KinodynamicFabrics.jl

Julia implementation of the Kinodynamic Fabrics Whole-Body Control framework.

Project webpage: adubredu.github.io/kinofabs
Paper: arXiv
Authors: Alphonsus Adu-Bredu*, Grant Gibson*, Jessy W Grizzle
*: Equal Contribution

Installation

1. Clone this repository using the command git clone -b sim https://github.com/adubredu/KinodynamcFabrics.jl
2. Navigate to the parent directory of this repo and type julia in your terminal to launch the Julia REPL.
3. Press ] on your keyboard to enter the package manager
4. Enter command activate . then enter command instantiate to download all package dependencies
5. Once installation is complete, press the Backspace key on your keyboard to return to the REPL

Usage

Checkout the examples folder for usage examples.

Dodge Ball

To run the dodge ball example, run the following command

include("examples/dodge_ball.jl")

You should see the visualization below:

Dodge Ball with End-Effector Goal

To run the dodge ball with end-effector goal example, run the following command

include("examples/dodge_ball_maintain_ee.jl")

You should see the visualization below:

Play Cornhole

To run the cornhole example, run the following command

include("examples/cornhole.jl")

You should see the visualization below:

Play Basketball

To run the basketball example, run the following command

include("examples/basketball.jl")

You should see the visualization below:

Deliver Package

To run the package delivery example, run the following command

include("examples/package_delivery.jl")

You should see the visualization below:

Note

The initial run of each example may have a delayed startup of the mujoco-python-viewer and other intermittent delays in-between simulation, due to the Julia pre-compilation process. If the interface asks, click Wait, and it may take up to 30 seconds to run. After the first run you should no longer have any delays.

Citing

@article{adubredugibson2023,
    title={Exploring Kinodynamic Fabrics for Reactive Whole-Body Control of Underactuated Humanoid robots},
    author={Adu-Bredu, Alphonsus and Gibson, Grant and Grizzle, Jessy},
    journal={},
    url={https://arxiv.org/abs/2303.04279},
    year={2023}
}

Branches

• The sim branch is the main branch of this repository, containing all the code to run the simulation demos above
• The experiments branch contains code to replicate simulation experiments in the paper
• The physical branch contains code for the physical robot demos