Command and Control
Modeling for Synthetic
Battlespaces:
Flexible Group Behavior
Randall W. Hill, Jr.
Jonathan Gratch
USC Information Sciences Institute
ASTT Interim Progress Review
May 24, 1999
Agenda
Synthetic Forces Problem
Program Hypotheses
Technologies and R&D
Significant Results & Expected Results
Technology Transition Products & Efforts
Problem Areas
Programmatic Issues
Synthetic Forces Problem
Problem
Need cost-effective C2 modeling
Replace / augment human controllers with automated C2
Represent a wide range of organizations and situations
Need realistic C2 behavior
C2 models must make believable decisions
The outcomes of C2 operations need to be credible
Project Goals
Develop autonomous command forces
Act autonomously for days at a time
Reduce load on human operators
Behave in human-like manner
Produce realistic training environment
Perform C3I functions
Reduce the number of human operators
Create realistic organizational interactions
Program Hypotheses
Hypotheses
 Flexible behavior requires the ability to handle
situation interrupts
Flexible group behavior requires:
Understanding behavior of groups of entities
Planning a mission for groups against groups
Executing a mission in a coordinated manner
Hypotheses
Flexible group behavior interleaves the
processes of situation assessment, planning,
execution, and plan repair
 Coordinated group behavior requires a theory of
multi-agent interaction
Technologies and R&D
Technologies
Continuous Planning
Depends on understanding evolving situations
Implements planning as a dynamic process
Achieve goals despite unplanned events
Collaborative Planning
Coordinate group behavior
Requires understanding behavior of other groups
Reason about organizational constraints
Technologies
Situation Awareness
Current situation
Need a consolidated picture
Requires situation assessment at multiple echelons
Future situation
Integrate planning with future sensing requirements
Formulate Priority Intelligence Requirements (PIR)
Mission Capabilities
Army Aviation Deep Attack
Battalion command agent
Company command agents
CSS command agent
AH64 Apache Rotary Wing Aircraft
Suppression of Enemy Air Defense (SEAD) by
indirect fire (partially implemented)
Intelligence assets (partially implemented)
Battalion Deep Attack
MLRS
SLAR
SEAD
HA
HA
CSS
FARP
FLOT
C2 Architecture
Situation Report
(understanding)
Operations Order
(plan)
Situation Report
(understanding)
Battalion
Commander
Operations
Order
(plan)
….
Company A
Commander
Company X
Commander
Operations Order
(plan)
Situation Report
(understanding)
Company A
Company X
Pilot
Pilot
Pilot
Helicopter
Helicopter
Helicopter
Actions
Percepts
….
Pilot
Pilot
Pilot
Helicopter
Helicopter
Helicopter
ModSAF
Actions
Percepts
Architecture
 Planner
Implements continuous planning capabilities
 Plan manager
Augments collaborative planning with organizational
reasoning and Military Decision Making Process
 Time Manager
Manages temporal constraints
 Domain Theory
Maintains plan management and tactical knowledge
 Situation Assessment
Fuses sensors, reports, and expectations
Generates and updates current world view
C2 Entity Architecture
Plan
Manager
Management
Theory
(domain independent)
Tactical
Domain Theory
Planner
(General Purpose
Reasoner)
Management
Plans
Tactical Plans
World Model
Facts, inferences
Situation Assessment
Situation Reports, Sensing
Synthetic Battlespace
Expectations
OPORDER
Other Communications
Technologies and R&D:
Continuous Planning
Continuous Planning
Plan generation
 Sketch basic structure via decomposition
 Fill in details with causal-link planning
Plan execution
 Explicitly initiate and terminate tasks
 Initiate tasks whose preconditions unify with the current world
 Terminate tasks whose effects unify with the current world
Plan Repair
 Recognize situation interrupt
 Repair plan by adding, retracting tasks
What are Plans?
Hierarchically ordered sequences of tasks
Plans capture assumptions
Column movement assumes enemy contact unlikely
Plans capture task dependencies
Move_to_Holding_Area results in unit being at the HA,
(precondition to moving to the Battle_Position)
OPFOR and Co must be at the Engage_area simultaneously
Plan Generation Example
World Model
Attack(A, Enemy)
at(A,FARP)
at(Enemy,EA)
Destroyed(Enemy)
Destroyed(Enemy)
...
init
Move(A,BP)
at(A,FARP)
at(Enemy,EA)
at(A,BP)
Engage(A,Enemy)
at(A,BP)
Destroyed(Enemy)
Battalion Tactical
Plans
Co
Deep Attack
Move
Move
Engage
Co
Deep Attack
Return
Move
Move
Engage
Return
Company B plan
Move
Move
Move
Move
Company A plan
Move
Move
OPFOR Plan
Move
FARP
Operations
CSS
plan
Situation Interrupts Happen!
