software agents for coalition forces
DESCRIPTION
Software Agents for Coalition Forces. Second International Conference on Knowledge Systems for Coalition Operations, 23rd and 24th April 2002 Toulouse, France. By Zakaria Maamar , Paul Labbé, and Wathiq Mansoor. - PowerPoint PPT PresentationTRANSCRIPT
KSCO April 2002 Toulouse p. 1
Software Agents for Coalition Forces
Second International Conference on Knowledge Systems for Coalition Operations,
23rd and 24th April 2002 Toulouse, France
By Zakaria Maamar, Paul Labbé, and Wathiq Mansoor.
Presented by Paul Labbé, P. Eng., IEEE Senior
Defense R & D [email protected]
Tel.: +1 (418) 844-4000 x 4479
KSCO April 2002 Toulouse FranceThis presentation reflects the views of the authors and does not necessarily represent the plans and policies of the
Canadian Department of National Defence or of Zayed University
KSCO April 2002 Toulouse p. 2
The distributed, heterogeneity, and dynamic nature of the coalition context has raised the need for new advanced
technologies. These technologies aim at managing the coalition informational infrastructure, in terms of autonomy, adaptability, and scalability. To achieve this support, Software Agents (SAs)
seem to be a promising approach. To develop this approach, different aspects of a coalition has to be identified. These aspects include the coalition structure; the roles and responsibilities held by people within the coalition; the flow of information within the
coalition; the capabilities required or available within the coalition; and the context in which the coalition operates. For
many of these aspects, SAs can be used; . For instance, the coalition structure can be associated with several SAs of different
types and with different roles.
Abstract
KSCO April 2002 Toulouse p. 3
Allied warfare objective
The ultimate objective of Allied Warfare is to increase the overall Joint/Coalition Force mission and task success rate and geopolitical influence agreed by the driving Nations.
This objective imposes more difficult to meet requirements for demanding operations as in littoral warfare theater.
The generation of the necessary conditions to increase the speed and accuracy of Joint/Coalition command and control
include but are not limited to the material of this presentation.
Introduction
KSCO April 2002 Toulouse p. 4
From live coalition exercises
• The Force Over the Horizon Track Coordinator (FOTC) data base resulted in many large inaccuracies and inconsistencies in the Global Command and Control System- Maritime (GCCS-M) picture (identity attribute of tracks (ID) jumping and time lateness)
• Operators believed picture to be accurate• Stale data used to make decisions (Blue Force was
successfully ambushed by Orange ships)• FOTC held correct ID on several Hostile Tracks while
LINK 11 reported them as Unknown for long periods• Orange ships came within weapons range of high value
blue units (Kittyhawk) without being reported for long periods of time
• Sanitization rules within Radiant Mercury strip vital data (e.g. source data)
Requirement Specification
KSCO April 2002 Toulouse p. 5
How to deal with interoperability issues?
Interoperabilityenvironment
Users' needsissue- a -
Informationdisparity issue
- c -Distributionissue- b -
SAs
Dis. ObjectComputing
OntologyWFs
CodeMobility
Agent-based architecture
KSCO April 2002 Toulouse p. 6
Three dimensions of interoperability issues:
physical interconnectivity, application integration, and command collaboration
Application level
Physical level
rely on
build on
Commandment level
Application level
Physical level
rely on
build on
Headquarter HeadquarterCollaboration
Integration
Interconnectivity
Coordinationspace
MAS 1 MAS i
Coalitionmission
Services Services
invoke
use
invoke
Network
Military Forces 1 Military Forces i
use
Commandment level
use
KSCO April 2002 Toulouse p. 7
We propose a software agent architecture and structure to resolve some of these issues!
• Meeting infrastructure
• Information exchange control
• Multi-agent environment
• Security of execution and sharing
• Shared decision-making
• …..
