LogisticsAided Battle Command Planning

1
Logistics-Aided Battle Command Planning

• Presented by
• Christopher Matthews
• Chris.R.Matthews_at_us.army.mil
• U.S. Army RDECOM CERDEC
• Fort Monmouth, NJ

2
Talking Points

• What?
• Why?
• How?
• Commentary

3
Background

• Research, Development, Engineering Center (RDEC)
• RDECOM CERDEC Fort Monmouth, NJ
• Command Control Directorate
• Command Control (C2)
• Transitioning relevant, novel technologies to the
  warfighter faster
• Emphasis on C2 decision-aid software
• Development Team
• 2 Programmers
• 1 Pseudo-expert on tactical logistics

4
Logistics Command Control ATD (LogC2)

• Objective
• Integrate logistics into the maneuver planning
  process
• Faster OPTEMPO reduced logistics footprint
• Shorten Combat Service Support (CSS) planning
  times
• Products/capabilities
• Collaborative, cross-functional planning services
• Optimization tools for increased maneuver
  sustainment efficiency
• Adaptive, predictive consumption models and
  demand generation functionality
• Near real-time running-estimate decision aid
  software
• Dynamic re-planning
• Execution monitoring

5
Maneuver Sustainment Planner (MSP)
Goals

• Develop proof-of-concept prototype
• Integrate logistic-planning impacts into the
  maneuver planning process
• Develop a detailed logistics plan to support
  maneuver operations
• Model maneuver activities and generate expected
  logistics demand as a function of platform,
  posture, and optempo
• Showcase dynamic re-planning execution-monitorin
  g capability
• Evaluate Cougaar and agent benefits for C2
  planning

6
MSP Why Cougaar?

• Perfect fit for the program
• Maps elegantly to military force structures
• Tailored for the planning domain
• Essential, core planning capabilities included
• Crucial plumbing (asynch msging, threading, wf
  engine)
• Dynamic replanning execution monitoring
• Resource management components
• Good design
• Component-based architecture
• Loosely coupled cognitive modeling
• Logical Data Model (LDM)
• Rapid software development
• Amenable to incremental development
• Substantial Reuse
• Automated tools helper programs

7
MSP Overview
Maneuver Planning Domain

• Core Functionality
• Import maneuver plan elements
• Develop logistics plan (LogPlan)
• Demand generation modeling for Class III, V,
  IX
• Temporal asset utilization military units
  physical assets
• Utilize dynamic re-planning capability
• Provide logistic feasibility feedback loop

MSP
8
MSP Approach

• Functional analysis design
• Agent Enumeration
• Role/Relationship Analysis
• Plugin Enumeration
• Task Grammar
• Asset/Property Requirements Analysis
• Execution Monitoring and Dynamic Replanning
  Analysis
• External system interfaces
• Databases
• Maneuver Command Control (MC2) application
• MSP Plan Viewer (GUI)
• Cougaar Plugin development

9
MSP Constraints

• No control over the maneuver plan
• Maneuver Command Control (MC2) system
• Stove-piped system
• Inaccessible models methods
• Large, unwieldy, coarse XML plan data
• IMPACT Atypical Cougaar implementation
• Short-lived vs. 24x7
• IMPACT Demand generation
• CASCOM Equipment Usage Profiles (EUP)
• Identical to MC2
• Inability to dynamically add agents
• IMPACT A priori knowledge of UTO
• Time
• IMPACT Reduced functionality

10
MSP Developmental Items

• GUI Servlet
• Asset utilization
• Consumable information
• Distributed LogPlan and corresponding maneuver
  tasks
• UTO and tables of equipment (TOE)
• Planning ontology
• Task structure
• Asset classes
• Property groups
• Agent roles relationships

11
MSP Developmental Items

• Plugins
• Allocators
• Expanders
• Assessors
• Hybrids
• LDMs

12
MSP Results

• Objective
• Tightly integrate maneuver logistics planning
• Result
• primitive feedback to maneuver planning domain
• Maneuver impacts of LogPlan via swivel chair
• Risk Area
• Interoperability with maneuver application
• Time-intensive interface development required
• Assessment
• Moderately successful
• Can add functionality later

13
MSP Results (cont.)

• Objective
• Develop detailed LogPlan for maneuver sustainment
• Result
• Accurately modeled class III (fuel) demand
  generation
• Successfully demonstrated the collaborative
  problem-solving behavior of Cougaar agents to
  develop a LogPlan
• Considerable use of assumptions
• Risk Area
• Impact of modeled assumptions
• Generalized scoring functions
• Assessment
• Highly successful
• Assumptions can be removed w/ more detailed
  modeling or soldier-in-the-loop at critical
  decision points

14
MSP Results (cont.)

• Objective
• Incorporate dynamic replanning
  execution-monitoring functionality
• Result
• Dynamic planning demonstrated
• Simulated execution-monitoring real-time updates
• Dynamic replanning unfinished
• Risk Area
• Time
• Poor planning on our part (prioritization of SW
  tasks)
• Assessment
• Fail
• While execution-monitoring was included, dynamic
  replanning is more seductive

15
MSP Level of Effort (Cougaar)
16
MSP Agent Interaction
Allocator
Expander
Allocator
Assessor
LDM
Mvr Plan Task Generator
Maneuver Tasks
LDM
Allocator
Assessor
Expander
Retail Refuel Requests
Notional Logistics Wholesaler
Logistics Broker
Wholesale Refuel Requests
Allocator
Retail Refuel Requests
Forward Support Company
Wholesale Refuel Requests
Base Support Company
17
MSP Planning Thread
Expander
Bn 1-8 Infantry
Co C 1-8 Infantry
LDM
Allocator
Co A 1-8 Infantry
Maneuver Task Generator
Allocator
Co B 1-8 Infantry
Assessor
Expander
Logistics Broker
Notional Logistics Wholesaler
Abstract Asset
Allocator
Allocator
64 FSB BSC
64 FSB FSC
Allocator
18
MSP Task Generation Agent
Bn 1-8 Infantry
Maneuver Task
Expander
LDMSQL
TaskListening Role
Maneuver Task Allocator
Maneuver Task Generator
Unit Initialization
19
MSP Maneuver Units
Co A, Bn 1-8 Infantry
Subordinate Task
Subordinate Task Expander
LDMSQL
Subordinate Role
Logistics Task Allocator
Bn 1-8 Infantry
OrgRTData
UTO
OrgReport
20
MSP Planning Thread
Expander
Bn 1-8 Infantry
Co C 1-8 Infantry
LDM
Allocator
Co A 1-8 Infantry
Maneuver Task Generator
Allocator
Co B 1-8 Infantry
Assessor
Expander
Logistics Broker
Notional Logistics Wholesaler
Abstract Asset
Allocator
Allocator
64 FSB BSC
64 FSB FSC
Allocator
21
General Assessment

• Cons
• Large overhead
• Bandwidth intensive in tactical environment
• Steep learning curve
• Time-intensive source code reviews (the reuse
  issue)
• Impacts of architecture upgrades
• Scoring functions
• Difficult to assess (apples v. oranges)
• Difficult to accurately model
• Dynamic modification to agent society
• Some sparse documentation
• .DEF file definitions
• PropertyWriter, AssetWriter, MeasureWriter
• Complex data types

22
General Assessment (cont.)

• Pros
• Rapid software development
• Provides logical roadmap for application design
  development
• Ideal for military planning systems
• Domain independent
• Open source software
• Core planning information management
  functionality
• Well documented architecture developer guides
• Manageable application development
• Incrementally move from general and intricate
  cognitive models
• Configuration management easier
• Loosely coupled behavioral components
• Easy to incorporate real-time planning or
  running-estimate functionality
• Large, knowledgeable support community

23
Wrap-up

• Missed objectives
• 1st time around
• 2nd time around
• Still requires substantial functional analyses
  investment
• Reduces time-intensive coding
• Allows development to focus on this process the
  most important one
• Vision conventional wisdom
• Can share niche planning communities and services
• Repositories of reusable components
• User community temperament
• Fear
• Politics
• Success story
• Crude,limited prototype
• Cougaar experience subject matter expertise

24
Questions?
25
Backup Slides
26
MSP PlanViewer Snapshot
27
MSP PlanViewer Snapshot
28
MSP Force Structure Scenario
Force XXI UTO
3 BDE, 4ID
HHC 3 BDE, 1CD
64 FSB
1-12 IN
1-8 IN
1-68 AR
HHC
Co B
Co A
Co C
HHC
Co B
Co A
Co C
HHC
Co B
Co A
Co C
HHC
FSC
BSC
Medical
29
MSP Asset Class Hierarchy
Package org.cougaar.glm.ldm.asset
Package logc2.assets
All Assets in the logc2.assets package are
machine generated from the logc2props.def and
the logc2assets.def files. All ground
vehicle Assets will be of one of the types
depicted here.
This class is hand generated (the only one in the
package.) This is needed so That we can add
functionality later to all Assets w/o having to
mod lots of classes.
Truck
SelfPropelledGroudWeapon
FuelTruck
Carrier
Open to view LogC2 Vehicle List
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
MSP Roles Relationships
Superior
Provider
Customer
Subordinate
39
Superior
Provider
Subordinate
40
ALP Goals

• Technical Goals
• Distributed agent architecture research
• Distributed information management research
• Real-time information fusion research
• Functional Goals 
• Automated logistics plan generation
• Real-time logistics situation assessment
• End-to-end movement control
• End-to-end rapid supply

