This repository contains code and data for the CCWI paper named in the heading. In order to run the code, first install the requirements (src/REQUIREMENTS.txt). You may then start RL experiments by running src/cw_main.py with a config file, e.g.
python cw_main.py ../Configs/ccwi26/ccwi26.yaml -e ccwi26_ctown_pda_trials
Due to the nature of the learning process and the difficulty to determine the lowest possible feasible pump speed (see paper), the experiments may generate a lot of EPANET warnings. In this context, these warnings are rather a sign of the agent trying out different strategies than a reason for concern about the algorithms performance. If you want to reduce the amount of warnings, you may use the shell script instead of running python directly:
./cw_main.sh ../Configs/ccwi26/ccwi26.yaml -e ccwi26_ctown_pda_trials
Please refer to the documentation of ClusterWorks2 on how to run experiments in the config files.
If you have questions, don't hesitate to contact us