Dynamic Military Airlift Network Design in a Contested Environment

Abstract

The proposed effort aims to extend prior funded research in military airlift network design by developing novel models, approaches and algorithms at the interface between mathematical optimization and machine learning. Leveraging the two areas and the compelling need for decision tools to aid planners with dynamic airlift network design, we focus on learning effective model restrictions to improve solution times in integer linear programs. A recently completed thesis co-advised by the principal investigator presents a new Integer Linear Programming (ILP) formulation to address a key component of dynamic airlift network design. Specifically, the ILP prescribes what airfields to employ daily, the level of working maximum on ground (MOG) capacity to establish at each airfield to include movement of capacity between airfields, and the use of aircraft to move materiel and personnel. Using commercially available optimization software, ILP solution time can be excessive for some scenarios. Research is required to see how we can reduce solution time and test modeling assumptions as we explore additional scenarios and extensions to the ILP. This would include new optimization models, heuristics, and learning effective model restrictions. Discovering useful learned cuts can require a substantial computational effort to build a database of instances solved to optimality or near optimality. We propose to explore how the choice of formulation variables and-or constraint sets is likely to impact the value of learned cuts in a specific Air Mobility Command (AMC) example and then seek to distill how these choices may apply to other ILPs where a learned pattern can potentially be used to eliminate solutions. A static network and predictable routine routes are arguably optimal when assuming air dominance. However, without air dominance, predicable routes and airfield use would provide an adversary with information that would certainly increase the risk of aircraft and-or airfield loss. We propose extending an ILP to enforce different patterns of aircraft and airfield use over time (to limit predictability) and to quantify the impact such patterns would have on risk.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 29, 2024
Source ID
FA95502310168

Entities

People

  • Robert Dell

Organizations

  • Air Force Office of Scientific Research
  • Research Foundation for the State University of New York
  • United States Air Force

Tags

Fields of Study

  • Computer science

Readers

  • Aerospace logistics and air mobility.
  • Distributed Systems and Data Platform Development
  • Operations Research

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks