Loren Data's SAM Daily™

FBO DAILY ISSUE OF OCTOBER 24, 2008 FBO #2524
SPECIAL NOTICE

A -- Request for Information For Information Fusion and Understanding Applied Research, Experimentation and Demonstration

Notice Date

10/22/2008
 

Notice Type

Special Notice
 

NAICS

541712 — Research and Development in the Physical, Engineering, and Life Sciences (except Biotechnology)
 

Contracting Office

Department of the Air Force, Air Force Materiel Command, AFRL - Rome Research Site, AFRL/Information Directorate, 26 Electronic Parkway, Rome, New York, 13441-4514
 

ZIP Code

13441-4514
 

Solicitation Number

Info_Fusion_And_Understanding_RFI
 

Response Due

11/30/2008
 

Point of Contact

Craig S. Anken,, Phone: 315-330-4833
 

E-Mail Address

Craig.Anken@rl.af.mil
 

Small Business Set-Aside

N/A
 

Description

Requests for Information (RFI) For Information Fusion and Understanding Applied Research, Experimentation, and Demonstration 1.0 GENERAL INFORMATION Document Type: Requests for Information Announcement Type: Initial announcement 2.0 FEDERAL AGENCY NAME Department of the Air Force, Air Force Materiel Command, AFRL - Rome Research Site, AFRL/Information Directorate, 26 Electronic Parkway, Rome, NY, 13441-4514 3.0 INTRODUCTION This publication constitutes a Request for Information (RFI) as defined in Federal Acquisition Regulation (FAR) 15.201(e), Exchanges with Industry before receipt of Proposals, Requests for Information". Respondents should note that no funding has been specifically reserved for this announcement. We are solicitating Requests for Information abstracts only. Do not submit a white paper or proposal at this time. 3.1 REQUEST FOR INFORMATION (RFI) This RFI announcement is an exploratory request to determine the existence of ideas or work in the Information Fusion and Understanding research area. The RFI seeks to obtain technical concepts, approaches, and the merit of the ideas or work in the subject field of research. Further, it seeks to obtain information about pricing, delivery, and other market information or capabilities for possible use in an upcoming Broad Agency Announcement (BAA). This announcement is not a request for proposals; therefore, responses to the RFI are not considered offers and cannot be accepted by the Government to form a binding contract. NO PROPRIETARY OR CLASSIFIED INFORMATION SHOULD BE INCLUDED IN THE RFI RESPONSE. Refer to section five (5) of this announcement for instructions on submitting an RFI abstract. 3.2 RFI ABSTRACTS To help guide the RFI process the following questions would be appropriate and should be considered when responding to this request. 1. What are you trying to do? 2. How is it done today? 3. What is new or innovative in your approach? 4. If you are successful, what difference will it make? 5. What are the risks and payoffs? 6. How much will it cost? How long will it take? 7. What are the midterm and final "exams" to check for success? Abstracts should contain in sufficient detail to enable the Government to determine whether the technical concept and/or capabilities should be reflected in a future BAA. 3.3 SYNOPSIS Submission of an abstract is voluntary and is not required to propose to subsequent Broad Agency Announcements (if any) on this topic. Respondents are advised that AFRL is under no obligation to provide feedback with respect to any information submitted under this RFI. 4.0 TECHNICAL BACKGROUND: The Air Force Research Laboratory's Information Directorate has seven key technology areas called Core Technical Competencies (CTCs). Information Fusion and Understanding (IFU) is one of the seven CTCs. The Information and Fusion CTC is defined as the continuous process that provides world-wide situational awareness in order to enable decision superiority. Activities include research, development, integration and testing of innovative technologies that enable continuous assessment of global conditions and events; that establish and maintain battlespace situational awareness, including an understanding of adversarial capabilities and intent; and that locate, identify, and track red, blue, green, and gray forces anywhere, anytime in near real-time. 4.1 SUB-IFU CTCs Within the IFU CTC there are a number of technology focus areas being addressed. This RFI is seeking ideas addressing four specific areas under the IFU umbrella. These include Reasoning & Learning, Link and Group Discovery, Advanced Analysis, and Space Situation Awareness. Descriptions of these areas follow. 4.1.1 Reasoning and Learning The goal of the Reasoning and Learning area is to research and develop technologies to enable computer systems that can reason, learn from experience, be told what to do, explain what they are doing, reflect on their experience, and respond robustly to surprise. Specific areas of interest under this RFI include but are not limited to: Learning and Reasoning in the Large The problems of information fusion, planning to gather information, inferring the intentions (plans, activities) of friends and enemies, and deciding how to behave to achieve goals are fundamental questions in AI research right now. In the very simplest of cases (when there are just a few states of the world), we know how to solve them exactly. But the problems that we are talking about below are on a huge scale: data is highly multi-modal and high volume; there are large numbers (unboundedly many, in fact) of individuals (trucks, units, sensors, missions, soldiers, aircraft, etc) and