Awesome

Multi-contact linearized model predictive control

Source code for https://hal.archives-ouvertes.fr/hal-01349880/document

Installation

On Ubuntu 14.04, once you have installed OpenRAVE, do:

sudo apt-get install cython python python-dev python-pip python-scipy python-shapely
sudo pip install pycddlib quadprog pyclipper

Then, clone the repository and its submodule via:

git clone --recursive https://github.com/stephane-caron/3d-walking-lmpc.git

If you already have pymanoid installed on your system, be sure that its version matches that of the submodule.

Usage

The walk.sh script from the top-level directory opens the staircase simulation.

For a more detailed analysis, there are three subfolders in this repository, corresponding to different Sections of the paper:

• Section IV: cones/ contains scripts to display and play with the CoM acceleration cones
• Section VI: staircase/ walks the humanoid model around a circular staircase with tilted stepping stones
• Appendix: sep/ compares four algorithms calculating the static-equilibrium polygon

Due to the copyright problem, we cannot release the COLLADA model HRP4R.dae used to produce the accompanying video and paper illustrations. We have replaced it with JVRC-1, which has the same kinematic chain.