Sequence-Based Simulation-Optimization Framework With Application to Port Operations at Multimodal Container Terminals

It is evident in previous works that operations research and mathematical algorithms can provide optimal or near-optimal solutions, whereas simulation models can aid in predicting and studying the behavior of systems over time and monitor performance under stochastic and uncertain circumstances. Given the intensive computational effort that simulation optimization methods impose, especially for large and complex systems like container terminals, a favorable approach is to reduce the search space to decrease the amount of computation. A maritime port can consist of multiple terminals with specific functionalities and specialized equipment. A container terminal is one of several facilities in a port that involves numerous resources and entities. It is also where containers are stored and transported, making the container terminal a complex system. Problems such as berth allocation, quay and yard crane scheduling and assignment, storage yard layout configuration, container re-handling, customs and security, and risk analysis become particularly challenging. Discrete-event simulation (DES) models are typically developed for complex and stochastic systems such as container terminals to study their behavior under different scenarios and circumstances. Simulation-optimization methods have emerged as an approach to find optimal values for input variables that maximize certain output metric(s) of the simulation. Various traditional and nontraditional approaches of simulation-optimization continue to be used to aid in decision making. In this dissertation, a novel framework for simulation-optimization is developed, implemented, and validated to study the influence of using a sequence (ordering) of decision variables (resource levels) for simulation-based optimization in resource allocation problems. This approach aims to reduce the computational effort of optimizing large simulations by breaking the simulation-optimization problem into stages. Since container terminals are complex stochastic systems consisting of different areas with detailed and critical functions that may affect the output, a platform that accurately simulates such a system can be of significant analytical benefit. To implement and validate the developed framework, a large-scale complex container terminal discrete-event simulation model was developed and validated based on a real system and then used as a testing platform for various hypothesized algorithms studied in this work