41
UltraLog Goals

• Expanding the ALP vision
• Military logistics domain
• Enhance the Cougaar framework
• Security trusted systems under information
  warfare attacks
• Scalability stability for large, distributed
  network of agents
• Robustness high state of survivability in
  chaotic environments
• System integration combining all of above to
  achieve desired systemic effects

42
Agent Development Framework

• Cognitive Agent Architecture (Cougaar)
• Darpa initiative
• Advanced Logistics Program (ALP) (FY96 FY01)
• UltraLog Program (FY01 FY04)
• Features
• Distributed, large-scale workflow engine
• Open source Java software
• Multi-Resolutional Logical Data Model (LDM)
• Built-in dynamic re-planning execution-monitorin
  g capability
• Asynchronous communication protocol
• Classic publish/subscribe blackboard
• Domain independent

43
Cougaar Agent basics
Agent
Blackboard (PLAN)
Publish
Subscribe


Plugins
Message Queue
44
Cougaar The Distributed Plan
Asset
Asset
Asset
Task
Task
Task
45
Cougaar - Plugins
LDM
Populating society with data from external
systems.
Allocator
Allocates tasks to other agents/assets.
Decomposes tasks into more manageable Subtasks.
Expander
Assessor
Evaluates the projected and actual results of
allocated tasks
Provides the web-based front end for viewing
activity within society.
GUI
46
Cougaar Re-planning Concept
Dynamic Re-planning Execution Monitoring
Self-Assessment
Task Allocations
Allocation Results Rescinds / Reallocations
Real-Time Data Fusion
Negotiate between agents to iteratively improve
plan Continually recheck assumptions
Constraints, Data, Policies, Metrics Updates
from data sources as time progresses
47
Cougaar Agent behavior
Agent Relationships
Superior
Agent Roles
Provider Customer
Subordinate Superior
Provider
Customer
Subordinate
48
Cougaar UltraLog
UltraLog Society
Cougaar
Domain Agent
Specific PlugIns
Generic PlugIns
Generic Agent

• Military specific processes
• Interfaces to military systems
• Specific to Logistics Domain

• Basic building blocks
• Easy to specialize
• Domain independent

49
Plan Element - Allocation
50
Plan Element - Expansion
Plan Element (Expansion)
Task-1
R
Reported AR
Workflow
Subtask-1.2
Subtask-1.1
51
Plan Element - Aggregation
Task-1
Task-2
Plan Element (Aggregation)
Plan Element (Aggregation)
Composition
Task-3
52
Cougaar Logical Data Model (LDM)

• Requirements
• Support over 6 million items
• Support efficient transport and distribution of
  objects
• Support modification and extension during
  execution
• Principles
• Based on the properties of objects and not what
  they are
• Represent all the properties of assets needed to
  reason about them
• Over a range of granularities
• Supporting their time-varying nature
• Refering to specialized properties of assets
• Implementation
• Use prototypes and delegation to reduce classes
  needed
• Prototype classes determine the required
  properties of all instances
• Related properties are collected in Property
  Groups
• Asset instances delegate properties to their
  Prototype instances

53
Cougaar LDM Asset
54
Cougaar LDM examples
LDM Classes
Asset Prototypes and Property Groups
5-Ton Truck ltprototypegt
VehiclePG ltinstancegt
NIIN 123456999 VehicleProp ContainerProp
MaxSpeedMPH50 FuelUseMPG 13 FuelTypeDiesel
2.5-TonTruck ltprototypegt
VehiclePG ltinstancegt
ContainerPG ltinstancegt
NIIN 123456789 VehicleProp ContainProp
MaxSpeedMPH50 FuelUseMPG 15 FuelTypeDiesel
MaxWtSTON 3.0 MaxVolMTON 2.0
Truck-T770 5-Ton Truck With Long Bed
Asset Instances
Truck-T789 5-Ton Truck
Truck-T123 2-1/2 Ton Truck
Truck-T789 5-Ton Truck
Truck-T123 2-1/2 Ton Truck
VINT789 ( )
Truck-T789 5-Ton Truck
VINT123 ( )
Truck-T123 2-1/2 Ton Truck
VINT770 ContainerProp
VINT789 ( )
VINT123 ( )
Truck-T123 2.5-Ton Truck
Truck-T789 5-Ton Truck
VINT789 ( )
VINT123 ( )
VINT123
VINT789
55
References

• Barger, Mark, Wong, Jason. (2004). Cougaar
  Training Slides.
• BBN Technologies. (Version 10.0). (2003).
  Cougaar Architecture Document.
• BBN Technologies. (Version 10.0). (2003).
  Cougaar Developers Guide.
• Berliner, Jeffrey, Thome, Michael, Cerys,
  Daniel. (2003). Multi-Resolutional Knowledge
  Representation Using Prototypes and Properties.
• Cougaar Open Source Web Site,
  http//www.cougaar.org.
• UltraLog Web Site, DARPA, http//www.ultralog.ne
  t/.