You are here

DIRECT TO PHASE II – Cognitively Inspired Artificial Intelligence for Automated Detection, Classification, and Characterization


OUSD (R&E) MODERNIZATION PRIORITY: 5G;Artificial Intelligence (AI)/Machine Learning (ML);Autonomy


TECHNOLOGY AREA(S): Information Systems


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 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: Develop a human-level/human-style artificial intelligence (AI) that can perceive and explain signals implicit in magnetics, electro-optical and infrared (EO/IR), and acoustics data to achieve long-range detection, tracking, and classification of maritime surface and subsurface contacts, which is an essential and imperative Naval capability.


DESCRIPTION: Nontraditional adaptive human-level/human-style artificial intelligence (AI) signal-processing algorithms have the potential to increase detection range in high littoral noise environment while extracting precise target signatures that optimize detection, tracking, and classification.


The consistently novel, noisy, and nonlinear aspects of magnetics, EO/IR, and acoustics data, critical for Navy detection, tracking, and classification, present particularly difficult problems for the standard techniques of machine learning (ML) (e.g., deep learning [Ref 1]). To be functional, the artificial neural networks of the latter must be trained on large curated data of “signal” in order to filter out noise. Such pattern recognition has revolutionized many domains [Ref 2], from image classification to game playing. However, such a methodology fails catastrophically in domains where data are frequently changing, and are neither large, curated, nor efficiently transformable into linear-vector form [Ref 3]. This is because engineers in standard ML fail to understand that “making comes before matching” [Ref 4]: a competence to generate general/adaptable pattern schemata prior to data processing is necessary to recognize novel (untrained) patterns in novel and potentially sparse and noisy/nonlinear data. Fundamentally, to be of greatest value to the Navy, these schemata ought to be conjectured explanations for the signals, beyond their mere detection. Such explanatory competence cannot be implemented in standard ML [Ref 5], hence the failure of ML to solve non-linear, signal-to-noise over noise problems.


However, such a competence is characteristic of some “good old fashioned artificial intelligence” architectures and the human intelligence they emulate [Ref 6]. Humans can routinely recognize novel signals in variable noise environments, and the Navy has relied on this human intelligence to process magnetic, EO/IR, and acoustics data for detection and classification of objects and environmental footprints. Indeed the evolutionarily-optimized, pattern-matching of the human mind/brain can expertly and efficiently recognize (i.e., “make-and-match”) patterns in novel/noisy time series as expertly and efficiently as it recognizes grammatical patterns in language [Ref 7]. Of profoundest importance—equipped with language—human intelligence seeks to explain these patterns. It generates causal knowledge. Obviously, however, it would be an intractably Herculean task for humans to process and interpret all the magnetics, EO/IR, and acoustics data necessary to satisfy naval objectives of near-real-time, long-range detection, tracking, and classification of surfaced and submerged objects with 90% probability of detection (Pd) rates. Hence, the optimal solution would be to implement the competence of human intelligence in the machinery of artificial intelligence. Thus, the Navy requires human-style AI.


Work produced in Phase II may become classified. Note: The prospective contractor(s) must be U.S. owned and operated with no foreign influence as defined by DoD 5220.22-M, National Industrial Security Program Operating Manual, unless acceptable mitigating procedures can and have been implemented and approved by the Defense Counterintelligence and Security Agency (DCSA) formerly Defense Security Service (DSS). The selected contractor must be able to acquire and maintain a secret level facility and Personnel Security Clearances. This will allow contractor personnel to perform on advanced phases of this project as set forth by DCSA and NAVAIR in order to gain access to classified information pertaining to the national defense of the United States and its allies; this will be an inherent requirement. The selected company will be required to safeguard classified material IAW DoD 5220.22-M during the advanced phases of this contract.


PHASE I: For a Direct to Phase II topic, the Government expects that the small business would have accomplished the following in a Phase I-type effort. Have developed a concept for a workable prototype or design to address, at a minimum, the basic requirements of the stated objective above. The below actions would be required in order to satisfy the requirements of Phase I:


The development of Strong Artificial Intelligence: human-level/human AI capacitated with the linguistic competence to generate causal explanatory models via critical rationalism. Given a set of big or small data, constructs a Chomskyan grammar construct to model causal relations, thereby transcending descriptions that answer what is being observed, transforming data into evidence for/against conjectured explanations that answer why and how the data, or the phenomena underlying the data, - exist and behave.


