Date of Award
2023
Degree Type
Thesis
Degree Name
Master of Science in Systems Engineering
Department
Mechanical, Industrial and Systems Engineering
First Advisor
Manbir Sodhi
Abstract
In the realm of optimization and operations research, addressing complex combinatorial problems efficiently and effectively has always been a challenge. The Capacitated Vehicle Routing Problem (CVRP), a classic and well-known optimization problem, exemplifies this challenge. CVRP involves finding optimal routes for a fleet of vehicles to serve a set of customers while respecting vehicle capacity constraints and minimizing the total distance traveled.
Over the years, researchers and practitioners have employed a multitude of techniques to tackle the CVRP, ranging from traditional optimization algorithms to modern computational methods. In recent times, the convergence of machine learning and genetic algorithms has paved the way for innovative and powerful solutions to previously intractable problems.
In the first study, the work assesses the predictive capabilities of different machine learning models to identify the optimal algorithm for various problem domains. Performance metrics, such as convergence rate, accuracy, and computational efficiency, are analyzed to determine the most effective machine learning model for algorithm selection.
The second paper complements the first by exploring a novel approach that integrates reinforcement learning into the optimization process. A Q-learning reinforcement learning algorithm is employed to fine-tune the hyperparameters of the genetic algorithm. This hybrid approach aims to improve both performance and diversity in the genetic algorithm’s search for solutions.
Recommended Citation
Quevedo, Jose D., "IMPROVING THE PERFORMANCE OF GENETIC ALGORITHMS USING REINFORCEMENT LEARNING." (2023). Open Access Master's Theses. Paper 2357.
https://digitalcommons.uri.edu/theses/2357
Terms of Use
All rights reserved under copyright.