Moma

.github/workflows/pytest.yml

@@ -2,34 +2,34 @@ name: JUSTICE Unit Tests
 
 on:
   push:
-    branches:
-      - main
+    branches: [main]
   pull_request:
-    branches:
-      - main
+    branches: [main]
 
 jobs:
-  build:
+  test:
     runs-on: ubuntu-latest
 
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v4
         with:
-          lfs: true # Enable Git LFS support
+          lfs: true
 
-      - name: Set up Python 3.9
-        uses: actions/setup-python@v2
+      - name: Set up Python
+        uses: actions/setup-python@v5
         with:
-          python-version: 3.9
+          python-version: "3.10"
 
       - name: Install Poetry
         run: |
-          curl -sSL https://install.python-poetry.org | python3 -
-          echo 'export PATH="$HOME/.local/bin:$PATH"' >> $GITHUB_ENV
+          python -m pip install --upgrade pip
+          python -m pip install "poetry==2.3.2"
+          poetry --version
 
       - name: Install dependencies
         run: |
-          poetry install
+          poetry config virtualenvs.in-project true
+          poetry install --no-interaction --no-ansi
 
       - name: Run Pytest
         env:

.gitignore

@@ -232,8 +232,13 @@ limitarian_analysis.py
 paper_data_figures_generator.ipynb
 paper_data_figures_generator_codeocean.ipynb
 reproducing_data_and_plots.py
-figs/*
 ml_importance_plots/*
+figs/*
+solvers/moea/*.so
+solvers/moea/*.dylib
+solvers/moea/borgMOEA.py
+third_paper_visualizations.ipynb
+fourth_paper_analysis.ipynb
 # tests/verification_data/*
 #Mac
 .DS_Store

CITATION.cff

@@ -0,0 +1,28 @@
+# This CITATION.cff file was generated with cffinit.
+# Visit https://bit.ly/cffinit to generate yours today!
+
+cff-version: 1.2.0
+title: JUSTICE Integrated Assessment Modelling Framework
+message: >-
+  If you use this software, please cite it using the
+  metadata from this file.
+type: software
+authors:
+  - given-names: Palok
+    family-names: Biswas
+    email: p.biswas.nl@gmail.com
+    affiliation: Delft University of Technology
+    orcid: 'https://orcid.org/0000-0003-3830-7203'
+identifiers:
+  - type: doi
+    value: 10.5281/zenodo.15145121
+    description: Zenodo
+repository-code: 'https://github.com/pollockDeVis/JUSTICE'
+keywords:
+  - integrated assessment model
+  - climate-economy
+  - deep uncertainty
+  - normative uncertainty
+license: BSD-3-Clause
+version: 0.2.0
+date-released: '2025-06-23'

README.md

@@ -2,6 +2,7 @@
 
 ![GitHub Actions build status](https://github.com/pollockDeVis/JUSTICE/actions/workflows/pytest.yml/badge.svg?event=push)
 [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.15145122.svg)](https://doi.org/10.5281/zenodo.15145122)
+
 <p align="center">
   <img title="JUSTICE Logo" alt="JUSTICE Logo" src="/docs/diagrams/JUSTICE LOGO.svg" width="400"/>
 </p>
@@ -12,7 +13,9 @@ JUSTICE is designed to explore the influence on distributive justice outcomes du
 
 ### JUSTICE Overview
 
-<img title="JUSTICE Framework" alt="Flowchart of JUSTICE" src="/docs/diagrams/JUSTICE Flowchart.jpeg">
+<p align="center">
+  <img title="JUSTICE Framework" alt="JUSTICE Framework" src="docs/diagrams/JUSTICE.svg" width="600"/>
+</p>
 
 ### Documentation
 
@@ -25,34 +28,23 @@ JUSTICE is developed by the [HIPPO Lab](https://www.tudelft.nl/ai/hippo-lab) at
   <img src="docs/diagrams/tudelft_logo_dark.png" alt="TU Delft Logo" width="200" style="margin-left: 20px;"/>
 </p>
 
+# Citation
 
-### Installation
-Create and activate a virtual environment and install as a package using
-
-```
-pip install .
-```
-
-### Running JUSTICE MOMARL
-
-#### Training
-
-JUSTICE-MOMARL can train multi-agent and multi-objective policies and currently support optimising for two objectives. JUSTICE-MOMARL converts the multi-objective training into a single objective by generating various policies with different weight combinations that weight each objective. The entire training process generates **100 different weight combinations** and trains each policy by linearising and normalising the reward and using **Multi Agent Proximal Policy Optimization (MAPPO)** as the reinforcement learning algorithm.
+To cite this code, please use the information in [CITATION.cff](CITATION.cff) and the following bibtex entry:
 
-Instead of training all the 100 weights together, we divide the training into batches and train a subset of weight combinations in each batch. For example, in order to train the first 10 uniform weight combnations, balancing the global economic output and inverse global temperature reward functions, run the following:
-
-```bash
-python train.py --start_uniform_weight 0 --end_uniform_weight 10 --env-config.rewards "global_economic_output" "inverse_global_temperature" --seed=<Seed> --base_save_path="Path where results are saved"
 ```
-
-You can view the additional argument to modify the JUSTICE modules and the RL hyperparameters by looking at the `Args` and `EnvConfig` dataclasses in `rl/args.py`.
-
-This script will save the checkpoints, pickled configs and multi-objective vectorised returns for trained policies for each weight combination in the base_save_path provided.
-
-#### Evaluation
-Once models have been trained for different weight combinations and the checkpoints for each of them are saved, you can run the evaluation for one specific weight combination and for one evaluation seed by running:
-
-```bash
-python eval.py --checkpoint_path="Absolute path to the checkpoint for policy" --seed=0 --output_path="Directory where evaluation results are saved""
+@inproceedings{ijcai2025p1064,
+title = {Exploring Equity of Climate Policies Using Multi-Agent Multi-Objective Reinforcement Learning},
+author = {Biswas, Palok and Osika, Zuzanna and Tamassia, Isidoro and Whorra, Adit and Zatarain-Salazar, Jazmin and Kwakkel, Jan and Oliehoek, Frans A. and Murukannaiah, Pradeep K.},
+booktitle = {Proceedings of the Thirty-Fourth International Joint Conference on
+Artificial Intelligence, {IJCAI-25}},
+publisher = {International Joint Conferences on Artificial Intelligence Organization},
+editor = {James Kwok},
+pages = {9573--9581},
+year = {2025},
+month = {8},
+note = {AI and Social Good},
+doi = {10.24963/ijcai.2025/1064},
+url = {https://doi.org/10.24963/ijcai.2025/1064},
+}
 ```
-The evaluation script will run the MOMARL policy in the JUSTICE simulation and save the evolution of various economic and climate metrics over time where the Emission Control Rate and Savings Rates are set by the trained policy in each time step.