Choice-Based Optimization in Transport & Logistics

Date:

Time:

10.00 – 16.00 h

Location:

Utrecht

Lecturer:

Dr. Bilge Atasoy & Dr. Jie Gao

Days:

4

ECTS:

1 (attendance only) | 4 (attendance + passing assignment)

Course fee:

Free for TRAIL/Beta/OML/ERIM members, others please contact the TRAIL office

Registration:

See below.

Objectives:

The objective of this course is to provide knowledge on the integrated optimization and behavioral models in order to represent supply-demand interactions endogenously within optimization models. The course begins with a brief introduction to choice modeling methodology and then focuses on various formulations of choice-based optimization models, along with solution methods to address their computational complexity.

Course description:

Traditionally, optimization problems consider the demand as an input which is exogenous. For example, in transportation problems where we decide on the capacity and location of certain facilities, we have a demand matrix representing the demand across the network and across the time horizon. Similarly, when we have a demand model, the supply side is typically considered as a given input, e.g., capacity of the system is an input parameter. However, there are strong interactions between demand and supply. The decisions on the supply side will influence the resulting demand, e.g., the optimized service level has a direct impact on the resulting demand for that service. The changes in the demand patterns also have an impact on the supply as the transport operators will update their decisions, e.g., capacity, route, schedule, based on the evolving demand. Therefore, in this course, we focus on choice-based optimization, which integrates behavioral choice models into optimization frameworks to explicitly capture the mutual influence between supply and demand. Students will learn how to model the behavioral choices of both demand and supply agents, and how to incorporate these models into optimization problems to support system-level decisions that respond to user behavior.

 

The course demonstrates the different formulation possibilities with important transportation problems of facility location, routing, revenue management and pricing. The complexity of these problems is discussed and ideas on potential solution methods are discussed with a few demonstrations. Below is the outline of the course:

  1. Discrete Choice Models
    1. Introduction to random utility models
    2. Choice models, e.g., logit, mixed logit, nested logit
  2. Choice-based Optimization
    1. Recap of mathematical models of optimization
    2. Formulation of choice-based optimization problems
    3. Potential solution methods
  3. Different Classes of Choice-based Optimization Problems
    1. Revenue Management and Pricing
    2. Facility Location
    3. Routing and scheduling (e.g., last-mile logistics, on-demand transport, ride-hailing)
    4. Assortment Optimization and Recommender Systems in Transportation

Assignment:

Students will work on an assignment in groups of 2 or 3. They will work on a transportation related problem that they will choose. For those groups that do not have a particular case, the teachers will help to have an example transport problem and network to start with. Until the second week of lectures, the students will work on a mathematical formulation of the problem they have chosen. They will also implement the mathematical model in Python. During the second week of the lectures, they will present their problem and the formulation. They will receive feedback on their formulation as well as the potential solution methods for their problem. After the lectures, the students will finalize their assignment where they will discuss the solution methods based on the nature of the problem, and they will evaluate the modeling choices, the associated complexity and the potential future developments.

Program:

Literature:

Methodology:

Random Utility Theory, Choice Modeling, Discrete Optimization, Mixed Integer Linear Programming, Linearization and Reformulation Techniques, Approximation Methods and Heuristics.

Course material:

Prerequiste:

Knowledge in mathematical modeling of optimization problems, i.e., operations research.

Pre-registration form


Member of research school: