Influence-Augmented Online Planning for Complex Environments

This is the codebase accompanying the NeurIPS2020 paper "Influence-Augmented Online Planning for Complex Environments" by Jinke He, Miguel Suau and Frans A. Oliehoek.

Disclaimer: the codebase has only been tested on Ubuntu and Arch Linux.

Update: We fixed a few minor bugs found in the codebase and updated part of the results in the arxiv paper.

Dependencies

For online planning experiments:

• Singularity
• libtorch (C++ version of PyTorch): download and unzip libtorch into third-party/libtorch
• yaml-cpp: git clone https://github.com/jbeder/yaml-cpp third-party/yaml-cpp

For training the influence predictors and plotting the results:

• python3
• pytorch
• matplotlib
• jupyter notebook

Training the influence predictors

• instructions for training the influence predictors will be added soon.
• pretrained influence predictors are available under models/.

Online Planning Experiments

Singularity Container

We implemented our online planning experiments in C++ and provided a Singularity definition (singularity/FADMEN.def) file to resolve the dependencies.

To run the code, first build the singularity container with the command: sudo singularity build singularity/FADMEN.sif singularity/FADMEN.def.

To execute a command under the singularity container, use ./run + command.

Compile

./run bash scripts/build.sh

Reproducing results

General

./run ./scripts/run_benchmark + path to config file
for example, ./scripts/run_benchmark configs/GAC/5agent/global.yaml

Grab A Chair - num-agents: Figure 2(a) and 2(b)

./run ./scripts/run_benchmark configs/GAC/{X}agent/{Y}.yaml
for X in num_agents=[5,9,17,33,65,129] and
for Y in algos=["global", "inf_gru_H8_D1000", "inf_rand"]

Grab A Chair - coupling: Figure 3

./run ./scripts/run_benchmark configs/GAC-coupling/p{X}/{Y}.yaml
for X in [01,02,03,04,05,06,07,08,09] and
for Y in algos=["global", "inf_gru_H8_D1000SPLIT[0-3]", "inf_rand"]

Grid Traffic Control - real-time: 5(b), 5(c) and 5(d)

./run ./scripts/run_benchmark configs/GTC/{X}sec/{Y}.yaml
for X in seconds=[1,2,4,8,16,32,48,64] and
for Y in algos=["global", "inf_rnn_H2_D1000SPLIT[0-*]", "inf_rand"]

Plotting results

see notebooks/Plot-Results.ipynb

Acknowledgment

This project had received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 758824 — INFLUENCE).