Optimization under uncertainty and conflict: Algorithms for heterogeneous quadratic programs
Abstract
Short Work StatementThe proposed research will develop efficient algorithms for solving quadratic programming problems.ObjectiveThe objective of the proposed research is to develop a tool box of efficient algorithms based on rigorous mathematical theory for solving optimization problems modeled as heterogeneous Quadratic Programming (QP) problems. Here, heterogeneous means that theQP problem can contain mixed-integer variables, nonconvex objective and/or constraint functions, unknown/uncertain data, conflicting, multiple objective functions.ApproachThe approach consists of several sub-tasks. For example, to deal with mixed-integer variables, the PIs will employ new convexification techniques. These techniques should yield methods for deriving valid linear and non-linear inequalities, which can then be incorporated into a branch-and-cut procedure. A second sub-task will be to address parametric mixed-integer QPs, which build upon the first sub-task. To address these type of problems, first the convexification procedure from sub-task 1 will be applied, followed by techniques from Lagrangian duality.Overall Merits and ONR Mission Relevance: The proposed research will develop innovative foundational and algorithmic techniques for solving an important class of optimization problems. Quadratic optimization problems arise frequently in several application domains of naval interest. These type of problems are found in the control of mechanical and electronic systems, signal and image processing, and in clustering problems in statistics.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141612725
Entities
People
- Margaret Wiecek
Organizations
- Clemson University
- Office of Naval Research
- United States Navy