Use of Flow Field Forecasting to Anticipate Events in a Multi-Threat Environment
Abstract
Allocation of resources within or beyond the battle space must often be done in an environment of uncertain or evolving parameters. In such situations forecasting can reduce uncertainty and dramatically enhance the mission-effectiveness of the resource allocation. Flow field forecasting, a statistical learning methodology introduced by us in 2011, has a mix of technical capabilities—model independence, autonomous operation, computational efficiency—that no standard general-purpose forecasting tool can match. Our team, which combines expertise in mathematics, statistics and computer science with experience in academia, naval leadership and military consulting, proposes a one-year project to advance the capability of flow field forecasting in two key respects. We propose to extend flow field forecasting to explicitly address multivariate data flows and to remove the need for a user-specified history space. Data from modern sensor and reporting systems are typically multi-faceted, and the ability of a forecasting tool to accept multivariate data is a necessity. To remove the need in flow field forecasting for a user-specified history space, we propose a novel, transformative re-design in which all histories—of all forms and lengths—are searched for correlative past histories that are relevant to the present. This will give flow field forecasting a new mechanism for adapting to rapidly changing dynamical structure in arriving data. Accomplishing these objectives will significantly advance the science of forecasting.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 09, 2016
- Source ID
- N002441510052
Entities
People
- Kyle Caudle
Organizations
- South Dakota School of Mines and Technology
- United States Air Force