1 Failures in C2 Technology Why command and control has stagnated Doug Dyer April 03 Joint Vision 2010 provides an operationally based template for the.

1

Failures in C2 TechnologyWhy command and control has stagnated

Doug Dyer April 03

Joint Vision 2010 provides an operationally based template for the evolution the Armed Forces for a challenging and uncertain future. … This vision of future warfighting embodies the improved intelligence and command and control available in the information age...”

1997: Joint Vision 2010

2

Overview

Why haven’t past technology efforts succeeded in improving C2?• How can current and future efforts (DJC2, JC2, SDE) do better?

Structured information• What it is• How to develop, integrate, extend, and sustain it• Reasons to believe we’ll succeed

Creating Structured Information with 5th Generation Applications• Bottom-up, inside-out, do-it-yourself • Applications: workflow, structured email, interfaces to problem solvers, data editing

• Applications as creators and reasons to sustain structured data

3

Previous programs correctly identified goals

Example: Joint Vision 2010The basis for thisframework is found in theimproved command,control, and intelligencewhich can be assured byinformation superiority…

… These transformations willbe so powerful that theybecome, in effect, newoperational concepts: Dominant maneuver Precision engagement Full-dimensional protection Focused logistics.

-- John M. Shalikashvili, Chairmanof the Joint Chiefs of Staff

4

Previous programs correctly identified goals

Example: Joint Vision 2010

From the Air Force 1998 TAP:

GLOBAL AWARENESS• Consistent Battlespace Knowledge• Precision Information• Global Information BaseGLOBAL INFORMATION EXCHANGE• Distributive Information Infrastructure• Universal Transaction Services• Assurance of Service• Global Connectivity to Aerospace ForcesDYNAMIC PLANNING/EXECUTION• Predictive Planning and Preemption• Integrated Force Management and Execution• Execution of Time Critical Missions/Real TimeSensor-to-Shooter Operations• Joint, Combined and Coalition Operations

All of these goals are appropriate, but we’ve

only made progress on the service infrastructure

5

Previous programs correctly identified goals

Example: Joint Vision 2010Information Architecture

A little optimistic about the Common Operational Picture

Otherwise, good architecture

“When combined with extensive coverage of the order of battle of the opposing forces, faster than real time simulation of potential enemy courses of action, and exceptional high capacity communications, military commanders in 2010 will see the three domains pulled ever closer together. The outcome of this process should be a command and control system that bases its decisions and management actions on a bedrock of accurate and common understanding and acts (reacts also, but the initial responses of commanders that may be the most important could be those that are proactive) to make quality and timely decisions. Further, the quality of the information allows us to better parse the decision making process. Simple Decisions are those for which a stimulus or event requires a specific response. These are easily automated. Contingencies can be considered in advance and Contingent Decisions are those that require a response from a known list and can be categorized through "if-then" logic. These are also quite automatable. On the other hand, Complex Decisions are those in which the decision-maker must create the list and these are extremely difficult, if not impossible, to automate. Automatable support (e.g., M&S) can nevertheless assist the decision-maker in complex decision-making.” --- EBRI paper

6

Previous programs correctly identified goals

Example: Joint Vision 2010Information Architecture

A little optimistic about the Common Operational Picture

Otherwise, good architecture

“When combined with extensive coverage of the order of battle of the opposing forces, faster than real time simulation of potential enemy courses of action, and exceptional high capacity communications, military commanders in 2010 will see the three domains [Battlespace Awareness, Decision-Making, and Battle Management]pulled ever closer together. The outcome of this process should be a command and control system that bases its decisions and management actions on a bedrock of accurate and common understanding and acts (reacts also, but the initial responses of commanders that may be the most important could be those that are proactive) to make quality and timely decisions. Further, the quality of the information allows us to better parse the decision making process. Simple Decisions are those for which a stimulus or event requires a specific response. These are easily automated. Contingencies can be considered in advance and Contingent Decisions are those that require a response from a known list and can be categorized through "if-then" logic. These are also quite automatable. On the other hand, Complex Decisions are those in which the decision-maker must create the list and these are extremely difficult, if not impossible, to automate. Automatable support (e.g., M&S) can nevertheless assist the decision-maker in complex decision-making.” --- EBRI paper

7

So, what’s the problem?