The objective is a linguistically competent AI that can generate explanatory causal models. An example would be the classic board game Battleship. The strong cognitive AI is given a partially revealed board state and must, by its epistemic process of critical rationalism (i.e., conjecture-and criticism), discover the true state of the board. What the solution ultimately seeks is an explanation for the configuration of the partially revealed gameboard: “Why does the board appear this way?” The reason why is the complete configuration (i.e., the position of the hidden ships). It is an exercise in explanatory causal-model-construction. The Cognitive AI succeeds and surpasses humans in constructing “the ultimate question”: one question whose answer reveals the true state of the board. This is formally analogous to a complete explanation for some complex phenomena in the real world. Importantly, this problem cannot be solved by reinforcement learning, as Google DeepMind did for “Go” and “Chess”. It requires a cognitive strong intelligence: language and explanation. AI researchers have tested numerous techniques to solve this problem, from tree search to Bayesian models, but all fail to attain human-level competence, the common denominator being limitations imposed by hard-coded rules with single objectives that are insufficiently adaptive and creative.


FEASIBILITY DOCUMENTATION: Offerors interested in participating in Direct to Phase II must include in their response to this topic Phase I feasibility documentation that substantiates the scientific and technical merit and Phase I feasibility described in Phase I above has been met (i.e., the small business must have performed Phase I-type research and development related to the topic, but from non-SBIR funding sources) and describe the potential commercialization applications. The documentation provided must validate that the proposer has completed development of technology as stated in Phase I above. Documentation should include all relevant information including, but not limited to: technical reports, test data, prototype designs/models, and performance goals/results. Work submitted within the feasibility documentation must have been substantially performed by the offeror and/or the principal investigator (PI). Read and follow all of the DON SBIR 22.1 Direct to Phase II Broad Agency Announcement (BAA) Instructions. Phase I proposals will NOT be accepted for this topic.


PHASE II: Develop a prototype of a human-level/human-style AI that can perceive and explain “artificial” (i.e., invented) magnetics, EO/IR, and acoustics data in an idealized Navy war game simulation. Further development of the AI prototype and its demonstration on “natural” (i.e., real-world) data in a realistic Navy war game simulation. Perform sea trials data collection of individual vessels in terms of feature identification performance, operational agility, and accuracy. Perform limited sea trial test data analysis of surface and subsurface objects.


Work in Phase II may become classified. Please see note in Description paragraph.


PHASE III DUAL USE APPLICATIONS: Continue to refine magnetics, EO/IR, and acoustics data. Finalize sea trials data collection of individual vessels in terms of feature identification performance, operational agility, and accuracy. Complete final testing and perform necessary integration and transition for use in antisubmarine and countermine warfare, counter surveillance and monitoring operations with appropriate current platforms and agencies, and future combat systems under development.


Commercially this product could have applicability in search and rescue operations; and could be used to enable remote environmental monitoring of geophysical survey, facilities, and vital infrastructure assets.



  1. Fridman, L. (2020). MIT deep learning and artificial intelligence lectures. Massachusetts Institute of Technology. Retrieved on August 24, 2020, from
  2. Sejnowski, T. (2018). The Deep learning revolution. Cambridge: MIT Press.
  3. Mitchell, M. (2019). Artificial intelligence: A guide for thinking humans. Farrar, Straus and Giroux.
  4. Eccles, J. C., & Popper, K. (1977). The self and its brain: an argument for interactionism. Routledge.
  5. Marcus, G., & Davis, E. (2019). Rebooting AI: building artificial intelligence we can trust. Pantheon Books.
  6. Watumull, J. (2012). A Turing program for linguistic theory. Biolinguistics, 6(2) 222-245.
  7. Watumull, J., & Chomsky, N. (n.d.). Rethinking universality. In A. Bárány, T. Biberauer, J. Douglas, & S. Vikner (Eds.), Syntactic architecture and its consequences II: between syntax and morphology (Vol. 2).
  8. Godin, O. (2006, October 17). Anomalous transparency of water-air interface for low-frequency sound. Physical review letters, 97(16).
  9. Defense Counterintelligence and Security Agency. (n.d.).
  10. Department of Defense. (2006, February 28). DoD 5220.22-M National Industrial Security Program Operating Manual (Incorporating Change 2, May 18, 2016). Department of Defense.
  11. Silver, D., Hubert, T., Schrittwieser, J., & Hassabis, D. (2018, December 6). AlphaZero: Shedding new light on chess, shogi, and go. Deepmind.


KEYWORDS: Artificial Intelligence; AI; Nonlinear; signal processing; Cognitive; Nontraditional

US Flag An Official Website of the United States Government