Current World
Attack(A, Enemy)
at(A,FARP)
at(Enemy,EA)
destroyed(Enemy)
destroyed(Enemy)
active(A)
Engage(A,Enemy)
Start of OP
Move(A,BP)
at(A,FARP)
at(A,BP)
active(A)
at(A,BP)
active(A)
ADA
Attack
destroyed(Enemy)
Reacting to Situation Interrupt
Situations evolve unexpectedly
Goals change, actions fail, intelligence incorrect
Determine whether plan affected
Invalidate assumptions?
Violate dependency constraints?
Repair plan as needed
Retract tasks invalidated by change
Add new tasks
Re-compute dependencies
Technologies and R&D:
Collaborative Planning
Collaborative Planning
Represent plans of others
Extend plan network to include others’ plans
Detect interactions among plans
Same as with “normal” plan monitoring
Apply planning modulators:
Organizational roles
What others need to know
Phase of the planning
Stance of the planner wrt phase and role
Plan Interaction Example
Move(A,BP)
at(A,BP)
at(A,FAA)
Engage(A,Y)
at(A,BP)
Dead(Y)
at(gas,FAA)
Attack Helicopter Company Plan
Move(CSS,HQ)
at(gas,FAA)
at(gas,HQ)
at(CSS,FAA) at(CSS,HQ)
resupplied(HQ)
Combat Service Support Plan
Planning Stances
Authoritative
Order subordinate to alter his plans
Deferential
Change my plans to de-conflict with superior
Helpful
Help peer to resolve conflicts in plan
Self-serving
Adversarial
Try to introduce conflict in other agent’s plan
Elaboration: Being Helpful
Planning issues
Propose doing activities that facilitate others’ plans
Avoid introducing threats into others’ plans
Communication Issues
Collaboration protocols: propose, accept, counter
Relevance reasoning
Which of my tasks would others want to know
• e.g. “Honey, I’m going to the market”
Elaboration: Self-serving
Planning issues
Notice things that others might do for me
Ignore threats I introduce into other’s plans
Unless that keeps them from doing things for me
Communication Issues
Deception
e.g. Someone might not help me if the knew what I was
really planning
Plan Management
Must model when to use different stances
Involves organizational issues
Where do I fit in the organization
Stances may need to change over time
During COA Analysis, adopt an adversarial stance towards ones
own plans
Must model how stances influence planning
How do we alter COA generation
C2 Entity Architecture
Plan
Manager
Management
Theory
(domain independent)
Tactical
Domain Theory
Planner
(General Purpose
Reasoner)
Management
Plans
Tactical Plans
World Model
Facts, inferences
Situation Assessment
Situation Reports, Sensing
Synthetic Battlespace
Expectations
OPORDER
Other Communications
When to Use a Stance
Model the collaborative planning process
Includes management tasks that modulate the
generation of tactical plans
Tasks refer to specific tactical plans
Specify preconditions on changing stance
Includes knowledge of one’s organizational role
Planner constructs management plans
Use same mechanisms as tactical planning
Management Plan Example
 Explicitly model the Military Decision Making Process
Tasks
COA
Development
COA
Analysis
Stances
Authoritative towards subordinates
Deferential towards superiors
Adversarial towards OPFOR
Authoritative towards OPFOR
Adversarial towards self (war gaming)
Implementing Stances
Implemented as search control on planner
Plan manager
Takes executing management tasks
Generates search control recommendations
Example: Deferential Stance
When giving orders to subordinates
Indicate subset of plan is fixed (defer to this)
Indicate rest of plan is flexible
Plan manager enforces these restrictions
Interaction Example
Deferential towards
Move(A,BP)
at(A,FAA)
Make CSS Planner
defer to Company
A’s Plan
at(A,BP)
at(gas,FAA)
Move(CSS,HQ)
at(gas,FAA)
at(gas,HQ)
at(CSS,FAA) at(CSS,HQ)
Combat Service Support Plan
C2 Entity Architecture
Plan
Manager
Management
Theory
(domain independent)
Tactical
Domain Theory
Planner
(General Purpose
Reasoner)
Management
Plans
Tactical Plans
World Model
Facts, inferences
Situation Assessment
Situation Reports, Sensing
Synthetic Battlespace
Expectations
OPORDER
Other Communications
Technologies and R&D:
Situation Awareness
Situation Awareness
Planner needs a consolidated picture of
the current situation in the battlespace
Determines which goals and tasks are achievable
Influences the choice of strategies and actions
Allows the detection of imminent plan failure
Enables re-planning
Situation assessment produces a current
World Model
Monitor plans with respect to world model
Situation awareness = world model + plans/tasks
Situation Assessment
Performed at multiple echelons
Scouts performing reconnaissance of battlespace
C2 staff assimilates scouting and sensor reports
General process:
Identify entities
Classify groups of entities as units
Determine units’ functionality, capabilities, plans, intent
 Technical Issues
Pilot awareness and information overload
Situation assessment techniques
Pilot Situation Awareness
Synthetic worlds are information rich
100’s of other entities
Vehicle instruments
Terrain, weather, buildings, etc.