the time scales at which the activity recognition and planning must be done are long (potentially years, in the case of gathering intelligence about terrorist networks). So, the crucial technical emphasis has to be on solving huge problem instances by finding decompositions of the problem or special cases that can be solved efficiently. It may be that after applying certain decompositions, it will no longer be able to get the optimal answer. Our goal is to get the best answers we can under potentially severe time constraints. "Learning and Reasoning in the Large" refers to large amounts of data, large underlying world domain and large time scales. Reasoning and data mining A significant opportunity in gathering, extracting, integrating and fusing all of this related information is the ability to then perform various types of reasoning or data mining over the data to make new inferences. For example, in previous work, some initial techniques were developed for exploiting telephone books to identify buildings in imagery. This is done by using constraint satisfaction techniques to determine the possible labels for buildings in an image. However, to make these techniques work on real-world data actually requires the combination of the various steps. First, we need to extract the locations of the buildings in an image, which could be done using infrared data. Next, we need to know the names of the streets that the buildings are located on, which might be extracted from maps. Finally, we can then do the constraint reasoning to determine the mapping between the phone book entries and the buildings in the image. Similarly, we would also like to have a general set of data mining techniques that can identify patterns or relationships in the data. More generally, we need to develop a library of these techniques that can be applied over different data sources and in different situations. Temporal Entity Databases Recent advances in information extraction have made it possible to extract and collect facts about named entities, such as companies, persons, and locations. This presents an opportunity for automating the collection of information. However, current technology is often insufficient for compiling this information into a coherent "story". The main problem in accurately aggregating and integrating information from multiple sources is the broad range of inferences required. For instance, entity resolution is typically required when collecting facts about entities from multiple information sources. Entity resolution refers to the process of determining whether or not two facts refer to the same entity or not. For example, one fact may refer to "John Smith" the "Chief Marketing Officer" of "Acme Corp." while another fact may refer to "Mr. J. Smith" the "CMO" of "Acme Products, Inc.". Entity resolution involves determining whether these two facts refer to the same individual and company. Machine learning and inference techniques have recently been developed so that rules do not need to be hand coded for entity-resolution systems. However, the entity resolution problem is compounded when information changes over time, such as when John Smith gets promoted to CEO, or Acme Corporation is acquired by another company. Existing machine learning techniques will not work in such cases. In general, research is needed on methods that intertwine learning and reasoning. The focus of this topic is designing and developing an approach for aggregating information about entities over time. The system must be able to actually resolve references to the same entity, even when that information is changing over time. In particular, new approaches must be developed where temporal reasoning can be accommodated as part of the learning process. Automatically Harvesting Information about Networks Social networks are often used to model, or map, relationships between entities (such as people, companies, as well as more complex phenomenon such as disease transmission or airline routes. To operate, these analysis techniques require that the data be represented as a graph -- a set of nodes tied by one or more specific types of interdependency. In many cases, the data required to create the social network is not explicitly available, but must be acquired by extracting and integrating data from a host of different sources. This topic explores automated techniques for learning, or inferring, the structure of a network based on information that is available from multiple sources. Challenges include • Information Extraction: extracting relevant information from structured, semi structured and unstructured sources • Data Normalization and Cleaning: converting facts harvested from different sources into a source independent format • Entity Resolution: determining if two facts refer to the same entity • Information Integration: converting facts into nodes and links that can be incorporated into the social network, handling errors and inconsistencies, and deleting nodes and links based on new data. 4.1.2 Link and Group Discovery (LGD) This area is focused on techniques for analyzing large collections of data to discover valuable knowledge that may be present as hidden patterns or links among seemingly unrelated items. LGD will discover the hidden structure of organizations, relate groups, identify fraudulent behavior, model group activity, connect bits of information into patterns that can be evaluated and analyzed, learn