1. MURI Annual Review Meeting Randy Moses November 3, 2008
Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation Summary
MURI Annual Review Meeting
Randy Moses
November 3, 2008

2. Today’s Presentations
Optimal, Robust Information Fusion in Uncertain Environments (Willsky)
Graphical Models for Resource-Constrained Hypothesis Testing and Multi-Modal Data Fusion (Castañón, Fisher, Hero)
Sparse Reconstruction and Feature Extraction (Çetin, Ertin, Karl, Moses, and Potter)
Algorithms and Bounds for Networked Sensor Resource Management (Castañón, Fisher, Hero)
Tools for Analyzing Shapes of Curves and Surfaces (Fisher, Srivastava)
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

3. Key Accomplishments
Scalable, flexible information fusion using graphical models
Multiscale structures
Learning structure for inference
Unified sensor signal processing approaches
Incorporate sensor and object physics
Feature estimates and feature uncertainties
Statistical shape theory
Scalable, information-based sensor management
Low-complexity algorithms with guaranteed performance
Information metrics for sensor management
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

4. Synergistic Interactions
Graphical Models as a synergy catalyst
Front-end Signal Processing
Coordinated, well-integrated research program
Unified approach to address statistical fusion and performance assessment
Multiple modalities; multiple operating modes
Information-based resource management
Information cost metrics and performance bounds
Physical scattering models
Extended visits:
Cetin to MIT/BU summers; Hero sabbatical at MIT
Student research co-advising and thesis committees
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

5. Transitions Strong, Fruitful Connection to AFRL
Student summer interns
Collaborative data collections
Research collaborations
Transitional research topics (e.g. ATR Center)
Technology transfer to industry via DoD programs
DARPA Visibuilding, ATIF, POSSE
AFRL Gotcha
SBIRs with AF, MDA
Student internships
Several at AFRL; also GD-AIS, NG, BAE-AIT
Non-DoD technology transitions
Medical imaging and shape estimation
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

6. What’s next – Signal Processing
Closing the gap between parametric and nonparametric approaches
Bayesian estimation perspective for sparse reconstruction
Pixels versus parameters
Multi-sensor and multi-modal performance
Connect shape and point features; connect across modalities
Learning, Adaptation, and Context
Efficient algorithms and approaches for closing the loop
Shape estimation features for ATE
Graphical Models for Studying Configurations of Shapes
Sensor Processing to achieve multiple objectives
E.g. moving and stationary targets
Update these!
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

7. What’s next – Information Fusion
More on learning behavioral models and multi-target tracking
More on learning tractable models for fusion and discrimination
E.g., introducing hidden variables to capture hidden causes
Inference over object interactions
Learning graphical dependency structures to infer interaction states.
How to do this efficiently
Integrated  learning of embedded graphical models
Joint clustering/classification and manifold learning
Distributed topological inference
Update these!
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

8. What’s next – Sensor Management
More on scalable algorithms
More on performance bounds
Integration of graphical fusion models and performance estimates into algorithms
Algorithms for unknown target classes
Integrated SM and front end processing for imaging
SM driven by info theoretic imaging criteria
Incorporating inverse scattering models
Image priors e.g., sparsity, smoothness, shape
Update these!
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

9. Expected End States Decision-directed front-end signal processing
A unified understanding and mechanism for uncertainty characterization monitoring, incorporation of context and higher-level fusion information, and adaptation.
Robust sensor processing algorithms
Combine robustness of nonparametric techniques with accuracy and tractability of parametric techniques for sensor signal processing
New methods for building and using graphical models
Scalable and tractable for large, militarily-relevant problems
Designed to enable fusion across levels and sensor resource management
Graphical Models for tracking and behavior extraction
Incorporate learning from experts and non-conventional information sources
General theory for information-seeking resource management
Heterogeneous sensors
Inference metrics and predictions drive sensing actions
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation

10. Questions or Comments?
MURI: Integrated Fusion, Performance Prediction, and Sensor Management for Automatic Target Exploitation