An exponential increase in data sources and sensors provide today’s intelligence analysts and war fighters with a significantly larger, much more enhanced scope of operation.
However, with these new capabilities comes the ever-more pressing need to combine the increasing number of diverse and dispersed sources of data into information and knowledge for a comprehensive picture of the operational area. This training provides a detailed understanding of the cutting-edge and most commonly used technologies for multisource and multisensor data fusion. You will learn learn how to analyze, combine and make sense of large volumes of structured and unstructured data from disparate sources, such as physical sensors, operational transactions, human intelligence, news, blogs, and social networking sites. Specific questions to be answered:
- How do requirements for low vs. high-level fusion differ?
- What are the current trends and state-of-the-art in the data fusion area?
- What are the drivers for data fusion in urban operations, homeland security, missile defense, cyber warfare, air, space and maritime surveillance, process control, and health and status estimation?
- How are net-centric warfare and effects-based operations driving the developing of new tools and techniques?
- How are unstructured textual data handled for sentiment and social network analyses?
The course will provide solutions to “level 2” fusion (aka high-level situation assessment) requirements, including an analysis of the various available cognitive and agent-based multi-sensor data fusion architectures. However, solutions for lower-level fusion problems will also be addressed, including Kalman and particle filtering for multi-target tracking. Additionally, this training will also provide guidelines for using various models and techniques to deal with higher level problems associated with decision making in complex, uncertain environments.
Examples and demos will be drawn from a broad range of critical operational scenarios – from urban operations, to anti-terrorism, air operations, missile defense, and platform/system health monitoring. Available software tools will be discussed, and participants will engage in analyses of several example military scenarios, including building appropriate Bayesian belief networks for assessing enemy situations and developing appropriate response recommendations.
Attendees receive comprehensive slides, texts, CDs and software tools to take away for future references. The content of these tutorials are drawn heavily from books by in-house experts, especially two recent ones, namely, “High-Level Data Fusion” and “Foundations of Decision Making Agent: Logic, Modality and Probability”. Each attendee will receive a complimentary copy of the fusion book.
This seminar is intended for scientists, engineers, software professionals and technical managers in defense and other industries who need to develop solutions to difficult data fusion and decision-making problems.
Subrata Das with contributions from other team members. Each attendee will receive a complimentary copy of Dr. Das’s most recent book, “High-Level Data Fusion”.
Lesson 1: Multi-Sensor Data Fusion – Key Issues in Multi-Sensor Data Fusion, Low vs. High Level Fusion, Sensor Types and Characteristics, Impact of Sensor Types on Fusion System Design, Data Fusion and Decision Making, DoD and Service Initiatives.
Lesson 2: Architectures for Multi-Sensor Data Fusion and Decision Making – Joint Directorate of Laboratories (JDL) Architecture, Observe-Orient-Decide-Act (OODA) Loop, Situational Awareness vs. Situation Assessment, Cognitive Architectures, Rasmussen’s Hierarchy of Human Information Processing, Domino/Envelope Framework for Decision Making
Lesson 3: Multi-Sensor Data Fusion Application Domains – Conventional Warfare, Operations Other than War, Military Operations in Urban Terrains (MOUT), Counter-Bioterrorism and Other Anti-Terrorism Applications, Theater Missile Defense, Air Operations Center (AOC) Operations, Effect-based Operations (EBO), System Status and Healthy Monitoring, Example DoD Fusion Systems and Programs.
Lesson 4: Foundational Technologies for Multi-Sensor Data Fusion – Theory of Probability and Statistics, Statistical Distributions, Conjugate Distributions for Bayesian Inference, Monte Carlo Techniques, Syntax and Semantics of Propositional, First-Order, and Model Epistemic Logics, Bayesian Belief Networks, Resolution Theorem Proving for Classical/Non-Classical Logics, Approximate Inferencing via Particle Filtering, Intelligent Agents
Lesson 5: Software Tools for Multi-Sensor Data Fusion – iDAS for decision aiding, aText for text analytics, 5th Generation Application Development Environment, Bayesian Belief Network Engine, Argumentation Engine, SAS, MATLAB
Lesson 6: Techniques for Handling Uncertainty – Bayesian Probability, Possibility Theory and Fuzzy Logic, Dempster-Shafer Theory of Belief Functions, Certainty Factor, Transferable Belief Model, Handling of Confidence.
Lesson 7: Level 1 and Level 2 Fusion – Gating and Data Association, Single and Multi Target Tracking, Interacting Motion Models, Kalman Filtering for Level 1 Fusion, Unit Aggregation via Spatiotemporal Clustering, Static and Dynamic Bayesian Belief Networks for Situation Assessment, Follow-On Threat Assessment and Course-of-Action Generation, Sensitivity Analysis and Collection Management, Agent-Based Information Fusion
Lesson 8: Decision Making in Uncertain Environment – Expected Utility Theory, Rule-Based Expert Systems, Influence Diagrams, Symbolic Argumentation and Aggregation, Measurement of Experts’ Consensus.
Lesson 9: Temporal Modeling for Multi-Sensor Data Fusion – State Space Model, Hidden Markov Model, Dynamic Belief Networks, Rao-Blackwellised Filtering, Extended and Unscented Kalman Filtering.
Lesson 10: Unstructured Data Handling – Supervised and unsupervised text classification techniques, Natural language processing for parsing and stemming, Information extraction and structuring.
Lesson 11: Measuring Performance – Hit Rate, False Alarms, ROC Curve, etc., Subjective Evaluation, Cramer-Rao Lower Bound.
Lesson 12: Network Centric Warfare and Distributed Fusion – Publish and Subscribe Architecture, Pedigree Meta-Data Handling, Distributed Multi-Agent Fusion, Shared Situational Awareness, Distributed Sensor and Resource Management, Sense and Respond Logistics.
Lesson 13: Key Directions for Future Multi-Sensor Data Fusion – Data Mining/Machine Learning, Handling Unstructured Text Data, Knowledge Acquisition, Human Role in Data Fusion Process, Visualization.