ECOLIGHT
--------

We used OpenSource CityFlow as the simulator for our experiments:
https://github.com/cityflow-project/CityFlow

For the Thresholding EcoLight experiments, we can embed our algorithm into below OpenSource project:
https://github.com/wingsweihua/colight

In the root folder of colight, we can add attached EcoLight_agent.py, and add its config at config.py as given below -
DIC_ECOLIGHT_AGENT_CONF = {
    "ALPHA": 0.17,
    "TIME_CYCLE": 150,
    "MIN_GREEN": 5,
    "MAX_DENSITY": 40,
    "MODE": 0
}
where MODE signifies {Random:0, Timed(1dim):1, Timed(2dim):2}

For any doubts, you can check the usage of any baseline algorithm under the baseline folder, and include for our algorithm accordingly.
OR
As a hack, replace the code and config of any other baseline algorithm, and run that algorithm.

Run the code as below
python run_baseline.py --model=EcoLight

NewYork datasets are present under 
https://github.com/wingsweihua/colight/tree/master/data/NewYork/16_3
https://github.com/wingsweihua/presslight/tree/master/data/template_lsr/NewYork
As conveyed in the paper, developing region dataset will be provided on paper acceptance.

Metrics can be gathered using the methods described in the open source project. We have also developed our own ways to gather metrics by modifying the anon_env.py file. Though we plan to share the whole work project on paper acceptance, still if you require it during review, kindly let us know.

