Adaptive Bilevel Mixed-Integer Programming
Abstract
Bilevel optimization provides a general framework for decision making in complex systems that do not have a centralized decision-maker who controls and-or optimizes all actions in the system. Bilevel optimization can be used to capture adversarial relationships between multiple decision-makers, which naturally arise in various types of interdiction and defender-attacker settings. Furthermore, bilevel optimization is also applied in collaborative environments, in particular, for applications that involve decentralized resource allocation, planning, logistics, and scheduling. Bilevel programming problems have been extensively studied over the past several decades; however, the vast majority of these works study bilevel programs that are inherently deterministic and-or that only have two decision-makers (an upper-level decision-maker called the leader and a lower-level decision-maker called the follower). Furthermore, the limited number of studies that consider uncertainty-multiple follower settings do not take into account the practical consideration that the decision-makers can often make some decisions after some or all of the uncertainty is realized. In many defense and security related contexts, for example, the decision-makers prefer to have available several pre-computed strategies that can be activated whenever some, but not necessarily all, uncertainty about the actual situations is realized. The proposed project will study adaptive bilevel optimization problems in which the upper-level decision-maker has the capability to adapt some of his-her decisions after uncertainty (or at least some part of it) about the followers is realized.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 06, 2024
- Source ID
- FA95502310372
Entities
People
- Jourdain Lamperski
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Pittsburgh