Skip to content

mit-drl/pyrmm

Repository files navigation

PYRMM: (Learned) Risk Metric Maps in Python

Overview

Dubins Example in 2D

The pyrmm library is used for development of neural network models that predict risk (e.g. probability of failure) for kinodynamic systems (e.g. cars, airplanes, etc.)

Citation

TODO

Reproduction of Paper Results

Please see instructions in the experiments sub-package

Disclaimer

Research was sponsored by the United States Air Force Research Laboratory and the United States Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative Agreement Number FA8750-19-2-1000. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the United States ​Air Force or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.

Installation

Due to several 3rd party non-Python dependencies, the install process is somewhat involved and manual.

The installation process is written for an Ubuntu Linux machine.

It is strongly recommended to use a conda environment to manage the installation. These instructions only cover a conda environment installation process.

Clone and Conda Env

git clone --recurse-submodules git@github.com:mit-drl/pyrmm.git
cd pyrmm

# creates a conda env called pyrmm, this may take a while
conda env create -f environment.yml 
conda activate pyrmm

# build the mazegenerator submodule used for procedurally
# generating 2D maze images
cd mazegenerator/src
make

# build the V-HACD library used to create convex decompositions of procedurally generated 
# rooms
cd ../v-hacd/app/
cmake -DCMAKE_BUILD_TYPE=Release CMakeLists.txt
cmake --build .

Install OMPL with Python Bindings

Next we will download and install OMPL following the OMPL install instructions.

We use the Open Motion Planning Library (OMPL) for it's state and control space definitions and sampling functions. OMPL is a C++ library but provides python bindings that must be built by running an installation script. These instructions were written when ompl 1.5 was the most recent release, but now ompl 1.6. The install process should be the same for ompl 1.6, but it hasn't been tested.

# Download OS-specific script from https://ompl.kavrakilab.org/core/installation.html
# For ubuntu, the next steps look like
chmod u+x install-ompl-ubuntu.sh
./install-ompl-ubuntu.sh --python # this will install the latest release of OMPL with Python bindings

⚠️ Troubleshooting The python bindings can take hours to build. It is recommended to use PyPy3 to speed up the process, but these instructions don't go into the level of detail needed to walkthrough PyPy3 vs CPython

The above steps should install ompl to your home directory, e.g. /home/user/ompl-1.5.2/. The python bindings are not immediately "discoverable" by python; we need to point the conda environment to the ompl install using a .pth file. Assuming your conda environment is stored in ~/miniconda3/envs/pyrmm, you run something similar to (adjusting paths as need to match install locations and username on your machine)

touch ~/miniconda3/envs/pyrmm/lib/python3.8/site-packages/ompl.pth
echo "/home/user/ompl-1.5.2/py-bindings" >> ~/miniconda3/envs/pyrmm/lib/python3.8/site-packages/ompl.pth

Install Blender

We use blender in order to convert .stl meshes produced by pcg-gazebo into .obj files that can be loaded by pybullet. These instructions were written with blender 3.2.1, but the pyrmm library should be compatible with later version of blender

Download and install blender following instructions here: https://www.blender.org/download/

To make blender more easily callable without the entire path to the install, create the following script (assuming blender 3.2.1 was installed to home directory) and save to /usr/local/bin/blender

#!/bin/sh
~/blender/blender-3.2.1-linux-x64/blender "$@"

Then make that script executable by running

sudo chmod a+x /usr/local/bin/blender

Verify Installation

TODO

Example Usage: Dubins Vehicle in PPM Mazes

Procedural Generation of Obstacle Space

TODO

Approximate Risk Metric Data Generation

TODO

Risk Metric Model Training

TODO

Example Usage: Parallel Autonomy for Dubins Vehicle

TODO

About

Risk Metric Maps in Python

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages