Air force 17. 1 Small Business Innovation Research (sbir) Phase I proposal Submission Instructions

2. Howard, C.(2013). UAV Command, Control and Communication.Military and Aerospace Electronics,Volume 24(7). Retrieved from  http://www.militaryaerospace.com/articles/print/volume-24/issue-7/special-report/uav-command-control-communications.html.

3. Stansbury, R. S.. Vyas. M. A.,& Wilson, T. A. (2009). A Survey of UAS Technologies for Command, Control, and Communication (C3).Journal of Intelligent and Robotic Systems.Volume 54(1-3),61-78.

4. USAF Strategic Master Plan 2015, Page 59;Vector: Continue the Pursuit of Game-Changing Technologies; Retrieved fromhttp://www.af.mil/Portals/1/documents/Force%20Management/Strategic_Master_Plan.pdf.

5. Autonomous Horizons, System Autonomy in the Air Force -A Path to the Future, Volume I: Human-Autonomy Teaming, AF/ST TR 15-01 June 2015 Retrieved fromhttp://www.af.mil/Portals/1/documents/SECAF/AutonomousHorizons.pdf? timestamp=1435068339702 .

KEYWORDS: Autonomous communications, resilient communications, access-denied communications, adaptive communications, shared situational awareness, command and control

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 section 5.4.c.(8) of the solicitation and within the AF Component-specific instructions. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. Please direct questions to the AF SBIR/STTR Contracting Officer, Ms. Gail Nyikon, gail.nyikon@us.af.mil.

OBJECTIVE: Overcome the Big Data challenges of analyzing multi-int data to enable scalable analytics for Activity Based Intelligence.

DESCRIPTION: The number, variety and fidelity of sensors continue to increase yielding increasingly complex analytical challenges. Activity-based intelligence (ABI) has emerged as a new intelligence methodology aimed at exploiting this opportunity. The Under Secretary of Defense for Intelligence defines ABI as or "a multi-INT approach to activity and transactional data analysis to resolve unknowns, develop object and network knowledge, and drive collection." [1] Fundamentally, ABI is a discipline of intelligence where analysis and subsequent collection are focused on the activity and transactions associated with an entity, population, or area of interest "in decision advantage." While ABI has emerged within the IC, another shift has occurred within technology that creates the opportunity to realize the benefits currently obscured by the variety, volume, and velocity of the data.

Open source Big Data technology now enables the storage, retrieval, analysis and exploitation of data for which traditional data processing applications were inadequate. The details and aims of Big Data technologies vary, however the unifying theme tends to be the usage of commodity hardware and parallel computing. In addition, open source Big Data technologies are increasingly creating an ecosystem in which a variety of technologies build upon each other or work well together. Hadoop [2] and Spark [3] present new opportunities in batch analytics while systems like Samza [4] and Storm [5] tackle unwieldy data streams. In addition, generic architectures like the Lambda Architecture [6] have emerged that seek to combine Big Data technologies to enable scalable and fault-tolerant data processing architectures.

Unfortunately, the full potential of big data technologies has yet to be realized with regard to ABI and the processing of geospatial activity and transactions. Typically, algorithmic advances with ABI applications have been developed within traditional data processing architectures. In addition, Big Data technologies are often utilized for basic storage and retrieval, or very basic statistics. This SBIR aims to develop advanced analytics within Big Data architectures to produce a game changing capability to understand and act on activity and transactional multi-INT data from a variety of different sources. The only agreed upon notion of what constitutes Big Data is data sets that are so large or complex that traditional data processing applications are inadequate. The one of the appealing aspects of solutions provided within Big Data architectures is the potential to achieve horizontal scalability so the solution may be scaled up or down as necessary. In addition to transforming intelligence, and the ways in which people intermingle and interact with the physical and virtual world.

PHASE I: Develop novel algorithms utilizing Big Data technology to discover relevant patterns, determine and identify change, and\or detect and analyze trends within multi-int data. Evaluate the performance and viability of these methods using realistic data sets (simulated or collected) (demonstrate feasibility)

PHASE II: Implement algorithm prototypes in a realistic environment that enables thorough testing of algorithms. Incorporate applications to support testing, e.g., operator displays, decision support systems. Demonstrate and validate algorithm(s) effectiveness. Deliver an algorithm description document, engineering code and test cases. Explore and document other potential methodologies identified in Phase I.

