Notice Type: Presolicitation Notice
Subject: Computational Methods for Decision Making
Classification Code: A - Research & Development
Solicitation Number: ONRBAA14-010
The purpose of this applied research (6.2) topic is to identify, understand, and resolve key issues, develop and mature algorithms and methods; determine and demonstrate performance of algorithms, methods, techniques, and strategies for automated computational methods and information systems that support decision making. The algorithms, methods, techniques, and strategies must support autonomous information processing systems that can successfully and securely execute a variety of missions in complex environments while exploiting multiple sources of sensor and open domain data. The program will pursue a wide variety of approaches that enable automated systems to, within the context of a mission, automatically analyze multiple sources of data supporting interpretation of the data; combine data and generate interpretations from multiple data sources to provide understanding of the battle space, provide management of sensor and other resources to maintain and improve the battle space picture, and to enable and build high performance software systems that are defect free and trustworthy to implement these algorithms, methods, techniques, and strategies. Background:
The development of automated decision systems provides a number of significant technical challenges including processing, interpreting and developing decisions using diverse data sources, multiple modalities, unstructured data, and large volumes of data with varying latencies while compressing the time-line for arriving at a decision. Additional challenges occur when we consider that the computing hardware and software environment must protect the data, function correctly, while simultaneously providing security and trustworthiness. These issues will likely be exacerbated in practical implementations that are distributed and employ networks. The quality of the decisions developed by the system is dependent upon the quality of the underlying data and it's relation to the mission. The quality of the decisions is also impacted by the security of the data and the computing hardware that also impact the trustworthiness of the decision.
The processing and interpretation of data requires understanding of the context of the mission. The context of a mission enables a set of hypotheses, expressed as models, to provide a viewpoint that enables a system to determine data that is relevant and important to producing a picture of the battle space (situational awareness). Missions also provide a context in which the inherent uncertainty and imprecision of the data can be identified and understood with respect to subsequent processing steps involving data and inferences over the data. The presence of multiple data sources introduces additional technical issues associated with aligning the data prior to fusion, schemes for fusion, and assessing, understanding, and controlling the effects arising from incompleteness, imprecision, and contradiction in the data upon inferences and decisions.
A key issue for Naval Forces in developing situational awareness is to understand what is known, how well it is known, what is unknown and to provide strategies to determine new data that should be collected to maintain or improve situational awareness. In turn this requires capabilities to perform optimization of scarce resources in order to support a mission. If the process is to be automated and timely relative to a mission then algorithms must be implemented that can sense, interpret, reason and successfully act in an open world with uncertain, incomplete, imprecise, and contradictory data. These information processing systems should also be capable of autonomously validating their hypotheses and derived models, as well as autonomously developing new hypotheses and models as warranted. Achieving operational capabilities such as Persistent Pervasive Tactical Surveillance or Adaptive Tasking, Collection, Processing, Exploitation, and Dissemination could be straightforward if information processing systems were capable of understanding the information and quality of information that they need to produce and maintain a model of the world given its hypotheses and mission goals. This applied research topic aims to develop knowledge and understanding of key technologies that will enable rapid, accurate decision making by autonomous processes in complex, time varying highly dynamic environments that are probed with heterogeneous sensors and supported by open source data. The applied research results should lead to understanding, computational theory, algorithms, techniques, strategies, and practical implementations providing security and trustworthiness that enable information processing systems and decision aids to adapt in an open, complex, and uncertain environment over an arbitrary set of missions.
