Fire Containment: Optimization & AI

These projects introduce the novel Moving Firefighter Problem (MFP), offering solutions through Mathematical Optimization and comparing their effectiveness with advanced Deep Reinforcement Learning (DRL) algorithms. The portfolio showcases the use of cutting-edge techniques to address real-time firefighting challenges in dynamic environments.

To explore the MFP in detail, including comprehensive insights, additional solutions, and extensions, click 'Learn More' to view my PhD thesis, The Moving Firefighter Problem.

My Work

The Moving Firefighter Problem

The Moving Firefighter Problem

This repository introduces "The Moving Firefighter Problem," a novel approach in graph theory to model fire spread and containment. Using a mixed-integer quadratically constrained program (MIQCP), it finds optimal defense strategies. This work is part of a published paper, which you can access here.

View Repository
Firefighter-DRL

Firefighter-DRL

This repository applies Deep Reinforcement Learning (DRL) to simulate firefighting strategies in a dynamic environment. By optimizing containment strategies with various reward functions, it showcases the power of DRL to adapt to real-world scenarios. The environment used here, known as "Bulldozer," models the spread of fire and the effectiveness of different firefighting approaches.

View Repository
Instance Translator

Instance Translator

This project builds a bridge between mathematical optimization and simulation, translating instances of the Moving Firefighter Problem (MFP) into the Bulldozer environment. It enables a direct comparison between two powerful techniques: Deep Reinforcement Learning (DRL) and Mixed-Integer Quadratically Constrained Programming (MIQCP), highlighting their strengths in addressing complex firefighting challenges.

View Repository
Gym Cellular Automata

Gym Cellular Automata

This repository presents the Bulldozer environment, a benchmark designed to simulate real-world firefighting scenarios. By leveraging cellular automata and Reinforcement Learning (RL) algorithms, it serves as a testing ground for developing and evaluating strategies to contain and control the spread of fires in dynamic situations.

View Repository

Get In Touch

If you'd like to learn more about my work or discuss opportunities, feel free to reach out to me.