PHASE III DUAL USE APPLICATIONS: DUAL USE APPLICATIONS: Develop and mature the technology for use within the DoD, Intelligence Community, and Homeland Security as well as other viable commercial applications (marketing, business intelligence, cyber security...).

REFERENCES:

1.  C. P. Atwood, "Activity-Base Intelligence: Revolutionizing Military Intelligence Analysis." 01 April 2015. [Online]. Available:http://ndupress.ndu.edu/Media/News/NewsArticleView/tabid/7849/Article/581866/jfq-77-activity-based-intelligence-revolutionizing-military-intelligence-analys.aspx. [Accessed 20 04 2016].

2. "Apache Hadoop, [Online}. Available:http://hadoop.apache.org/. [Accessed 20 04 2016].

3. "Apache Spark,"[Online]. Available:http://spark.apache.org/. [Accessed 20 04 2016].

4.  "Samza," [Online]. Available:http://samza.apache.org/.   [Accessed 20 04 2016].

5. "Storm." [Online]. Available:http://storm.apache.org/. [Accessed 20 04 2016].

6.  "Lambda Architecture." [Online]. Available:http://lambda-architecture.net/. [Accessed 20 04 2016].

KEYWORDS: Activity Based Intelligence, Big Data, Analytics

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 section 5.4.c.(8) of the solicitation and within the AF Component-specific instructions. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. Please direct questions to the AF SBIR/STTR Contracting Officer, Ms. Gail Nyikon, gail.nyikon@us.af.mil.

OBJECTIVE: Develop collaborative/collective intelligence system where human interaction is independent but cooperative with autonomous analytics for information exploitation.

DESCRIPTION: The analyst community faces ever-increasing volume of data products from numerous data sources. The ISR community is relying heavily on improvements to machine analytics. However, one of the challenges to machine analytics is the reliance on historical data and to adapt to the dynamic nature of human decision-making. New scenarios or events often calls for centralized organizations to use subject matter experts (SME) to create new processes or techniques. The SME faces the daunting task of reaching out to other experts and singularly judging the credibility of their analysis.

The product ranking in Amazon may be considered collaborative filtering [1,2). Similarly, the ISR community could rank the quality of data, refer products to peers and even propose alternative uses for information in other mission areas.

Another suitable use of collaborative intelligence is to break "stovepipes" and reduce bureaucratic inter-agency overhead. Rather than conduct collaboration within organizations, on-line communities could emerge over areas of expertise. In this way, a seismologist working on earthquakes could collaborate with a seismologist working on nuclear treaty monitoring.

The focus of this topic is to implement commercial success in collaborative intelligence for information exploitation. The concept goes beyond human-in-the-loop techniques but puts the human component as neither superior or inferior to the machine component.

Collaborative intelligence will serve as a powerful tool to address the challenges of understanding human interaction. For instance, in 1999, Microsoft sponsored a chess match between Garry Kasparov and volunteers around the world. Kasparov, considered the leading chess player in the world played a game against 50,000 people in 75 countries in a game of chess. Ultimately, the ''world" lost but Kasparov considered it one of the most complex games in history [3]. The world players also had coaches in the game which would be analogous to the SMEs in their respective fields.

PHASE I: Identify commercial and government social and collaborative intelligence processes for use within the ISR community. Determine solutions to obstacles in implementing such solutions for the DoD and IC communities.

PHASE II: Implement technologies identified in Phase I and integrate within current mission systems. Assess the value of such technologies in the decision making process.

PHASE III DUAL USE APPLICATIONS: Targeted technologies will transition to AFRL, 25AF and other DoD partners. They will be incorporated into the ISR PED and US Government cloud infrastructure.

REFERENCES:

1.  Adomavicius G. and Tuzhilin, A., "Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions," in IEEE Transactions on Knowledge and Data Engineering, vol. 17, no. 6, pp. 734-749, June 2005.

2.  Long-Zhen, W. Gui-Fen and R. Yan, "Collaborative Filtering with Improved Item Prediction Approach for Enhancing the Accuracy of Recommendation, "2012 Fourth International Conference on Multimedia Information Networking and Security, Nanjing, 2012, pp. 349-352.