what patterns discriminate between legitimate and suspicious behavior, and provide early detection of emerging threats. Specific areas of interest under this RFI include, but are not limited to the following areas. Dynamic Networks Over Time Most networks of entities and relationships change over time. Computers connect and disconnect from the internet. Web pages and hyperlinks appear and disappear from the worldwide web. People enter and leave social networks. Leaders emerge and then disappear. Many approaches have been developed to detect patterns in static snapshots of these networks, but discovering patterns in the dynamics of these networks can provide new understanding of how they evolve and new ways to categorize networks based on their dynamics. There is growing evidence that networks influence both individual and collective group activity. Networks play a key role in determining how information is disseminated and how organizations evolve (e.g., how new members are recruited and how they leave). Additionally, these networks have a significant impact on the choices an individual makes. These observations apply not only to social, but also to geospatial, information and technological networks. Despite growing interest in social network analysis (SNA) and related fields, our understanding of how these networks influence individuals and groups is still very limited. There is no general methodology for combining the effect of multiple, heterogeneous networks within a single model. In other words, there is a large knowledge gap in the mathematical methods and formalisms for modeling dynamic networks and their influence on the individual and group dynamics. Some of the outstanding problems in social network analysis are how to combine different sources of evidence, how to include patio-temporal data, and how the community structure changes over time. We seek to expand various aspects of social network modeling. Existing techniques emphasize static models. We need to be able to handle dynamic models that account for relationships/friendships drifting over time. We need to make it tractable to learn such models from data, even as the number of entities gets large. We seek to develop metrics to assess and identify change within and across networks and develop techniques to determine whether differences observed over time in networks are due to simply different samples from a distribution of links and nodes or due to changes over time in the underlying distribution of links and nodes. We would like to ultimately be able to understand the entire problem space well enough to predict/forecast change in existing networks. In addition, to accurately combine information from multiple sources, we need to address ways to manage the pedigree of the information. Abnormality Detection in Counter Terrorism Evidence Anomaly detection looks for entities whose characteristics, behaviors or connections to other individuals are in some way abnormal. Since we usually do not have a particular pattern of abnormality available to look for, we need to somehow derive from data what is normal and then identify who deviates most from that. Often it is the combination of several seemingly normal things that is abnormal. Much of the evidence we have today consists of networks of entities and their interrelationships, yet most approaches for detecting abnormalities in this evidence consider only the attributes of the entities, or focus on the presence or absence of specific relationships, without considering the context of the abnormality. For example, in the case of fraud, the perpetrator attempts to mimic normal behavior in order to avoid detection. That is, abnormality only makes sense in the context of what is normal. In the case of networks, normal behavior can be represented by recurring normative patterns in the structure of the network, and abnormalities can be defined as unexpected deviations to these normative patterns, not just unexpected deviations within the network as a whole. In addition to methods for detecting outlier attributes and relationships, we also need methods for detecting anomalies to the prevalent, normal behavior reflected in the transactions among entities in the domain of interest. Such methods will provide several benefits, including the description of abnormalities within the context of the normal behavior they defy and the ability to detect these abnormalities without prior training in possible abnormalities; thus, providing the ability to detect previously unforeseen abnormalities. Combining these methods with the above methods on detecting patterns in dynamic networks will provide the ability to detect anomalies in both the structure and the dynamics of networks. Single anomalies are often the result of noise, data mistakes, or other one-time-only phenomena and do not have significant practical interest. Instead we may want to find a collection of records that are each anomalous but have some pattern of self-similarity that make them indicative of a larger event. For example, a single strange case in the hospital emergency department is probably nothing more than that, but several of these across a city are indicative of a disease outbreak, or potential malicious activity. We need to extend traditional anomaly detection methods to link data to find linked groups of anomalous records. This will identify linked groups of individuals engaged in anomalous activity even if none are previously known