due to my hardware interface. But the simulation is independent of ROS. However, if you do not have ROS installed, you might need to install another dependency:

pip install catkin-pkg

Then you may install this package with

pip install -e .

Generate URDF file

There are multiple PuzzleBot URDFs that can be generated. Currently I'm using the simplified one. Run the following in the package root directory.

python scripts/generate_urdf.py 1

This should generate puzzlebot.urdf in the urdf/ directory.

Run Simulation

Run the following for the simulation.

python bin/run_sim.py

Let me know if you encounter any errors.

Run on Hardware

Since ROS 1 is based on python 2 while our package uses python 3, the user is suggested to resolve this conflict on their computer based on their local environment. This link may be helpful.

After resolving the python conflict, you may first move the package to your catkin workspace source directory, e.g. catkin_ws/src. Then install this package with

catkin build
source devel/setup.bash

You can then launch the package with

roslaunch puzzlebot_assembly run_multi.py N:=$NUMBER_OF_ROBOTS

Disclaimer

This code is only tested fully on Ubuntu 18.04 with python version 3.7. It is partially tested on Macbook but the simulation and optimization parameters may need additional tuning.