3.  Marko, P. and Haworth, G. M. (1999) The Kasparov-World Match. ICGA Journal, 22 (4) pp. 236-238.

KEYWORDS: Collaborative intelligence, collective intelligence, multi-source, multi-int, machine learning.

AF171-050

TITLE: Human-Machine Collaboration for Automated Patterns of Life Analysis

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 section 5.4.c.(8) of the solicitation and within the AF Component-specific instructions. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. Please direct questions to the AF SBIR/STTR Contracting Officer, Ms. Gail Nyikon, gail.nyikon@us.af.mil.

OBJECTIVE: Develop a fully interactive system that encourages human-machine collaboration for Patterns of Life analysis over multi-modal data.

DESCRIPTION: Automated pattern recognition and predictive analytics technologies have emerged to help intelligence analysts more efficiently process large streams of multi-INT data feeds. However, automated solutions still provide only limited value. For example, while development of annotation and benchmark datasets [5] has advanced large-scale single-image scene classification and object detection models, state of the art computer vision algorithms can only detect the objects they are well trained on, and still produce detection rates much lower than humans [4]. Even with existence of annotated imagery datasets, the fully supervised learning is inefficient, and semi-supervised learning have been developed. Further, video-based activity recognition algorithms can't assume the presence of large training datasets due to manpower requirements in annotating videos [3], and recently focused on semi-supervised learning from sparse set of example [2] to improve classification accuracy. Recently, a significant interest of the computer vision community is on enabling collaboration between humans and machines in producing better classification performance [5].

All-source analysts are faced with even greater challenge: they need to combine data from different modalities, and the types of "useful" results depend highly on the types of tasks that analysts have. Consequently, development of tools for multi-source multi-modal fusion has to account for the ever changing needs of the analysts, and hence must be adaptive to the needs of the users. Learning patterns of life in multi-source multi-model domains, and especially in Anti-Access Areas of Denial (A2AD) environments [1], is akin to discovering "unknown unknowns". But how can automated information fusion algorithms discover such patterns? Drawing on the advances achieved in computer vision, we hypothesize that semi-supervised multi-int learning algorithms that use supporting concept libraries (Ye et al., 2015) and interactively engage the human analysts may facilitate the technology development in this area.

The focus of this topic solicitation is to create a full user-in-the-loop feedback system that encourages the analysts to provide expert guidance in patterns of life specification. Solutions are needed that leverage minimal inputs from humans to help multi-int fusion algorithms develop models that are tailored to the analyst's preferences, the characteristics of the data and domain specific knowledge. Such a system may have algorithmic core that can learn patterns from large volumes of unlabeled training data in unsupervised manner, present representative candidates to the analysts to validate and annotate, and incorporate their feedback in semi-supervised manner to re-label and re-structure the resulting patterns. The user-should be able to provide corrections to annotations as well as patterns of life that automated solution develops, as well as define their own patterns re-using the information provided by other users or algorithms as "building blocks". Overtime, the pattern library is populated with highly-accurate, user-vetted patterns which are tailored to the user, INT type, data source and analysis task.

Solutions of interest will be focused on implementing the user-in-the-loop feedback system described above. Novel methods for online-learning of patterns of life from incremental, multi-source, multi-INT data streams are encouraged. The proposed solution burdensome on the human but maximally valuable to the machine. The solution must be able to accommodate dynamic patterns of life and conflicting observations in the input data.

PHASE I: Using representative open source data of at least two modalities (e.g. vision or text), develop and demonstrate a set of algorithms that learn patterns of life and accept user-provided corrections. Techniques should be able to process noisy data and demonstrate scalability. Demonstrate a proof-of-concept to AFRL.

PHASE II: Using online-streaming multi-INT, multi-source data, develop and demonstrate a set of algorithms with the Enhanced Exploitation and Analysis Tools (E2AT) environment located at AFRL/RRS All Source Process and Exploitation (APEX) Center that would provide rapid learning and re-learning of patterns of life using analyst feedback. Techniques should demonstrate improvements over state-of-the-art processing. Demonstrate a robust proof-of-concept with the E2AT environment.

PHASE III DUAL USE APPLICATIONS: Targeted technologies would be to transition to matured AFRL E2AT, NASIC, and PCPAD-X programs. Any application, whether military or commercial, requires methods and capabilities to develop and learn pattern of life indicators and hot to apply to analysis of customer requirements.

