Description:
TECHNOLOGY AREA(S): Materials
OBJECTIVE: Develop an analytical capability to overcome challenges inherent in predicting potential influencing factors, performance, and mission effectiveness within complex, multi-domain challenges that the future USAF will face.
DESCRIPTION: Development Planning efforts across the Air Force include gap analysis, emerging technology assessments, war-gaming, experimentation, performance analysis, operational analysis, requirement development, acquisition strategies, and investment strategies. In general, these efforts aim to inform investment decisions by estimating the cost and mission capability of potential alternatives. Despite these efforts, the Air Force has had mixed results in identifying multi-domain problem elements which have the largest impacts on mission success. One reason for the mixed results is that the Air Force needs to evaluate a vast number of cross-domain solutions and strategies in order to have confidence in its understanding of the trade-space. During Development Planning, the vast number of potential solutions are often difficult, if not impossible, to evaluate using sophisticate methods, like simulations, due to resource and computation limitation. An alternative to more sophisticate methods could include Bayesian networks. Bayesian networks have long been utilized to understand the probabilistic relationship between a set of variables and the relationship that may or may not exist between them. They can be a powerful tool for identifying the key interactions in complex systems and for evaluating alternative approaches to achieving an end goal. An important, poorly understood, limitation of current applications of Bayesian networks is the implied assumption that the conditional probabilities embedded in the network all have approximately similar uncertainty. Efforts here will seek to develop a toolset across the materiel and non-materiel spectrum that will extend approaches like the Bayesian networks to include the uncertainty in the conditional probabilities enabling a more precise understanding of those larger system elements that have the greatest overall influence on success or failure. For example, current Analysis of Alternatives tools do not include selected technology readiness level impacts on the solution. Further work will develop a toolset that provides straightforward inclusion of confidence/uncertainty levels into the Bayesian network analysis approach to predicting performance of complex systems. This would have a wide range of applications to include areas as diverse as: Improved health diagnostic tools, Analysis of Alternatives tools, performance analysis of complex systems, and multi-domain mission effectiveness analysis.
PHASE I: Develop and demonstrate a methodology to extend Bayesian networks (or another innovative approach) by including uncertainty in conditional probabilities. Provide general descriptions of how such a methodology could be applied to the Air Force’s development planning process and an example of how it could be used to estimate the effectiveness of alternatives.
PHASE II: Implement a Bayesian network (or another innovative approach) into a delivered tool to evaluate, understand, and predict key influential technology areas within a multi-domain, Air Force centric, future challenge. Demonstrate and validate the utility of such a tool to predict emerging concepts along the materiel and non-materiel spectrum that are most influential to the successful of a campaign and thus deserving of further investigation and/or maturation. Ideally this tool should run on a high-performance windows operating system laptop or desktop; however, a LINUX based operating system is also acceptable.
PHASE III: The technology developed in Phase I and demonstrated in Phase II will have application throughout government and industry.
REFERENCES:
1: Glenny, V., "A Framework for the Statistical Analysis of Probability of Mission Success Based on Bayesian Theory." (2014). Defense Technical Information Center (DTIC) report number ADA610732. http://www.dtic.mil/docs/citations/ADA610732
2: Uusitalo, L., "Advantages and Challenges of Bayesian Networks in Environmental Modelling." Ecological Modelling, 203(3), 312-318. (2007).
3: Chan, H., and Darwiche, A., "When do numbers really matter?" Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence (pps. 65-74), Morgan Kaufmann Publishers Inc. (August 2001).
KEYWORDS: Measures Of Effectiveness, System(s)-of-systems, Bayesian Nets, Probabilistic-based Methods
