R:SS 2018 paper on Differentiable Physics and Stable Modes for Tool-Use and Manipulation Planning
Contact me if you're interested using Logic-Geometric Programming yourself!
I'm currently developing nicer (user oriented) interfaces to the code, including python bindings. Please register with the "LGP code" mailing list at https://groups.google.com/forum/#!forum/lgp-code or contact me at email@example.com if you're interested being involved, e.g. in testing interfaces and providing feedback how to improve them, or even contributing yourself.
This repository is a supplement to the paper
M. Toussaint, K. R. Allen, K. A. Smith, and J. B. Tenenbaum: Differentiable Physics and Stable Modes for Tool-Use and Manipulation Planning. In Proceedings of Robotics: Science and Systems (R:SS), 2018.
Please also see the supplementary video to get an impression on the method.
The repository checks out the original commit 18-RSS tag (as submodule) and allows you to reproduce the experiments. Please see the notes at the end on ongoing developments of the core RAI code, including simpler interfaces for others to use these methods.
On Ubuntu 16.04, and assuming you have ssh authorization with github:
git clone firstname.lastname@example.org:MarcToussaint/18-RSS-PhysicalManipulation.git cd 18-RSS-PhysicalManipulation git submodule init git submodule update make -j1 initUbuntuPackages # calls sudo apt-get install; you can always interrupt make # builds libs cd demo; make; ./x.exe # run the demo
If you don't have ssh authorization, try
git clone https://github.com/MarcToussaint/18-RSS-PhysicalManipulation.git cd 18-RSS-PhysicalManipulation git config --file=.gitmodules submodule.rai.url https://github.com/MarcToussaint/rai.git git config --file=.gitmodules submodule.rai-robotModels.url https://github.com/MarcToussaint/rai-robotModels.git git submodule init git submodule update make -j1 initUbuntuPackages # calls sudo apt-get install; you can always interrupt make # builds libs cd demo; make; ./x.exe # run the demo
If you need to run in a docker, please have a look at the docker setup I use to test RAI. A slight modification of the Dockerfile and run scripts should allow you to setup the appropriate environment also for this repo, including X.
This repo is meant as a snapshot that documents the RSS paper's experiments. Since then, the RAI code and LGP method continuously developed, both to provide cleaner interfaces to collaborators, and in terms of extending the method. Development is mostly going on driven by the python bindings - see https://github.com/MarcToussaint/rai-python .