Skip to content

Latest commit

 

History

History
84 lines (62 loc) · 2.3 KB

README.md

File metadata and controls

84 lines (62 loc) · 2.3 KB

lgp

This repo implements Dynamic Logic-Geometric Programming for 2D agents on PyBullet.

Installation

This assumes you a;ready install the dependencies for Simon's master thesis repo and humoro.

Clone Simon's master thesis repo:

git clone [email protected]:simon.hagenmayer/hierarchical-hmp.git

Then, clone humoro and lgp to hierarchical-hmp folder, checkout MASimon branch on humoro and install dependencies of lgp:

cd hierarchical-hmp
git clone [email protected]:philippkratzer/humoro.git
git clone https://github.com/humans-to-robots-motion/lgp
cd humoro
git checkout MASimon
cd ../lgp
pip install -r requirements.txt

Also clone bewego into hierarchical-hmp folder:

cd hierarchical-hmp
git clone https://github.com/anindex/bewego --recursive
cd bewego
mkdir -p build && cd build
cmake .. -DCMAKE_BUILD_TYPE=RelWithDebInfo -DWITH_IPOPT=True -DPYBIND11_PYTHON_VERSION=3.5
make
make install

Finally, please download MoGaze dataset and unzip it into lgp/datasets/mogaze.

mkdir -p datasets && cd datasets
wget https://ipvs.informatik.uni-stuttgart.de/mlr/philipp/mogaze/mogaze.zip
unzip mogaze.zip

And also run this script to initialize Pepper URDF:

cd lgp
python3 examples/init_pepper.py

Usage

To see an example of Dynamic LGP, please run:

python3 examples/test_lgp.py -d True -v True -p False

where -d is the option to run in dynamic mode or not, -v True is to enable trajectory optimization visualization and -p is the option to enable human prediction.

To run Dynamic LGP with human prediction, please use the following segment ('p2_1', 137536, 139256):

python3 examples/test_lgp.py --segment "('p2_1', 137536, 139256)" -d True -v True -p True

To reproduce the paper results, please run the following experiment scripts.

  • Dynamic LGP with Human Ground Truth:
python3 examples/experiment_ground_truth.py
  • Dynamic LGP with Long-term prediction:
python3 examples/experiment_prediction.py

The experiment data will be saved as pickle file in folder lgp/data/experiments, rename it to a more shorter name. To visulize data result, please run:

python3 examples/process_data.py --name <your data>.p