Preferential access and fairness in waste management

This project is based on the theory and application of quadratic optimization models within the context of undesirable facility location problems. All data and code for the associated papers are provided in this repository.

For the theoretical side, see and cite the published article Quadratic optimization models for balancing preferential access and fairness: Formulations and optimality conditions by C. Schmitt and B. Singh in the INFORMS Journal on Computing. Here, we propose a Mixed Integer Quadratic Programming (MIQP) Model for a Facility Location Problem that seeks to assign users to facilities in a manner that balances access for users and fairness among facilities. We refer to the article for further information on methods and assumptions.

For the motivation of this project, see and cite the published article Selectively closing recycling centers in Bavaria: Reforming policy and reducing disparities by M. Schmidt and B. Singh in Networks. Here, we seek efficient ways to close recycling centers in Bavaria to meet climate neutrality goals, while still not overburdening the remaining centers and ensuring residents have good access to recycling.

Repository content

The repository contains the following content:

• catchment_population presents an efficient algorithm to compute the "catchment population" of each recycling center that was used to estimate the capacities. An implementation of this algorithm is contained in catchment_population.py. Further, this directory contains two corresponding input data files in csv format: bavaria_grid_population.csv is a file containing the latitude and longitude of the centroid of each 100m x 100m grid in Bavaria as well as the residing population. rc_locations.csv contains the latitude and longitude of each recycling center in Bavaria.
• data contains the two input data files that are used in the MIQP model: users_and_facilities.xlsx contains all ZIP codes and recycling centers related data like the population, centroid and regional spatial type (rural/urban) of each ZIP code as well as the capacity, centroid and regional spatial type of each recycling center. travel_dict.json.pbz2 is a compressed json file that contains the travel probabilities from each ZIP code to each recycling center.
• model contains scripts for the optimization of the MIQP model and visualization of the results: model.py contains functions for building and optimizing the MIQP model, while greedy_heuristic.py applies a greedy heuristic to achieve a feasible solution. plotting.py and results.py contain functions tasked with visualizing results through various different plots as well as excel tables. They also contain superordinate functions that create the corresponding results first by running functions from model.py and/or greedy_heuristic.py before creating the corresponding visualization. These functions are called by functions in tables_and_figures.py to create the exact same figures and tables that are included in the paper from scratch. Lastly, utils.py contains helper and utility functions.
• subsequent_workcontains a copy of the files in modelthat have been further expanded on for the application of this topic.
• results contains the excel tables and figures visualizing the results that are included in the paper.
• appendix.pdf that supplments the main text. This appendix includes proofs, sources of data, and figures and tables.

Requirements to run code

The code uses some open-source Python packages. The ones that the reader may be most unfamiliar with are:

• Pyomo, a Python-based optimization modeling language that allows building optimization models.
• Gurobi, a software well-equiped for solving complex optimization models such as MIQPs.
• Geopy, which was used for calculating geodesic distances (i.e. shortest distances on the surface of the earth) between two locations.