Ph.D. Thesis Defense

Public Presentation
Thursday May 5, 2022

10:00 AM

Location: Hanes 125

Wei Liu

  Data-driven aircraft assignment and stochastic models for service systems

Under the direction of Vidyadhar Kulkarni and Vinayak Deshpande

This dissertation consists of two parts: data-driven aircraft assignment and stochastic models for service systems. In Part I, we propose a data-driven approach to reduce the delay propagation by optimizing the assignment between incoming and outgoing flights flown by an airline. There are two projects in this part. In the first project, we consider the aircraft assignment problem at a single airport. We propose a data-driven approach to estimate the assignment cost by considering covariates including scheduled arrival time, originating airport and aircraft type of the flights. We conclude that the stochastic assignment derived from this data-driven approach significantly outperforms the actual assignment. In the second project in this part, we extend the previous project to a network of airports by optimizing the assignment between incoming and outgoing flights at each airport in the network. We propose a similar data-driven approach to estimate the assignment costs at each airport, and show that our approach performs better than the benchmark policies.

In Part II, we consider the stochastic models for service systems. There are two projects in this part as well. In the first project, we consider a joint staffing and admission control problem under minimal, partial and full information cases. We compare the profit under different information cases over the parameter space in detail.  In the second project, we consider the joint admission and service rate control problem for a general reward structure under an unobservable (minimal information case) single server queueing system. We show that when the per unit service cost is less than or equal to a critical value, it is optimal to admit all the customers, otherwise, it is optimal to admit none. We show that this socially optimal policy induces the customers to behave in a socially optimal way with self-regulation.

Share this Event