Graduation Year

2024

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Industrial and Management Systems Engineering

Major Professor

Hadi Gard, Ph.D.

Co-Major Professor

Changhyun Kwon, Ph.D.

Committee Member

Ankit Shah, Ph.D.

Committee Member

Yu Zhang, Ph.D.

Committee Member

Yi Qiang, Ph.D.

Keywords

Combinatorial Optimization, Dynamic Programming, Genetic Algorithm, Hybrid Methods, Vehicle Routing

Abstract

This dissertation explores novel approaches to address complex combinatorial optimization challenges in transportation and routing scenarios. Three sets of contributions are presented, each encapsulated in a chapter. The first set of contributions introduces a pioneering hybrid genetic algorithm meticulously crafted to address the intricacies of the Traveling Salesman Problem with Drone (TSPD) and the Flying Sidekick Traveling Salesman Problem (FSTSP). These emerging problems involve the strategic use of both ground-based trucks and aerial drones for efficient package delivery. Our algorithm stands out by leveraging sophisticated chromosomes and dynamic programming, allowing for broad exploration by the genetic algorithm and effective exploitation through dynamic programming and local searches. Notably, the genetic algorithm adeptly generates distinct sequences for trucks and drones, encoding them within type-aware chromosomes, whereas the dynamic programming determines the optimal launch and landing locations for drone. Local searches are applied to each chromosome, subsequently decoded by dynamic programming for fitness evaluation. The algorithm excels in surpassing existing methods, yielding superior solutions in both quality and computation time across a spectrum of benchmark instances.

The second set of contributions in this dissertation presents a hybrid genetic algorithm designed specifically for the Multiple Traveling Salesman Problem (mTSP) with an emphasis on minimizing the length of the longest tour. This problem variant poses unique challenges in coordinating multiple tours to achieve an optimal solution. Our algorithm employs TSP sequences as individual representations, incorporating a dynamic programming algorithm to evaluate individuals and determine the optimal mTSP solution for a given city sequence. A novel crossover operator enhances population diversity by combining similar tours from two parents, and a self-adaptive random local search, coupled with a thorough neighborhood search, further refines the generated offspring. Intriguingly, our algorithm excels in identifying and eliminating intersections between tours, particularly beneficial for min-max mTSP scenarios. Benchmark testing against existing methods showcases the algorithm’s superior performance, outpacing competitors on average and improving the best-known solutions for a substantial number of instances.

The third set of contributions proposes a versatile algorithmic framework applicable to a diverse array of arc routing problems with practical applications in mail delivery, snow plowing, and waste collection. Characterized by the common constraint that each edge must be visited no more than once, these problems present significant challenges. Our proposed framework systematically integrates genetic algorithms, dynamic programming, and local searches to offer a generic solution. By applying this framework to two specific problems, the min-max windy K-vehicle rural postman problem and the undirected capacitated arc routing problem with profits, we demonstrate its effectiveness and superiority over existing algorithms. The dissertation, through these three interconnected contributions, not only offers innovative solutions to distinct routing problems but also establishes a robust foundation for addressing a wide spectrum of challenges within the domain of transportation logistics.

Download

Share

COinS