to be dangerous or the method of attack they are pursuing is novel. We need to develop new efficient methods to scale this up to millions of links. Many times there are simply too many unique or rare events to make it useful to identify them all (not to mention having to determine how rare an event needs to be before it is reported). Sherlock Holmes stories typically see the detective observe a multitude of things that seem perfectly ordinary until he puts their significance together to solve the crime. So, we need be to recognize situations of potential interest, and to tell a story that explains why the situations are of potential interest, without those situations having been predefined in detail. What may be needed is some model of relevancy. An abnormality that can be explained away through a simple non-threatening story should probably not be reported. Conversely, an abnormality that cannot be easily explained with a non-threatening story, or that conjure up threatening or opportunistic stories should be reported (e.g., you come home and find a human finger, or $50,000, on the kitchen floor). So, we should be able to recognize situations of potential interest, and to tell a story that explains why the situations are of potential interest, without those situations having been predefined in detail. 4.1.3 Advanced Analysis The objective of the Advanced Analysis (A2) area is to lead the development of an integrated, service-based, interactive environment of tacit collaboration. The current model for information analysis is limited since it consists of client/server based systems supporting individual repository/database access and queries. The need for more complete informational awareness reveals the flaws of this system. The diversity of analytical tasks and approaches means that difficulty of solving this problem is in the understanding of user tasks, how to assist them in performing the tasks and retaining their analytical context. This process must be improved by understanding the value of available information, assessing the user needs/requirements and tracking user activity. To accomplish this there needs to be a better understanding of the three key components of an information system - information, user and actions. This understanding must be based on correlation of four aspects of analysis: relevance, novelty, credibility and persuasiveness. As an example, intelligence analysis is a complex and intensive process that requires analysts to make use of large data sets. The analyst must capture, identify and understand higher level knowledge that is constructed from that data. This is done in order to formulate, evaluate, and defend hypothesis against all other possible explanations. Adding a weak collaborative aspect to this process may only serve to increase the complexity of the problem. Instead this automated collaboration must provide relevant and/or novel information sharing. It must be tacit collaboration. With these issues in mind, the goal of the A2 area is to focus on the needs of the information analyst by effectively modeling their tasks, processes and datasets. This will enable the development of an integrated "end-to-end" machine assistant that can facilitate information sharing with an automated real time discovery and delivery of relevant information. Such a capability will allow users to individually and collaboratively produce more deeply reasoned analytical results in less time. Users will increase their productivity with the automated parts of information gathering allowing more time for their analysis. Additionally tacit collaboration will help them to discover the unknowns related to analytic context, identify meaningful collaborators and knowledge sources, produce a searchable corporate memory and provide enhanced support in performing risk analysis, developing option sets and creating deeper hypothesis and report analysis. Primary research areas of interest under A2 include, but are not limited to, Analysis Monitoring, Information Sharing, and Analysis Integration. The focus of Analysis Monitoring is modeling the needs and requirements of each user (which can vary with job operations) and how those requirements interact with the machines' understanding of the user needs. Information sharing research is developing techniques to help a user discover information that he/she does not even know is required and ensuring that the content is relevant and/or novel. Information Sharing also includes developing the means to accurately identify other users holding relevant knowledge to further collective intelligence capabilities. The focus of Analysis Integration is in the development of a system that leverages the best of breed capabilities while minimizing coordination costs and enhancing analytic tradecraft. The goal of the research is to address Air Force needs through the transition of effective user-focused collaboration technologies. These transitions will be applications and techniques that advance tacit collaboration within dedicated Air Force operations (e.g. Air Operations Centers (AOC), National Air and Space Intelligence Center (NASIC), Air Force Intelligence Analysis Agency (AFIAA)); Joint operations (e.g. US Central Command (CENTCOM)); coalition forces (e.g. North Atlantic Treaty Organization (NATO)) and with various government agencies (e.g. Homeland Security, Federal Aviation Administration (FAA) and Federal Emergency Management Agency (FEMA)) 4.1.4 Advanced