Fisika & Matematika editorial
Open AccessOA2023
A Survey on Bilevel Optimization Under Uncertainty
This survey provides a comprehensive overview of bilevel optimization under uncertainty, covering stochastic and robust techniques, limited observability, near-optimal decisions, and intermediate solution concepts, with applications in energy, security, management, and networks.
Yasmine Beck; I. Ljubiฤ; Martin Schmidtยท European Journal of Operational Researchยท 2023ยท DOI 10.1016/j.ejor.2023.01.008
The core problem
Bilevel optimization is a highly active field of applied mathematics, primarily because bilevel problems can model hierarchical decision-making processes. However, this modeling power comes at the cost of computational and theoretical challenges. Significant algorithmic advances over the past two decades have enabled the solution of much larger and more complex bilevel problems than previously possible. This survey focuses on a particularly challenging class: bilevel optimization under uncertainty. It reviews classic approaches using stochastic or robust techniques and highlights that uncertainty in bilevel settings is richer than in single-level problems, as uncertainty can affect not only problem data but also the observation of decisions of the two players. The survey also covers limited observability, near-optimal decisions, intermediate solution concepts between optimistic and pessimistic cases, and applications in energy, interdiction games, security, management sciences, and networks.
Innovation
The survey systematically reviews methods for handling uncertainty in bilevel optimization. It categorizes approaches into stochastic and robust techniques. Stochastic methods assume probabilistic uncertainty, while robust methods consider worst-case scenarios. The authors emphasize that uncertainty in bilevel problems arises from multiple sources: data uncertainty (as in single-level problems) and uncertainty in the observation of decisions (limited observability). The survey also discusses solution concepts that lie between the optimistic and pessimistic cases, and problems where the follower may make near-optimal decisions. The review is structured around these themes, providing a comprehensive overview of the state of the art.
Introduction
Bilevel optimization is a highly active field of applied mathematics, primarily because bilevel problems can model hierarchical decision-making processes. However, this modeling power comes at the cost of computational and theoretical challenges. Significant algorithmic advances over the past two decades have enabled the solution of much larger and more complex bilevel problems than previously possible. This survey focuses on a particularly challenging class: bilevel optimization under uncertainty. It reviews classic approaches using stochastic or robust techniques and highlights that uncertainty in bilevel settings is richer than in single-level problems, as uncertainty can affect not only problem data but also the observation of decisions of the two players. The survey also covers limited observability, near-optimal decisions, intermediate solution concepts between optimistic and pessimistic cases, and applications in energy, interdiction games, security, management sciences, and networks.
Why it matters
The authors discuss the implications of uncertainty in bilevel optimization and identify open research directions. They note that the richness of uncertainty sources makes bilevel problems under uncertainty particularly challenging, both theoretically and computationally. The survey calls for further research into solution methods that can handle limited observability and near-optimal decisions, as well as the development of intermediate solution concepts. The wide range of applications demonstrates the practical relevance of these problems. The survey concludes that bilevel optimization under uncertainty is a vibrant and evolving field with many opportunities for future contributions.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member contentโฆ