WardropNet: Traffic Flow Predictions via Equilibrium-Augmented Learning

This repository comprises the code to learn a Combinatorial Optimization Augmented Machine Learning (COAML) pipeline that predicts traffic flow from contextual information.

This method is proposed in:

Kai Jungel, Dario Paccagnan, Axel Parmentier, and Maximilian Schiffer. WardropNet: Traffic Flow Predictions via Equilibrium-Augmented Learning. arXiv preprint: arxiv:2410.06656, 2024.

This repository contains all relevant scripts and data sets to reproduce the results from the paper.
To run the code for reproducing the results we assume using slurm, and require a Gurobi license.
We used Gurobi version 10.0.0.
We used Python version 3.8.10.
We run the code on a Linux system.

The structure of this repository is as follows:
./matsim-berlin: Multi Agent Transport Simulation to simulate oracle scenarios
./runnables: comprises all scripts to execute the code
./surrogate: code of the Combinatorial Optimization Augmented Machine Learning (COAML) pipeline
./toySimulation: code to construct toy simulations that create toy scenarios

Each of these folders comprises respective README.md files.

To reproduce the results from the paper follow these steps:

1. Install MATSim

In the matsim-berlin directory, install MATSim: https://github.com/matsim-scenarios/matsim-berlin

2. Install python dependencies

We recommend to install the packages in a virtual environment called venv.

pip install -r requirements.txt

3. Create scenarios

1. Install the random road network creator in ./surrogate/generating: https://github.com/arun1729/road-network
2. To create a scenario run

bash runnables/scenario_name.sh

4. Train and evaluate the COAML pipeline

1. Recommendation: Run the training on a cluster using slurm.
2. We provide pre-caculated instances in surrogate/data. Please unpack them, if you did not run step 3. Create scenarios
3. On the cluster, define the parameters in the runnables/cluster/master.sh file.
4. On the cluster, in runnables/cluster/master.sh uncomment the training or evaluation, depending on what you want to run.
5. On the cluster, run

bash runnables/cluster/master.sh

5. Visualization of results

1. Unpack ./surrogate/results_paper.zip and ./surrogate/data/read_in_best_learning_iteration.zip
2. Move to ./surrogate/visualization and run:
   • learning_results.py to compare the benchmark models
   • input_data.py to visualize the scenarios
   • output_data.py to visualize the traffic flow

Please note that the benchmark names in the paper were updated: multicommodityflow / MCFP -> CL, wardropequilibria / WE -> PL, wardropequilibriaRegularized / WE-reg -> ER