MORL4Water: Toolkit for testing and building water resource management simulations from scratch

The goal of this project is to create a suite of water mamagement-related Mo-GYM environments for training reinforcement learning agents with multipe objectives. These environments simulate problems related to managing reservoirs around the world. Currently, the number of environments for MORL available is limited and represents mostly toy-examples. With the suite of water management-related environments, it will be possible to test different MORL algorithms on real-world examples with high impact. The modular framework will also allow researchers from water management to build their simulations in an easier, unified manner.

This work required knowledge of water systems and reinforcement learning and was done 100% in Python.

You can use MORL4Water and access the documentation here

This works has been accepted at AAMAS 2025.