Contextual Reasoning for Threat Classification Refinement

RT&L FOCUS AREA(S): Artificial Intelligence/ Machine Learning TECHNOLOGY AREA(S): Air Platform; Information Systems; Weapons The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), 22 CFR Parts 120-130, which controls the export and import of defense-related material and services, including export of sensitive technical data, or the Export Administration Regulation (EAR), 15 CFR Parts 730-774, which controls dual use items. Offerors must disclose any proposed use of foreign nationals (FNs), their country(ies) of origin, the type of visa or work permit possessed, and the statement of work (SOW) tasks intended for accomplishment by the FN(s) in accordance with section 3.5 of the Announcement. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. OBJECTIVE: Evolve beyond classification based on a predetermined set of features for a given threat type and reason on a larger set of information to include scene based reasoning and contextual evidence. DESCRIPTION: Advances in technology have allowed adversaries to employ numerous types of reentry vehicles and decoys in an attempt to thwart counter techniques. Understanding and classifying inbound objects quickly provides critical information necessary to perform optimal midcourse defense through precision assignment and targeting of the highest priority objects. To enhance the accuracy of this classification, we pursue advances in contextual reasoning. Recognizing that the threat may not appear as was predicted, that which can be discerned from the context about the engagement and threat presentation could be key. DARPA has expanded the wave' concept of Artificial Intelligence (AI) to a third wave of AI including contextual adaptation, and we aspire to incorporate this paradigm for threat identification. The first wave of AI was expert systems with handcrafted reasoning rules that were defined by subject matter experts. The second wave brought in the new developments in machine learning, primarily supervised deep learning in the classification domain. Although this was a remarkable advancement, it generally dealt with individual classification and obscured the reasons for the classification call. For complex scenes, hierarchical reasoning, scene based reasoning, or evidence based reasoning for an expanded concept and exploitation of evidence, may enable us to achieve additional accuracy for threat identification, as well as gain confidence knowing why the determination was made. The goal is to develop a better classifier, with disparate sensor data. that assesses classification of objects, including over time (as opposed to single shot classifiers) so that salvo firing control can be better informed to target high priority objects. The classifier must be mission assured by informing the operator of an unknown classification of an object rather than best guess by way of some metric in order to prevent misclassifications on zero-day events. PHASE I: Demonstrate proof of principle with technology prototype. From self-generated representative data from two different sensor types, design a classifying technology utilizing additional scene, or context, information. Deliver Algorithm Description Document (ADD) to accompany final report with test results and analysis. PHASE II: Using realistic, relevant data, produce matured technology prototype. Define preferred data and message content for optimal technology performance. Validate concept with flight and/or ground test data. Assess performance across a range of government supplied scenarios and data compositions. Deliver ADD with test results. This phase will likely be classified due to the nature of the data. PHASE III DUAL USE APPLICATIONS: The topic has numerous unclassified applications, as well as alternate classified applications. For example, disease detection, mental illness diagnosis, terrorist identification, disaster management, and natural language processing. REFERENCES: 1. Looking for a synergy between human and artificial cognition: Br zillon, P., Blackburn, P., Dapoigny, R. (eds.) CONTEXT 2013. LNCS (LNAI), vol. 8175, pp. 45 58. Springer, Heidelberg (2013). ; 2. Representation of procedures and practices in contextual graphs. Br zillon, P. Knowl. Eng. Rev. 18, 147 174 (2003). ; 3. Reasoning with Contextual Knowledge and Influence Diagrams. Erman Acar, Rafael Pe aloza, arXiv:2007.00571v1. ; 4. Relational inductive biases, deep learning, and graph networks, P.W. Battaglia et al, arXiv:1806.01261v3, October 2018 ;