Answer #1: Failure to build an enduring structured data model

8

Structured information is the foundation for new C2 capability

Structured Data Model

Tailored Feeds Sentinels

Standard Presentation Formats

Rapid situation understanding

Assessment of coverage

Rules Simulations

Constraint checking

Case-based reasoning

Generative Planning

Dialog & machine learning

Web Services

Procedures

In general, any technology intended for decision support requires structured information: variables and values in context

Once you have a structured data model of the battlespace, you can stack up an impressive array of technologies resulting in new command and control capability

Effects: Reduced workloadBetter situation awarenessEasier coordinationBetter decisions faster Increased span of control

9

Joint Vision 2010 Plan and Situation Data Models

Again, JV2010 had the right idea… But JV2010 never

successfully implemented the models…

Without a structured data model, you can’t get smart algorithms

It makes sense to figure out why past attempts failed and how to

do it better in the future

No other C2 effort has

created an enduring

data model

10

So, what’s the problem?

Answer #2: Failure to define, create, and sustain smart algorithms

11

Smart algorithms are the keys to decision speed and qualityAssumed: future operations will be more complex

Structured Data Model

Tailored Feeds Sentinels

Standard Presentation Formats

Rapid situation understanding

Assessment of coverage

Rules Simulations

Constraint checking

Case-based reasoning

Generative Planning

Dialog & machine learning

Web Services

Procedures

Some might disagree based on past efforts, but machine-amplified brain-power has the best potential to make better decisions faster… for getting agile

How? Tons of ways…Creating information in common formatsIdentify missing informationTailoring information to people’s roleTimely alertsWorkflow and coordinationImplication calculationResource allocation and optimizationPlanning and goal satisfactionHypothesis generationPossible futuresProblem-solvingProcess improvementAdaptation of our knowledge base

12

A Joint Vision 2010 Functions and Algorithms

Again, JV2010 had the right idea… But even JV2010

demos couldn’t meet the vision

Without smart algorithms, you can’t get intelligent assistance from your computer

It makes sense to figure out why past attempts failed and how to

do it better in the future

No other C2 effort has

fieldedsmart,

large-scale automation

13

Defining Structured Information

14

“Structure” as defined by Webster’s Dictionary

Two defining characteristics…

A number of parts…• i.e., an enumerated list of things (known; not infinite)

… that are put together in a specific way

Example: An object in an object-oriented computer language• Has a number of attributes or variables associated with it• Is bound by a set of defined relationships• Interacts with other objects via a specific set of interfaces

15

Structured Information for Command and ControlDefining an “Information Element”

Context Variable Value Meta-Data

For OPLAN 3400, Operation Bullfrog, root branch, according to planning by Cmdr Newton for SEAL Team 4’s ingress plan…

The ingress resources required are:

2 Mark-V Assault Boats

As decided at 12:04:13 Z on 16-Mar-03 using a default rule which was apparently accepted by the user

Example:

An information element is an atomic unit of structured information It includes:

16

Structured InformationDefining an “Information Element”

Context Variable Value Meta-Data

For OPLAN 3400, Operation Bullfrog, root branch, according to planning by Cmdr Newton, for SEAL Team 4’s ingress plan…

The ingress resources required are:

2 Mark-V Assault Boats

As decided at 12:04:13 Z on 16-Mar-03 using a default rule which was apparently accepted by the user

Example:

An information element is an atomic unit of structured information It includes:

Plans and situations can be partially described by sets of variables and values

Context dictates which variables belong together

Meta-Data includes other useful information

17

A structured data model represents plans and situationsSDM: a set of information elements

Structured Data Model

Context Variable Value Meta-Data

C1C1C3C2C4C1C2C4C11C22C1C4…

V1V2V1V3V7V2V3V6V4V1V2V3…

Val AVal BVal AVal CVal DVal FVal XVal UVal TVal GVal EVal M…

Meta-data {a, b, d}Meta-data {l, k, c} Meta-data {q, j, v}Meta-data {t, w, i}Meta-data {o, p, r}Meta-data {g, h, g} Meta-data {k, u, u}Meta-data {z, p, d}Meta-data {a, v, e}Meta-data {p, c, y}Meta-data {i, y, d} Meta-data {s, w, a}…

The set of variables associated with any particular plan or situation are found by matching on context

18

How to build a structured data model

19

Building and Exploiting a Structured Data Model

Structured Data ModelA way to build it• Incentives• Resources• Approach• Technology

A way to introduce and integrate it

• Workflows• Briefings• Email• Planning

A way to update and improve it

• Dynamic state change• Process improvement

Requirements for it• Manual search for information• Decision support• Bandwidth limitations• Multi-level security

Neat things you can do with it• Format to speed understanding• Heuristics automate details• Decision logic to aid reasoning• Deep analysis for hard problems

20

Two ways to build a structured data model

Traditional approach: Top DownDevelopers use knowledge engineering to

learn the domainCreate models that cover the domain

Ontology Database schema XML definitions Frequent patches to get coverage, correct errors … ~3-18 months