Communications (messages)
Amount of information will continue to increase ….
Perceive, understand, decide and act
Comprehend dynamic, complex situations
Decide what to do next
Do it!
Information Overload
Roots of the Problem
Naïve vision model
Entity-level resolution only
Unrealistic field of view (360o, 7 km radius)
Perceptual-Cognitive imbalance
Too much perceptual processing
Cognitive system needs inputs, but …
It also needs time to respond to world events
Approach
Create a focus of attention
Apply attention mechanisms to entity perception initially
Incorporate filters
Implement a zoom lens model (covert attention)
Stages of perceptual processing
Attention in different stages: preattentive & attentive
Control the focus of attention
Goal-driven
Stimulus-driven
Zoom Lens Model of Attention
(Eriksen & Yeh, 1985)
Attention limited in scope
Multi-resolution focus
Magnification inversely proportional to field of view
Low resolution
Large region, encompassing more objects, fewer details
Perceive groups of entities as a coherent whole
High resolution
Small region, fewer objects, more details
Perceive individual entities (e.g., tank, truck, soldier)
Low Resolution
Perceptual Grouping
K
Preattentive
Gestalt grouping
Involuntary
Proximity-based
Other features
Dynamic
Voluntary grouping
K
K
Group Features
Quantity and composition
Activity
Moving
Shooting
Location
Center-of-mass
Bounding-box
Geometric relationships wrt pilot
Slant-range, azimuth, etc.
High Resolution
Entity Features










Location (GCS)
Speed
Velocity
Orientation
Slant Range
Force
Object, Object Type
Vehicle Class
Function
Sense Name










Altitude
Angle Off
Target Aspect
Magnetic bearing
Heading
Status
Lateral Range
Lateral Separation
Closing Velocity
Vertical Separation
Control of Attention
Goal-driven control
Agent controls the focus / resolution of attention
Low resolution: Scouting groups of enemy; escorting group
High resolution: Search for air-defense entities; engage target
Sets filters that select entities for WM
Stimulus-driven control
Attention can be captured involuntarily by a visual event
Muzzle flash (luminance contrast, abrupt onset)
Sudden motion (abrupt onset)
Goal-driven Attention
Overwatch
Position
Land
Sea
Overwatch
Position
Transport Carrier
Rendezvous
Point
Escort task
• Orient on group
• Voluntary grouping
Escort Carrier
Stimulus-driven Attention
Low Resolution
High Resolution
Situation Awareness
at Higher Echelons
Command
Entity
Situation Reports
Command
Entity
Situation Reports
Command
Entity
Situation Reports
Situation Assessment
Identify entities
Fuse scouting reports
Classify groups of entities as units
Cluster entities into unit-sized groups
Classify units into functional types
Determine capabilities, plans, intent
Clustering and Classification
Bottom-up and top-down approach
Bottom-up clustering based on proximity
Identify a group of entities close to each other
Other useful features: color, orientation, speed
Top-down classification based on doctrine
Threat templates
Issues: which template, partial matching
Bottom-up Clustering
Hierarchical Clustering
Partitioning starting at the top until a satisfactory level (e.g.
individual units)
Robust Clustering
Nearest-neighbor using center of mass
Works well for hierarchical clustering
Requires a parameter of minimal distance
Density-based clustering
Works well on different shapes of patterns
No parameter is required (or can be learned)
Top-Down Classification
Classification and prediction
Classification based on threat templates
Doctrine of situations, actions, formation and capacities
Matching clustered units with templates for classification
Partial matching to predict the location of missing units
Encoding threat templates
Encoding spatial information for symbolic processing
kD-tree to encode spatial relationships
Adding possible actions to nodes (units)
Future Situation Awareness
Model how tactical intelligence influences
planning
Future situation: knowledge goals
What will I need to know for this plan to work?
