Bayesian approach for feasibility determination and spatiotemporal scheduling

Abstract

This thesis mainly consists of four parts. The first three parts explore Bayesian methods in solving the feasibility determination problem that commonly arises in the study of simulation and the last part considers spatiotemporal scheduling in manufacturing. More specifically, we propose a new reward function for Bayesian feasibility determination which emphasizes the importance of barely feasible/infeasible systems whose mean performance measures are close to the threshold. We utilize our proposed reward function in developing a Bayesian procedure and show its advantage in comparison with the benchmark procedures in the first part. Then, we present new two-stage and sequential Bayesian procedures that are not only easy to compute but also effective in solving the feasibility determination problem in the second part. In the third part, we focus on solving feasibility determination using a Gaussian process and propose our novel acquisition function. Finally, the last part considers a different topic, namely, spatiotemporal scheduling which often occurs in a manufacturing site where products are large and tend to be customized, such as ships, aircraft, and constructional structures. We propose how to generate a reasonably good time and spatial schedule on the manufacturing process. We propose a two-phase approach in solving this scheduling problem with an application to block scheduling in shipbuilding.