Agency: Department of Defense | Branch: Air Force | Program: SBIR | Phase: Phase I | Award Amount: 150.00K | Year: 2015
ABSTRACT:Establishing timely, accurate, and comprehensive Space Situation Awareness (SSA) requires analytic methods that put into context the information drawn from multiple data sources. A vital feature of an SSA system is the capability to predict, and understand, space-based activities that might threaten our satellites. The Activity Learning and Inferencing for Space Situational Awareness (ALISSA) program will address this important operational need by developing, evaluating, and transitioning an innovative capability to learn models of normal satellite operation, and detect operationally relevant deviations from satellite normalcy, built on the synergistic combination of machine learning and multiple hypothesis reasoning methods. In the Phase I program we will: (1) develop an algorithmic approach and overall framework for the problem of improving SSA by learning normal behavior, detecting changes, and reasoning over the implications of those changes; (2) conduct a series of focused investigations and evaluations of algorithms for normalcy learning, anomaly detection, event reasoning; and (3) evaluate the implications of the proposed algorithms on the overall SSA system. The results of the Phase I program will demonstrate the feasibility and promise of the systems concept to be realized in Phase II.BENEFIT:Activity Learning and Inferencing for Space Situational Awareness (ALISSA) offers significant potential benefits to the US Government and commercial satellite applications. Learning normal behaviors, and detecting and reasoning about anomalies, are vital to ensure the physical safety of commercial and military satellites (e.g., through collision avoidance), in addition to enabling the development of accurate space situational awareness. Our work in ALISSA will ensure that further government investment in space situational awareness (SSA) is optimized by maximizing the ability to provide detection and understanding of anomalous satellite events. The current most likely paths for transition are SSA for military organizations (e.g., Joint Space Operations Center (JSpOC) Mission System (JMS), NASIC, NRO, and AFSOPS) and commercial satellite owner/operators.
Agency: Department of Defense | Branch: Missile Defense Agency | Program: STTR | Phase: Phase II | Award Amount: 993.90K | Year: 2015
Boston Fusion, together with our teammate Syracuse University, propose a program of research and development, Uncertainty Characterization Using Copulas (UC)2, that will result in a parametric framework based on the statistical theory of copulas for modeling uncertainties for the problem of object classification. (UC)2 will produce a mathematical framework, founded on rigorous theoretical analysis, along with algorithms and software that will accurately characterize uncertainties associated with different sensor outputs. (UC)2 will naturally lead to (1) a better understanding of both the performance and the limits of the underlying fusion architecture; (2) new and enhanced fusion algorithms with optimal or near-optimal performance under realistic operating conditions; and (3) recommendations for sensor or feature selection under system resource constraints. In Phase I, we developed the initial framework and demonstrated the feasibility of (UC)2 on real data. In Phase II, we will demonstrate the performance of the proposed approach via a high-fidelity testbed environment, reflecting the operational system requirements. Approved for Public Release, 15-MDA-8303 (1 July 15)
Agency: Department of Defense | Branch: Missile Defense Agency | Program: STTR | Phase: Phase I | Award Amount: 99.97K | Year: 2014
Boston Fusion, together with our teammate Syracuse University, propose a program of research and development, namely Uncertainty Characterization Using Copulas, which will lead to a parametric framework based on the statistical theory of copulas for modeling uncertainties in a centralized fusion architecture for the problem of target classification for ballistic missile defense applications. This program of research and development will produce a mathematical framework, founded on a rigorous theoretical analysis, which will result in accurate characterization of the uncertainties associated with different sensor outputs. This framework will naturally lead to: (1) a better understanding of both the performance and the limits of the underlying fusion architecture; (2) new ways for developing fusion algorithms with optimal or near-optimal performance; and (3) recommendations for sensor or feature selection under system resource constraints. In Phase I, we will develop a prototype system design based on a challenge problem to validate the feasibility of the proposed statistical modeling approach. The resulting modeling framework will not only help in characterizing the performance of the fusion architecture, but will also enable future development of new fusion algorithms with enhanced performance in Phase II. Approved for Public Release 14-MDA-7663 (8 January 14)
Agency: Department of Defense | Branch: Missile Defense Agency | Program: SBIR | Phase: Phase I | Award Amount: 99.97K | Year: 2016
Combining information from disparate sensors can lead to better situational awareness and improved inference performance; unfortunately, traditional multi-sensor fusion cannot capture complex dependencies among different objects in a scene, nor can it exploit context to further boost performance. Integrating context information within a fusion architecture to reason cohesively about scenes of interest has tremendous promise for refining the decision space, thereby improving decision accuracy, robustness, and efficiency. The Deep Inference and Fusion Framework Utilizing Supporting Evidence (DIFFUSE) program will produce a mathematical framework, founded on rigorous probabilistic analysis and learning theory, which will result in accurate modeling of information across different targets in the scene and context-dependent high-level semantic representation of target labels. In Phase I we will: (1) develop a probabilistic modeling and learning approach to model multi-sensor data and context-dependent target label correlations; (2) develop inference algorithms based on fused multi-sensor data and context for multi-target classification; and (3) demonstrate the implications of the proposed algorithms under various target classification scenarios. The results of the Phase I program will demonstrate the feasibility and promise of the DIFFUSE system concept to be realized in Phase II. Approved for Public Release 16-MDA-8620 (1 April 16)
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase I | Award Amount: 79.99K | Year: 2015
In addition to organic (i.e., military) sensors, non-organic (i.e., non-government) sensors are ubiquitous, but largely untapped. These sensors capture data that address critical information needs such as detecting emerging activities or pinpointing a targets location. What is needed is a novel sensor discovery and exploitation framework to harvest intelligence from opportunistic sensors and allow analysts to incorporate new sensor data and collaborate both with the system and other analystscorrelating new data with historical intelligence archives, persisting and sharing frame-level annotations, and refining sensor selection for changing needs. In response, we will design and develop an adaptable, plug-and-play system, SEE-DATA, to identify opportunistic sensors and extract features to characterize and exploit the data in real-time. In Phase I, we will: Develop methods to identify opportunistic sensors for information needs of users and algorithms, and update sensor selection for evolving operations. Design a system concept for opportunistic sensor identification and exploitation to provide plug-and-play, cloud-compatible tool integration for characterizing and exploiting real-time sensor data. Capture user requirements for specifying information needs, structuring analysis, and annotating data for search and collaboration. Demonstrate opportunistic sensor identification and exploitation using data from multiple sensors and sensor types to show the validity of our approach.