Markov Decision Process (MDP) Solver for Pacman

Overview

This repository contains an implementation of a Markov Decision Process (MDP) solver applied to the game of Pacman. The MDP solver employs the Value Iteration algorithm to compute optimal policies, enabling Pacman to navigate and make decisions in a dynamic and adversarial environment.

Introduction

This project focuses on the application of Value Iteration, a fundamental algorithm in Reinforcement Learning, to solve the Markov Decision Process inherent in the Pacman game. The primary objective is to enable Pacman to make intelligent decisions that maximize cumulative rewards while navigating through the maze and avoiding adversarial agents.

Requirements

The project has the following requirements:

• Python [2.7]
• NumPy [1.16.6]

Run MDP Pacman

Clone repository and examine various configurational options within the .txt file.

Acknowledgements

The development of this MDP solver is inspired by the referenced api.py module developed @ [UC Berkeley].