1) Resource Optimization The objectives of the Resource Optimization thrust are the development and application of mathematically rigorous techniques (e.g., mathematical optimization) that provide optimal or provably near-optimal solutions to resource-allocation problems. These techniques will serve as the basis of automated decision aids in support of naval planning and execution. Within the Resource Optimization there are currently two themes: Maritime Mission Planning; and Sensor Management and Allocation. Maritime Mission Planning seeks capabilities that improve power projection and achieve far better utilization of expensive
2) Automated Image Understanding The objective of the Automated Image Understanding thrust is to develop efficient computational methods based on principled approaches that advance the understanding of issues governing performance that are needed to support system engineering. Image understanding is a broad field that requires advances along many directions. Under this thrust, we plan to address the following issues: (a) developing principled methods for fusion of multiple imaging modalities based on the physics of image formation, leading to image enhancement and improved recognition capabilities; (b) methods for integrating images from multiple platforms for improved object recognition, scene modeling, and meaningful change detection; (c) developing methods for indexing images based on semantic content for storage and retrieval; (d) detection and tracking of objects on water or in urban areas and inferring the threat level they may pose; including real-time detection of partially occluded objects in urban clutter; (e) developing robust recognition methods that integrate low-level image processing with high-level knowledge, or generative and discriminative models. This will also require investigating best representations, or hybrids of representations, for description and recognition of objects and activities. Furthermore, we want to extend recent advances in reasoning with image/video that make recognition of objects and activities more robust. Domain knowledge plays a critically important role in reasoning; hence an additional area of applied research would be methods for building visual knowledge bases. This also involves investigation of suitable representations for high-level semantic knowledge, which may come in various forms including contextual information, background models, shape and appearance and behavior information, relationships among entities.
3) Information Integration The objectives of the Information Integration thrust is to develop efficient, theoretically sound, and consistent algorithms for organization, fusion of high-dimensional data sources, interpretation of the fused product, determination of the value of data and information, and to investigate their application and potential to support naval applications. The Information Integration thrust is currently developing, maturing and assessing algorithms that organize high-dimensional datasets of interest to Naval Operations. Current efforts include applied research focused on image, video, structured database, social or complex networks, hyper-spectral, multispectral, acoustic, sensor array, and other structured datasets as well as assessing the potential value of missing information. Issues that are to be addressed under this thrust include (a)methods that lead to structuring unstructured datasets in an organized and meaningful way are desirable and should facilitate more efficient and accurate processing tasks including data matching or alignment, data merging, data search, outlier detection, learning and classification, query response, reasoning and decision making; (b) automated algorithms that fuse high-dimensional datasets that are comprised of uncertain, incomplete, imprecise, and contradictory data for the purpose of recognizing and classifying features, objects, entities, activities, patterns of interest, and relationships; (c) assess and understand the quality of the resulting fused battle space picture and its impact on decision making.
4) Cyber Security The objective of the Cyber Security thrust is to develop a software development environment that enhances the robustness and security properties of the resulting codes, while minimizing penalties to code performance and overhead. Currently flaws in software are a major contributor to the vulnerability of cyber systems. Most if not all of these vulnerabilities originate from improper software implementations. Identified flaws that lead to improper implementations include, and are not limited to, buffer overflow, stack and heap overflow, dangling pointers, input data format violation, race conditions, etc.. Methods for software implementation still lead to these deficiencies. Significant investment has been made to address this issue through techniques that seek to provide formal or other forms of software verification. However, complementary efforts to verification, that lead to understanding of techniques that enhance the development and generation of robust secure code is under-explored. Alternatively automated methods that capture and utilize work flow, thought/design-decision, and documentation during software coding that also are aware of software implementation issues could address this need. Only rarely are all of the details for the implementation of software is specified in advance. Currently programmers make instantaneous detailed design decisions during software coding. These instantaneous decisions (and assumptions) have far reaching effects, and they are often forgotten and lost. A tool that captures and documents these design decisions (and hence assumptions) automatically as coding is in progress can significantly enhance maintainability, robustness, and security of codes. The availability these tools also provide an opportunity to provide feedback to programmers to improve the correctness of their product and enhance productivity and efficiency.
Most Popular Stories
- More Hispanic Voters May Not Mean More Clout
- 2016 Camaro Shrinks, Moves to Caddy Platform
- Eric Garcia Appointed as Revenue Chief
- Apple Pay Debuts With Few Issues
- Government: 500 Million Records Stolen in 12 Months
- Stocks Subdued After Gains Earlier in Week
- Pistorius Gets 5-year Sentence in Shooting Death
- Mom Makes Toys R Us Pull 'Breaking Bad' Dolls
- Volatility No Reason to Bail on Stock Market
- Cuba Deploys More Medicos in Ebola Fight