NeurADP for Ride-pooling

Introduction

This is the unofficial implemented code in PyTorch for the paper "Neural Approximate Dynamic Programming for On-Demand Ride-Pooling" that appears in the Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence.

Requirements

We recommend using python<=3.7 and a conda venv.

conda create --name NeurADP python=3.7
cd path/to/NeurADP
conda activate NeurADP

Install the dependencies with:

pip install -r requirements.txt

Pay attention to the installation of docplex! If you only want to use the free version of docplex, simply:

pip install docplex

If you want to use the full version of docplex, you need to install it from source. (or contact the developer)

git clone https://github.com/IBMDecisionOptimization/docplex-docplex.git
cd docplex-docplex
python setup.py install

When you're done working on the project, deactivate the conda virtual environment with deactivate.

Data

Here is the structure of the data folder:

data/
    files_60sec/
        test_flow_5000_1.txt
        test_flow_5000_2.txt
    ignorezonelist.txt
    taxi_3000_final.txt
    zone_path.csv
    zone_traveltime.csv

Usage

To run the code, simply:

python main.py

The code will automatically process the data in the data folder.