Learning to Control Self-Assembling Morphologies
Winner of Virtual Creatures Competition at GECCO 2019, Prague
NeurIPS 2019 (Spotlight)[Project Website] [Demo Video]
Deepak Pathak*, Chris Lu*, Trevor Darrell, Phillip Isola, Alexei A. Efros
University of California, Berkeley
MIT
(* equal contribution)
This is a pytorch based implementation for our paper on learning to control self-assembling agents using deep reinforcement learning. We investigate a modular co-evolution strategy: a collection of primitive agents learns to dynamically self-assemble into composite bodies while also learning to coordinate their behavior to control these bodies. We learn compositional policies to demonstrate better zero-shot generalization. If you find this work useful in your research, please cite:
@inproceedings{pathak19assemblies,
Author = {Pathak, Deepak and Lu, Chris and Darrell, Trevor and
Isola, Phillip and Efros, Alexei A.},
Title = {Learning to Control Self-Assembling Morphologies:
A Study of Generalization via Modularity},
Booktitle = {arXiv preprint arXiv:1902.05546},
Year = {2019}
}
Installation and Usage
- Setting up repository
git clone https://github.com/pathak22/modular-assemblies.git
cd modular-assemblies/
git clone https://github.com/Unity-Technologies/ml-agents.git
cd ml-agents/
git reset --hard 6c5255e
cd ..
bash envs/setup_env.sh
python3 -m venv assemblyEnv
source $PWD/assemblyEnv/bin/activate
pip install --upgrade pip
-
Installation
- Requirements:
- CUDNN-5.1, CUDA-8.0, Python-3.5
- Detailed setup, skip to quick setup for exact replication:
# Install Pytorch from http://pytorch.org/ pip install http://download.pytorch.org/whl/cu80/torch-0.3.0.post4-cp35-cp35m-linux_x86_64.whl pip install torchvision pip install --upgrade visdom # Install baselines for Atari preprocessing pip install gym==0.9.4 # baselines install latest gym first automatically, but latest gym has moved to mujoco5 so first install old gym and then install baselines git clone https://github.com/openai/baselines.git cd baselines git reset --hard b5be53d pip install -e . # Additional packages pip install numpy pip install matplotlib pip install pillow pip install opencv-python # fold cd modular-assemblies/src/ git clone https://github.com/nearai/pytorch-tools.git cd pytorch-tools/ git reset --hard 09dccb2 python setup.py install
- Quick setup for exact replication:
pip install -r requirements.txt
- Requirements:
-
Run code
cd modular-assemblies/src/
python test_env.py
Acknowledgement
Builds upon Ilya Kostrikov's Pytorch PPO implementation.