Thesis Title: Operational and Tactical Strategies for Service Networks

Advisors: 

Dr. Martin Savelsbergh, School of Industrial and Systems Engineering, Georgia Tech

Dr. Natashia Boland, School of Industrial and Systems Engineering, Georgia Tech

Dr. Alan Erera, School of Industrial and Systems Engineering, Georgia Tech

Committee members:

Dr. Pascal Van Hentenryck, School of Industrial and Systems Engineering, Georgia Tech

Dr. Sushil Poudel, UPS Supply Chain Solutions

Date and Time: 8 - 10 am ET, Monday, November 30, 2020

Meeting URL (for BlueJeans): https://gatech.bluejeans.com/890727493

Meeting ID (for BlueJeans): 

890 727 493

Abstract:

This thesis focuses on two independent aspects of service network planning. The first aspect is operational and is related to load plan adjustment in Less-Than-Truckload (LTL) freight networks. The second one is both tactical and operational and is related to equipment management in small package networks. In the first part, we suggest near real-time routing and load plan adjustment strategies to improve the system-wide daily performance of the freight network. In the second part, we present different strategies for managing inventory levels of different equipment types in a large-scale network.

In Chapter 2, we design and implement decision support technology to assist dispatchers in the daily management of load plans in LTL networks. The freight volume that enters a service network on the day of operations deviates from the forecast freight volume used to create the load plan. These deviations cause inefficiencies when the capacity on planned freight paths is no longer sufficient and delays result in missed service promises. Near real-time load plan adjustments, i.e., rerouting freight on alternate paths, can improve on-time performance without incurring additional cost (e.g., without purchasing additional capacity). We model the problem of identifying effective alternate freight paths on a time-expanded network and we develop fast heuristics for its solution to ensure that there is sufficient time to put the adjusted load plan in place. The load plan adjustment technology has been extensively tested using data from a large US LTL carrier. The results show that on-time performance can be improved without increasing cost, i.e., by rerouting freight and using existing capacity in the service network.

In Chapter 3, we develop efficient and effective short-term equipment management strategies for small package express carriers. We start by investigating substitution-based equipment balancing for carriers operating multiple equipment types in their service network. The weekly schedule of movements used to transport packages through the service network leads to changes in equipment inventory at the facilities in the network. We seek to reduce this change, i.e., the equipment imbalance associated with the schedule of movements, by substituting the equipment types initially assigned to the movements. We model this problem using a hierarchical optimization approach and suggest two heuristics to solve it. We also explore the value of integrating empty repositioning decisions in the model. Furthermore, we performed a computational study using real-world instances to analyze the performance of an IP based solution approaches and assess the benefits of substitution-based equipment balancing and integrating empty repositioning.

In Chapter 4, we shift from the previous equipment balancing perspective to an inventory aware equipment management perspective where the time dimension is considered. We formulate a mixed integer program that tracks the inventory of each equipment type at each facility and seeks to minimize the cost of empty repositioning required to execute a given load plan and prevent stock-out occurrences, by substituting the equipment type assigned to loaded moves (respecting compatibility requirements) and adding new empty movements between facilities. We analyze the complexity of some special settings of the problem, propose a parsimonious time-discretization to control the size of the model, and introduce a dynamic variable generation algorithm to solve it. Computational experiments show that a significant reduction in the cost of empty movements required in the network can be obtained and using appropriately chosen equipment substitutions.