Heureka Corporation — Department of Defense SBIR Phase I: N151-051

Description

Heureka and Lockheed Martin will develop an AI-based approach to replicate human change request analysis. The approach utilizes state-of-the-art knowledge engineering to elicit expert tacit rules from past CRs. To replicate current processes, we will build and train models which replicate each product area's reasoning. This has several benefits. First, it emulates current human based processes (product areas often disagree). Second, it allows a product area expert to easily compare their reasoning with the model's thus aiding validation. Since subjectivity often creeps into the current CR process, we will augment model behavior by providing tools which automatically 1) "understand" Aegis CSS's static and dynamic characteristics and 2) use natural language processing to analyze Aegis documentation, test procedures, etc... This provides objectivity heretofore unavailable. We will also develop a model which consolidates all product areas' reasoning (again, mirroring current processes). Each step of our tool will be able to explain why a decision was made, thus allowing experts to understand the underpinnings of the process. Finally, we will apply sophisticated processing to not only prioritize CRs, but also take into account the incremental impacts of "similar" CRs. This allows more CRs to be included in a build than are currently possible.