RoboGrammar

[Paper] [Video]

A selection of best-performing designs generated with RoboGrammar for four different terrains.

Prerequisites

Commands were tested on Ubuntu 18.04.

CMake >= 3.8

Check with cmake --version

GLEW

sudo apt-get install libglew-dev

Python 3.6 or later + headers

Check the Python version with python3 —-version. If new enough, install Python headers: sudo apt-get install python3-dev
Otherwise, install the latest version of both: sudo apt-get install python3.9 python3.9-dev

Note: Newer versions of Python may be available through the "deadsnakes" PPA:

sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt-get update

Building (Linux, Mac OS)

git clone https://github.com/allanzhao/RoboGrammar.git

cd RoboGrammar

git submodule update --init

mkdir build; cd build

cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo ..

make -j8 (replace 8 with the number of CPU cores available)

Installing Python Packages

Make sure you are in the RoboGrammar directory.

pip3 install virtualenv

python3 -m venv venv

source venv/bin/activate

python3 examples/design_search/setup.py develop

python3 build/examples/python_bindings/setup.py develop

pip3 install numpy-quaternion
pip3 install torch==1.7.1+cpu -f https://download.pytorch.org/whl/torch_stable.html
pip3 install torch-scatter -f https://pytorch-geometric.com/whl/torch-1.7.0+cpu.html
pip3 install torch-sparse -f https://pytorch-geometric.com/whl/torch-1.7.0+cpu.html
pip3 install torch-geometric==1.4.3

Running Examples

Make sure you are in the RoboGrammar directory, and that the virtualenv is active:

source venv/bin/activate

Run MPC for selected designs, and visualize (change -j8 to use more CPU cores):

python3 examples/design_search/viewer.py RidgedTerrainTask data/designs/grammar_apr30.dot -j8 0, 7, 1, 13, 1, 2, 16, 12, 13, 6, 4, 19, 4, 17, 5, 3, 2, 16, 4, 5, 18, 9, 8, 9, 9, 8 -o

python3 examples/design_search/viewer.py FlatTerrainTask data/designs/grammar_apr30.dot -j8 0, 12, 7, 1, 12, 3, 10, 1, 3, 1, 12, 12, 1, 3, 10, 2, 16, 8, 1, 3, 12, 4, 1, 3, 2, 12, 18, 9, 18, 8, 5, 5, 1, 12, 6, 3 -o

python3 examples/design_search/viewer.py GapTerrainTask data/designs/grammar_apr30.dot -j8 0, 1, 1, 7, 1, 6, 10, 3, 2, 4, 10, 10, 3, 16, 4, 16, 18, 2, 5, 16, 8, 4, 8, 8, 18, 4, 5, 15, 9, 8, 8 -o

python3 examples/design_search/viewer.py FrozenLakeTask data/designs/grammar_apr30.dot -j8 0, 1, 1, 1, 6, 7, 10, 11, 13, 2, 4, 3, 4, 16, 8, 14, 4, 8, 3, 15, 15, 5, 3, 9, 8 -o

Run Graph Heuristic Search to find optimal designs (for flat terrain):

python3 examples/graph_learning/heuristic_search_algo_mpc.py --task FlatTerrainTask --grammar-file data/designs/grammar_apr30.dot --no-noise

Run the MCTS and random search baselines (5000 iterations, flat terrain):

python3 examples/design_search/design_search.py -a mcts -j8 -i5000 -d40 --log_dir logs_mcts FlatTerrainTask data/designs/grammar_apr30.dot

python3 examples/design_search/design_search.py -a random -j8 -i5000 -d40 --log_dir logs_random FlatTerrainTask data/designs/grammar_apr30.dot

The search algorithms output .csv log files containing each design/rule sequence tried and its reward.

FAQs

I get the error The RandR headers were not found

Install the X server development files: sudo apt-get install xorg-dev

I get the error RuntimeError: Could not open file "data/shaders/default.vert.glsl" when trying to run examples

Set the ROBOT_DESIGN_DATA_DIR environment variable: export ROBOT_DESIGN_DATA_DIR=$PWD/data/