INTEL for Space Situation Awareness Request information on new techniques for adversary modeling, anticipatory threat modeling, and early warning tripwire alerting of emerging counter-space adversary threats that could execute multiple coordinated attacks against multiple DoD (Department of Defense) and commercial space assets. Counter-space threats include non-kinetic (laser dazzling, RF Jamming, Cyber attack against satellite command & control ground stations, etc.) and kinetic: (Direct Assent Anti-Satellite Ballistic Missile (DA-ASAT) or Co-Orbital ASAT kinetic kill vehicle (KKV) in Low Earth Orbits (LEO), Medium Earth Orbits (MEO), Geostationary Orbits (GEO)) that could be executed by multiple adversaries through simultaneous, coordinated attacks against multiple satellites in the constellation (eg, could be DA-ASAT attack against Satellite A by country A teamed with country B who is concurrently attacking Satellite B by Laser dazzling & RF Jamming). Request information on techniques for space asset attack assessment which indicate the probability and type of space asset attacks, which coalition (air, ground, maritime, cyber) assets are vulnerable, the vulnerability periods, and confidence levels associated with the assessment. Dynamic adversary modeling and alerting techniques are needed to provide Joint Air and Space Operations Center (JSpOC) operators with advanced warning so that there is sufficient time for Defensive Counterspace actions to be able to be executed successfully. Request information on automated rule learning algorithms to adapt adversary patterns to evolving threats and changing adversary tactics, and tune detection and false alarm rates to optimize indications and warning (I&W) tools for JSpOC operators. Request information on adaptive computational approaches to fuse temporal I&W information from diverse ISR and foundational INTEL sources for timely situation awareness and allow analysts speculative what-if reasoning (i.e., what are expected to observe given a threat) that addresses uncertainty in the I&W data stream and incomplete ISR/INTEL data sets. Request information on new techniques to model the in-theater impact of space asset degradation effects that concern Joint Force Component Commanders including linkage of space asset degradation to coalition (air, ground, maritime, and cyber) operations to ensure unhindered integrated Command and Control (C2) of coalition forces. 5.0 REQUEST FOR INFORMATION (RFI) ABSTRACTS 5.1 CONTENT All abstracts shall state that they are submitted in response to this announcement and identify the IFU sub-CTC(s) to which the response is applicable. RFI responses shall include the company name, address and the title, telephone number, mail and e-mail addresses of the point of contact having the authority and knowledge to discuss the RFI submission. The abstracts should state the specific problem area, the technology proposed, the approach, and the potential advantage to the Air Force. A rough order of magnitude for the cost and a proposed duration of the effort should also be part of the RFI. NO PROPRIETARY OR CLASSIFIED INFORMATION SHOULD BE INCLUDED IN THE RFI RESPONSE. 5.2 SPECIAL CONSIDERATIONS Multiple abstracts within the purview of this RFI announcement may be submitted by each responder. 5.3 SUBMISSION RFI abstract due date is November 30, 2008. 5.4 FORMAT The abstracts will be formatted as follows: Section A: Title, Period of Performance, Estimated Cost, Name/Address of Company, Technical and Contracting Points of Contact (phone, fax, and email)(This section is Not included in the page count.); Section B: Task Objective and Approach; and Section C: Technical Summary and Proposed Deliverables. The abstracts shall be limited to 8 pages. All abstracts shall be double spaced with a font no smaller than 12 pitches. In addition, respondents are requested to provide their Commercial and Government Entity (CAGE) number, their Dun & Bradstreet (D&B) Data Universal Numbering System (DUNS) number, a fax number, an e-mail address, and reference this RFI with their submission. All responses to this announcement must be addressed to the POC's, as discussed in Section 6.0 of this announcement. Respondents are required to submit at least one electronic copy to the Government technical POC in Microsoft Office Word. 6.0 AGENCY CONTACTS Verification of government receipt or questions of a technical nature can also be directed to the cognizant TPOC. Primary TPOC Craig S. Anken Telephone: (315) 330-4833 Craig.Anken@rl.af.mil Link & Group Discovery TPOC Advanced Analysis TPOC Deborah Cerino Nancy Roberts Telephone: (315) 330- 1445 Telephone: (315) 330-3566 Email:rrs.ifu.lgd@rl.af.mil Email:rrs.ifu.a2@rl.af.mil Reasoning and Learning TPOC Space SA TPOC Roger Dziegiel Andrew Jeselson Telephone: (315) 330- 2185 Telephone: (315) 330-2411 Email:rrs.ifu.rl@rl.af.mil Email:rrs.ifu.ssa@rl.af.mil Questions of a contractual/business nature shall be directed to the cognizant contracting officer, as specified below: Lynn White Telephone (315) 330-4996 Email: Lynn.White@rl.af.mil
 

Web Link

FedBizOpps Complete View
(https://www.fbo.gov/?s=opportunity&mode=form&id=8a6685d8e783fdfeeb7c3fc9ea7c2b29&tab=core&_cview=1)
 

Record

SN01695077-W 20081024/081022215040-8a6685d8e783fdfeeb7c3fc9ea7c2b29 (fbodaily.com)
 

Source

FedBizOpps Link to This Notice
(may not be valid after Archive Date)

---

| FSG Index  |  This Issue's Index  |  Today's FBO Daily Index Page |