Establish Priority Intelligence Requirements (PIR)
What commander needs to know about opposing force
Drives the placement of sensors and observation posts
Constrains the pace of plan execution
 Rarely addressed in current C2 models
Intelligence Critical for
Realistic C2
Close interplay between intelligence and
COA Development
Intelligence guides COA development
COA development drives intelligence needs
Intelligence availability constrains actions
• Some COA must be abandoned if one can’t gather
adequate intelligence
Intelligence Critical for
Realistic C2
Intelligence imposes temporal constraints
When can a satellite observe?
How long to insert surveillance (LRSU)?
How long before I must commit to COA?
Intelligence critical for
realistic C2
Intelligence collection must be focused
Commanders must:
Prioritize their intelligence needs
Understand higher-level intelligence priorities
Provide intelligence guidance to subordinates
e.g. Simulation Information Filtering Tool
[Stone et. al]
Brigade Planning
(simplified)
 Attack 2nd echelon
tank division (TD)
AA
Lincoln
Identify Engagement Area
(EA Pad)
Should canalize OPFOR and
restrict movement
Identify launch time
Require 2-hour notice
EA Pad
Brigade PIR
AA
Lincoln
When will TD leave AA Lincoln?
Verifies enemy intent
When will TD reach PL Echo?
Satisfies the need for 2-hour notice
Further verifies enemy intent
Location of PL Echo driven by PIR
2hrs
EA Pad
Intelligence Plan
Assembly Area
LRSU
Trigger attack: TD
2hrs from EA Pad
SLAR
Monitor movement
from assembly area
EA Pad
Final Brigade Plan
Decision
Point
H
H-10 H-8
H+2
H+3
SLAR monitor AA
Insert LRSU
LRSU monitor PL Echo
Deep Attack
Execute Arrive
Mission at EA
Break
Contact
Automating PIR
Identify PIR in my own plans
Find preconditions, assumptions, and triggering conditions
that are dependent on OPFOR behavior
Extract PIR from higher echelon orders
Specialize as appropriate for my areas of operation
Derive tasks for satisfying PIR
Sensor placement
Ensure consistency of augmented plans
Identifying PIR
Examine COA dependencies on OPFOR
e.g. Precondition of engaging:
OPFOR will-be-at EA Pad at time H+2
Look for dependencies that:
Are not under my direct control
Are uncertain
 Implemented with PIR recognition schema:
Abstract rules that scan plans and assert PIR
Some domain-independent, some domain-specific
Interpreting Higher Level
Guidance
Need to convert into PIR at my echelon
e.g. Brigade’s PIR:
When will lead regiment reach forward defense
becomes Battalion PIR
When will lead battalion of lead regiment reach fwd def
Implemented by specialization rules
Encode doctrinal and terrain relationships
Deriving Sensor Plans
Implemented via tactical planning mechanism
PIR represented as “knowledge goals”
Domain theory augmented with sensing tasks
Sensing tasks achieve knowledge goals
Tasks encode maneuver / temporal dependencies
Planning process fills in details
Sensing tasks added to achieve knowledge goals
• e.g. Observe TD activity near PL_ECHO
Other tasks added to satisfy maneuver dependencies
• e.g. Use UH-60 to insert LRSU near PL_ECHO
Ensuring Consistency
Implemented via tactical planning mechanism
If PIR goals cannot be satisfied, COA is invalid
or
Use unsatisfied PIR to request external assets
Sensing plans constrain timing of events
If temporal constraints inconsistent, COA is invalid
Significant Results
Significant Results
 Continuous planning paradigm works well for
modeling C2 behavior in the joint synthetic battlespaces
Dynamic planning, monitoring, and execution
Handles situation interrupts in test cases
 Collaborative planning is made possible by adding a
few extensions to a general purpose planner
 A model of perceptual attention and situation
awareness implemented in RWA-Soar pilot
 Developed a technique for deriving Priority
Intelligence Requirements with planner
Significant Results (2)
Publications
Continuous Planning and Collaboration for Command and
Control in Joint Synthetic Battlespaces, CGF&BR ‘99
Deriving Priority Intelligence Requirements for Synthetic
Command Entities, CGF&BR ‘99
Modeling Perceptual Attention in Virtual Humans, CGF&BR ‘99
Perceptual Grouping and Visual Attention in a Multi-agent World,
Agents ‘99
Scope of Task Coverage
ATKHB Attack Mission
Achieve
Tactical
Disposition
Reduce
Enemy
Posture
Achieve
Culminating
Task
1-4-1305 (Section 6.1.2):
Integrate fire support
Attack (METL task)
Consolidate
1-4-1206:
Continuous Tasks
Achieve Readiness
1-4-1101: Personnel (S1) planning (C2)
1-4-1201: Intelligence (S2) planning (C2)
1-4-1301: Operations (S3) planning (C2)
1-4-1401: Logistics (S4) planning (C2)
1-4-1302: Establish and maintain
tactical operations center (C2)
1-4-1305: Coordinate maneuver with
CSS and rear ops (C2)
--------------------------------------------------1-2-0320: Provide supply support (CSS)
1-2-7723: Perform maintenance (CSS)
1-2-7728: Process ammo and fuel (CSS)
1-4-1103: Replacement operations (CSS)
1-4-1402: Coordinate supply/equip. (CSS)
1-4-1405: Plan and coordinate transport
assets (CSS)
Achieve Physical Posture
1-3:0001: Plan and organize move (Mnv)
1-2-0101: Move to and occupy assembly
area (Mnv)
1-4-1306: Establish and maintain tactical
command post (C2)
1-2-7726: Conduct FARP operations (CSS)
Legend
Implemented
Partially implemented
Desire to implement
Less relevant
1-2-xxxx: Establish satellite comm. (C2)
1-2-xxx0: Establish ground comm (C2)
1-2-7509: Establish voice comm (C2)
11-5-0104: Establish FM radio (C2)
1-4-1001: Perform C2 operations (C2)
1-4-1303: Control tactical operations (C2)
-----------------------------------------------------------1-4-1202: Implement security measures (Int)
1-4-1203: Process intelligence information (Int)
1-4-1311: Liaison operations (Int)
-----------------------------------------------------------1-4-1105: Provide admin services (CSS)
1-2-7708: Provide food support (CSS)
1-2-7710: Operate field mess (CSS)
1-2-7720: Establish med support (CSS)
1-2-7721: Conduct med activities (CSS)
1-4-1102: Perform strength management (CSS)
1-4-1104: Conduct casualty reporting (CSS)
1-4-1308: Direct army airspace C2 (CSS)
1-4-1310: Civil-military operations (CSS)
1-4-1403: Monitor equipment readiness (CSS)
1-4-1406: Provide logistic services (CSS)
Expected Results
 Detailed evaluation of planner
 Empirical
 Analytical
 Extended model of situation awareness at entity and C2
levels
 Attention, hierarchical clustering, classification, fusion
 Extended model of collaboration
 Abstract technical description of planner
 Journal articles and conference papers
Measures of Success
Collective Measure
 Ability of a group of entities (RWA Battalion) to achieve mission
objectives in scenarios containing a wide range of situation interrupts
Individual Measures
 Scalability: size of groups that can act autonomously
 Flexibility: classes of situation interrupts handled by group behavior
 Types of multi-agent reasoning integrated into framework
 i.e., collaborative, adversarial, temporal, ...
 Breadth and depth of domain knowledge
 e.g., # of tasks, echelon levels, functional categories (battlefield operating systems)
Evaluation
Empirical
Developed scenario generator, logging function
Will collect data from scenarios run in batch mode
Encode additional domain knowledge (WARSIM?)
Evaluate scalability
Analytical
Develop abstract description of planner
Complexity measures for scalability
Analyze properties of collaborative planner -- can it be decoupled from Soar-CFOR implementation?
Technology Transition
Efforts
Formulated concept for C2 in NASM
Demo at JPMR in February ‘99
Presented 3 papers at CGF&BR, May ‘99
Perceptual attention, C2 Modeling, PIR
JSIMS/ASTT workshop, May ‘99
WARSIM commonality (POC’s: Milks & Karr)
ONESAF?
Problem Areas
Focused Efforts Required
Not yet addressing role of learning
Need good evaluation
Scalability, robustness, efficiency, …
Programmatic Issues
Schedule
Milestone 4: 12/98
Design Review 2
Approach to learning improved group models
Approach to temporal planning
Schedule (2)
Milestone 5: 9/99 (revise to 12/99?)
Technology POP Demonstration 3
RWA Attack Battalion
Demonstrate advanced group understanding
Demonstrate more advanced group planning
• Temporal planning
• Group understanding: plan recognition
Demonstrate advanced group execution
• Commander utilizes teamwork model (scaled down)
Demonstrate group learning
• Improve group models through experience
Deliver software and domain independent descriptions of new
capabilities
Demonstration
Demonstration Scenario
 Attack Helicopter Battalion (AH-64)
Battalion Commander
3 Helicopter Companies
Company Commanders
Apache Pilots
1 Combat Service Support Commander
 Deep Attack Mission Scenario
Companies move from Assembly Area to Holding Area
Situation interrupt: unexpected enemy forces in Holding Area
Dynamically re-plan and execute mission