Machine learning and metaheuristic methods for optimizing dynamic truck dispatching in container ports

Abstract

Container port freight constitutes a pivotal component of contemporary global logistics, playing a decisive role in the convoluted matrix of international trade and supply chain management. Given the profound impact of container port operations on global trade dynamics, optimizing the transfer efficiency of container trucks within ports is of paramount significance. The efficiency of truck dispatching stands out as one of the core determinants determining port operational efficiency, thereby necessitating novel and advanced approaches to tackle the prevalent constraints and bottlenecks.

This paper explores machine learning (ML) methodologies to effectively navigate the complexities and uncertainties inherent in container port dynamic truck dispatching. Traditional methods in this domain frequently encounter limitations, particularly in their inability to adapt to real-time changes and handle the uncertainties characteristic of port operations. These conventional approaches, often reliant on static parameters, struggle to reflect port environments' constantly evolving and unpredictable nature accurately. In addressing these challenges, ML stands out as a highly suitable alternative, owing to its capacity to learn from vast datasets, adapt to novel scenarios, and make informed decisions under uncertain conditions.

The PhD project presented here is centered on two primary applications of ML: firstly, the development of innovative techniques for the generation and optimization of truck dispatching strategies, and secondly, the enhancement of operational efficiency and responsiveness amidst the fluctuating dynamics of port activities. By harnessing the adaptive and predictive capabilities of ML, this study aims to forge a more dynamic, responsive, and intelligent dispatching framework, one that is adept at overcoming the multifaceted uncertainties and fluidity inherent in contemporary container port operations.

Another facet of this PhD project seeks to augment the precision of evaluating truck dispatching strategies through refined ML methods. This segment is grounded in the aspiration to render the evaluation of diverse dispatching strategies more coherent and accurate, allowing for a meticulous assessment of their real-world efficacy and impact. The advancement in evaluation methodologies is anticipated to provide a nuanced understanding of individual dispatching strategies' intrinsic merits and demerits, contributing to developing more robust, effective, and tailor-fitted solutions.

Beyond these research works, this thesis delves into a comprehensive examination of the inherent challenges and opportunities residing within the interface of machine learning and truck dispatching. It elucidates the prospective advancements and innovations that machine learning can bring forth in optimizing dispatching mechanisms and strategies, emphasizing its potential to revolutionize container port logistics. This PhD research strives to discover new pathways for elevating operational efficiency and strategic intelligence in container port logistics through systematically synthesizing machine learning insights and domain-specific expertise.

The implications of this study are manifold, projecting substantial contributions to the enhancement of container port operational paradigms and global logistics frameworks. By incorporating machine learning into dynamic truck dispatching, this thesis aims to enhance efficiency and adaptability in the field, offering a more strategic approach to tackling the complexities and uncertainty involved. The findings and insights from this research are poised to provide valuable perspectives and pragmatic solutions for practitioners, policymakers, and academics, converging towards a more resilient, agile, and sustainable future in container port logistics and beyond.

Date of Award	Mar 2024
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Ruibin Bai (Supervisor) & Rong Qu (Supervisor)