Develop applications … ~ 6 months Integrate applications into users

processes... ~3 monthsAdapt applications as user processes

improve… ~3 monthsProcess ends when funding runs out

An alternative approach: Bottom UpDevelopers demonstrate tools, but users

extend examples and build new applicationsCreate forms for a specific purpose

Users pick the terms and define important relationships Forms define micro-ontology, namespace Database schema created automatically XML may be projected … ~1 day

Forms become distributed applications integrated by users and useful for

Workflow Data entry and viewing Planning and problem-solving … ~1 day to integrate using links

Users and developers adapt jointly Users can adapt the data model immediately Users can annotate requirements for AI and tailor

some rules Developers understand domain and requirements from

forms and annotations Developers may clean up data model

Process continues indefinitely because it’s cost-effective

21

Building and Exploiting a Structured Data Model Based on Forms

Active FormsA way to build it• Immediate payoff• Reduced resources• Bottom-up approach• Active Forms Technology

A way to introduce and integrate it

• Workflows• Briefings• Email• Planning• Users create forms to capture structured parts of these and get smart help for their effort

A way to update and improve it

• Edit state change once• Never duplicate effort• Improve as needed

Requirements for it• Manual search for information• Decision support• Bandwidth limitations• Multi-level security

Neat things you can do with it• Regular layout helps you find key info• Heuristic rules define defaults• Case-based reasoning based on experience• General procedures and web agents• Constraint checking• Machine learning from interaction

22

The World-Wide Web Analogy

Evidence that users can help technology scale

23

Technical Problem

We have a great information system (the web), but we have no large-scale system for smart automation… and none is coming

DARPA’s $80M investment in agents and the Semantic Web are useful but not sufficient because they are:

complex top-downslow to payoff

Source: Hobbes: http://www.zakon.org/robert/internet/timeline/

HTTPHTMLBrowsers

Number of Smart Agent Systems

1999: 50 2003: 30

Number of DAML Ontologies

1999: 112 2003: 112

24

We want this kind of growth curve for smart automation systems

We can achieve it with tools that enable anyone to create smart templates that can be stitched together to create complete systems

simple

bottom-up

immediate payoff

Technology Potential

And we’ll get composable micro-ontologies and structured data as a side-effect

25

5th Generation Applications

26

Bottom-up, inside-out, do-it-yourself

Bottom-up: don’t try to define a large model Depend on composition to get a larger model

Inside-out: expose the small model for use by othersDo-it-yourself: avoid the cost and scalability limitations of

knowledge engineering

Results: Faster application development Cheaper More scalable

27

Smart automation that’s “Web-simple”

HTTPHTMLBrowsers

Relational DatabaseXMLFormsRules and Case Depends

You can create a complete web site in a day…You should be able to create a smart workflow in a day too!

Users create these forms by specifying the variables, the appropriate widgets, and perhaps by enumerating values

28

Structure On-the-FlyBottom-up Micro-Ontology

Terms

Relationships between them

# If your destination is less than 300 miles away, # then you probably should just drive your own car rather than fly# If can't figure out how far away your destination is, don't suggest anythingset rule(Mode) { if {[destinationMilesLessThan 300]} {

set el(Mode.suggestedvalue) DrivePOV } elseif {[destinationDistanceUnknown]} {

reset el(Mode.suggestedvalue) } else {

set el(Mode.suggestedvalue) Fly }}

Every template defines a micro ontology that comes for free• layered context• namespace

But we do create “structure on-the-fly”

This semantic information might need cleanup

29

Structure On-the-FlyBottom-up Micro-Ontology

Terms

Relationships between them

# If your destination is less than 300 miles away, # then you probably should just drive your own car rather than fly# If can't figure out how far away your destination is, don't suggest anythingset rule(Mode) { if {[destinationMilesLessThan 300]} {

set el(Mode.suggestedvalue) DrivePOV } elseif {[destinationDistanceUnknown]} {

reset el(Mode.suggestedvalue) } else {

set el(Mode.suggestedvalue) Fly }}

Every template defines a micro ontology that comes for free• layered context• namespace

But we do create “structure on-the-fly”

This semantic information might need cleanup

30

Application categories

• Planning• Interfaces to external

problem-solvers• Information editing and

viewing

31

Application categories

• A structured form of email

32

5th Generation Applications Create and SustainStructured Data

Old riddle: “Which came first, the chicken or the egg?”

Answer for smart software: They must be co-developed

5th Generation Applications (e.g. distributed forms):Create structured information by capturing decisions, accepting inputs, and AIDepend on structured information from other sources, including other 5th Generation AppsAdd value to workflows and problem-solvingAre easily extended to cover new problems or to cover existing problems better

Viola! Mutual supportAll of these factors support the sustainability and extensibility of structured data and the family of smart algorithms needed for intelligent assistance and improved C2 capability