100507 1300 Bidder Conf Brief.ppt

Advanced Mathematics for DoD Battlefield Challenges BAA Bidders Conference 14 May 2010

This Briefing will be presented at the bidders conference, and is provided in advance for the convenience of potential bidders

Disclaimer This briefing is intended to supplement the Advanced Mathematics for DoD Battlefield Challenges BAA

Nothing in this briefing shall be construed to modify or alter the text of the BAA

Questions and answers will be recorded and released on the FedBizOps page for this BAA

Agenda: Advanced Mathematics for DoD Battlefield Challenges Bidders Conference 1200-1230: Registration

1315-1330: Proposal Mechanics (Mr Steve Biernesser)

1330-1345: Break

1345-Finish*: Q&A (Ms Robin Quinlan, Mr Tristan Nguyun, Mr Doug Cochran)

*NLT 1700

Where we were: One example of the Problem - Target Tracking

A A B C D B C D

Constant turnover of target designation limits warfighter ability to make decisions

Where we are:

The Low/Slow Air and Ground Picture

Low/Slow Air and Ground Problem [1of 2] Even more complicated than High/Fast

“Disadvantaged user” with limited bandwidth

Low/Slow Air and Ground Problem [2 of 2]

Data is incomplete, with errors, and needs to be optimally processed to give unambiguous tracks and target state vectors, including time

How do we “triage” data appropriately and provide actionable information to the warfighter?

Background Our capability to collect data well exceeds our ability to process it

Background (cont)

Background (cont) Some fields of mathematics have already contributed significantly, including linear algebra, systems theory, differential equations, probability, and statistics

DDR&E wants to bring a larger body of mathematical knowledge to bear on the challenges

Detecting poorly defined anomalies in large data sets is an important capability, particularly when adversaries are difficult to distinguish from non-combatants

Common Tactical Picture Challenge

Essential Components of the Challenge

Key Mathematical Challenge Areas Data/information fusion

Information Extraction / Data Interpretation

Inference and Prediction

Information Presentation / Visualization

Information Management and Architectures

Uncertainty Quantification

Challenge Problem 1 Creating the elements of a common tactical picture in the Low/slow air and ground domain

Use the data sets provided

Extract features to enable a clear and continuous composite track of an enemy target from multiple information sources, moving in space and time, in a densely cluttered environment

Locate, identify and predict the track of an enemy target amongst the clutter, using the data streams provided by multiple information and sensor types

Hold identification of the track for the lifetime of the track, using mathematical techniques that combine multiple sources of data, with discrimination from other closely situated friend and neutral entities

Challenge Problem 1 (cont)

Track Identification – Friend, neutral, hostile, or unknown

Track Continuity (i.e., maintains ID of track) – Average amount of time an object’s track is reported without its track identifier (e.g. friend, neutral, hostile, or unknown) changing

Track Ambiguity (low spurious track ratio, ideally one track per target) – Average value of uncorrelated tracks per object

Positional Accuracy

Velocity Accuracy

ID Ambiguity (percent confidence)

Proposed projects should include a demonstration milestone NLT Dec 2011

Challenge Problem 2 In advance of data/information fusion to form a common tactical picture, the reduction of unwanted local data is imperative to remove irrelevant information which can deleteriously affect a fusion procedure

This is also essential in eliminating the data deluge which results from indiscriminate collection of data through a multitude of sensors

As a concrete example, we presently have hundreds of sensors in Afghanistan that provide high fidelity but a very limited field of view. They see the world through a “soda straw”

If we could optimally place and task these sensors to look at specific targets of interest at specific times, thereby collecting the data we need rather than data that happens to be available, use of battlefield communication and computational resources would be optimized and the relevance of information provided to warfighters would be enhanced

Challenge Problem 2 (cont) Although optimal sensor placement is an active research discipline which predominantly hinges upon specific missions, this BAA focuses on sensor placement and management in direct support of Challenge Problem 1 that is described above

Advanced mathematical techniques that can be used to simultaneously tackle both Challenge Problems in a unified framework will be given precedence

Hence the challenge problem is to use available data to cue high fidelity/narrow field of view sensors in order to maximize sensor coverage and effectiveness on an identified activity in the data set provided

Challenge Problem 2 (cont)

Timeliness – Time required to identify an event of interest

Cueing Accuracy - distance between cue location and the actual target;

Sensor economy - minimal use of sensor resources to achieve a desired level of performance);

Ability to identify an event of interest from multiple or decentralized data streams.

Proposed projects should include a demonstration milestone NLT Dec 2011

Data Resources Two sources of Government Furnished Equipment (GFE) data available to support this BAA

Proposal Mechanics