REFERENCES:

1.  Gao, J.; Ling, Haibin; Blasch, Erik; Pham, Khanh; Wang, Zhonghai; Chen, Genshe. "Pattern of Life from WAMI Objects Tracking based on Context-Aware Tracking and Information Network Models". Retrieved 23 May 2013.

2.  Misra, I., Shrivastava, A., & Herbert, M. (2015). Watch and Learn: Semi-Supervised Learning for Object Detectors From Video. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 3593-3602).

3.  Oh, S., Hoogs, A., Perera, A., Cuntoor, N., Chen C. C., Lee, J. T., ... & Swears, E. (2011, June). A large-scale benchmark dataset for event recognition in surveillance video. In Computer Vision and pattern Recognition (CVPR), 2011 IEEE Conference on

4.  Borji, A., & Itti, L. (2014, June). Human vs. computer in scene and object recognition. In Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on (pp. 113-120). IEEE

5.  Russakovsky, O., Li, L. J., & Fei-Fei, L. (2015). Best of both worlds; human-machine collaboration for object annotation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 2121-2131).

6. G Ye, Y Li, H Xu, D Liu, SF Chang (2015). Eventnet: A large scale structured concept library for complex event detection in video. MM'15 Proceedings of the 23rd ACM international conference on multimedia, pages 471-480

KEYWORDS: machine learning, POL, pattern of life, anomaly, track, multi-int fusion, A2AD, automated patterns, multi-modal data

AF171-051

TITLE: Elevated Situational Awareness through Discovery and Characterization of Composable Free Market Analytics

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 section 5.4.c.(8) of the solicitation and within the AF Component-specific instructions. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws. Please direct questions to the AF SBIR/STTR Contracting Officer, Ms. Gail Nyikon, gail.nyikon@us.af.mil.

OBJECTIVE: Develop a modular, secure solution for analytics across multiple domains to be autonomously represented, characterized, and discovered for bandwidth-optimized summarization, transfer, evaluation, and tasking.

DESCRIPTION: DoD Situational Awareness (SA) has evolved from simple translation of information between operational domains to SA transference involving mission effects and impacts, role-based relevancy metrics, big-data predictions for prioritization, and complex event analysis. However, while the raw information has become more interoperable, discoverable, and sharable, the analytics through which domains interpret them have not. Not only are analytical tasks not portable, policy-negotiable, or interoperable between domains, but it is not feasible to transfer large datasets between communities with disparate bandwidth and connectivity in order to perform domain-specific analysis.

Analytical tools have become siloes just as data repositories used to be. However, while database standards have emerged for characterization, metadata, transaction processing, queries, and discovery, analytics engines and analysis tools have not. Unfortunately, this limits the scope and power of many advanced features that could be enabled through the orchestration and management of analytics across multiple toolsets and mission domains.

Just as operational situations are ever evolving, analytical approaches that could increase mission effectiveness over time should create robust technologies that are empowered with support for evolution, re-use, and orchestration. One means to empower mission and domain transferable analytics is by enabling a shared, discoverable representation of their requirements, injection formats, results, and provenance. The resulting technology would provide an analytics "free market" system that enables higher level analysis, optimized usage of multiple best-of-breed analytic engines, and more effectively match analytical resources with mission needs by profiling and capturing metrics for available analytical tools.

Developing such capabilities requires addressing several challenges, including:

- Characterization and discovery of analytical operations that represent and span multiple domains and toolsets.
- Higher level analytics decision making, management, and orchestration controls.
- Assurance for analytics tasking and translation

The expected results of this effort include strategies and approaches to successfully implementing distributed analytics orchestration, creating an interoperable understanding of available data analytics resources, and enterprise-level methods for discovery and collaboration. It also results in a foundation that enables a level playing field by allowing analytical tools to be evaluated in a more 1-to-1 manner, a layer of abstraction for feedback and qualitative analysis to occur, and for best-of-breed analytics to continue to live in their own ideal element, while also contributing to a higher level analytics ecosphere.

PHASE I: Design a prototype system that leverages/fuses multiple analytics engines as part of flexible architecture of discovery, access, and cross-tasking. The representation and optimization of analytics resources should be formed with the above objective in mind.