KSCO April 2002 Toulouse p. 8
A software agent definition
• An autonomous entity having the abilities to assist users when performing their operations, to collaborate with each other to jointly solve different problems, and to answer users' needs
Agent-based architecture
KSCO April 2002 Toulouse p. 9
A simplified CCIS model
• A structure and a set of functions and tasks
Planning Data fusion
Resources (software,hardware)
Communication
External environment
CCIS
Functions
Structure
Users
Agent-based architecture
KSCO April 2002 Toulouse p. 10
Architecture for interoperable CCISs
• Aspects to be dealt with
– Maintain the autonomy and independence of the CCISs
– Reduce the informational disparities of the interconnected
CCISs
– Protect the interconnected CCISs from the unauthorized
accesses
– Evaluate the communication channels performance,
particularly in low-bandwidth situations (QoS, CSNI)
– Help users satisfy their needs without worrying about the
characteristics of the CCISs
Agent-based architecture
KSCO April 2002 Toulouse p. 11
Interface-Agent 1 CCIS-Agent 1 Resolution-Agent 1
MAS 1
Control-Agent 1
Advertisement Infrastructure
SA-Supervisor
Interface-Agent 2 CCIS-Agent 2 Resolution-Agent 2
MAS 2
Control-Agent 2
Function 11 Function 1i
CCIS 1
Function 21 Function 2j
CCIS 2
Autonomy management
Remote interaction/Mobility
Low-bandwidth&unreliablenetwork
Agent-based architecture
KSCO April 2002 Toulouse p. 12
Architecture main characteristics
• Interface-Agent
• CCIS-Agent/Function-Agent
• Resolution-Agent
• Control-Agent
• Supervisor-Agent
• Advertisment infrastructure
Agent-based architecture
KSCO April 2002 Toulouse p. 13
Interface-Agent
• It assists users in formulating needs, maps needs into requests, forwards requests to the CCIS-Agent in order to be processed, and provides users with answers obtained from the CCIS-Agent.
Agent-based architecture
Formulationmodule
Requests
AnswersAnswersCCIS-AgentUser
Needs
Communication layer
KSCO April 2002 Toulouse p. 14
CCIS-Agent
• It processes user requests received from the Interface-Agent, but only if these requests require the involvement of the CCIS of this particular CCIS-Agent. In the proposed architecture, a CCIS-Agent has the ability to advertise its services by posting notes on the Bulletin Board of the Advertisement Infrastructure. To do so, the CCIS-Agent can either send a remote request to the Supervisor-Agent or can migrate to this infrastructure; the choice is based on the network status. In both cases, i.e., remote request or soft-mobility, a security level associated with the CCIS-Agent is used to identify the services this CCIS-Agent is authorized to advertise.
Agent-based architecture
KSCO April 2002 Toulouse p. 15
Turning CCISs into agents of MAS• Purpose: making a CCIS to behave like a SA
– Build a SA on top of the CCIS
Agent-based architecture
Function-Agent 11 Function-Agent 1i
CCIS 1
CCIS-Agent 1
MAS 1
Function 11 Function 1i
Requests/Results Functions initiation
Other MASs
Requests/Answers forwards
KSCO April 2002 Toulouse p. 16
Pre-processing
moduleAnswers
CCIScapabilities
Interface-AgentRequests
Communication layer
Definitionmodule
Processingmodule
Monitoringmodule
Answers
Requests
Updates
ServicesSupervisor-Agent
CCIS functions
Communication layer
Resolution-Agent AdminsitratorServices
RequestsSub-answers
AnswersSub-requests
CCIS-Agent and Function-Agent modules
KSCO April 2002 Toulouse p. 17
Resolution-Agent
Agent-based architecture
Pre-processing
moduleAnswers
CCIS-AgentRequests
Communication layer
Slavemodule
Adv. Infra.
Transfer(Help-Agent)Creation
Browsingmodule
Communication layer
Queries
CCISs
Bulletin BoardCCISs
Queries
Performancemodule
Answers
Route
CCIS-AgentsSub-queries
Sub-answers
Communication layer
Creation(Help-Agent)
Creation(Route-Agent)
Sub-queries
Sub-answers
KSCO April 2002 Toulouse p. 18
Control-Agent
Agent-based architecture
KSCO April 2002 Toulouse p. 19
Supervisor-Agent
Agent-based architecture
KSCO April 2002 Toulouse p. 20
Advertisement Infrastructure
• In an interoperating environment, CCISs are generally spread across networks and rely on low capacity and unreliable channels for communication. Moreover, a military user may use his Combat Net Radio to send and request information or may rely on mobile devices, such as portable computers, that are only intermittently connected to networks. In the proposed architecture, to avoid overloading the network, CCIS-Agents and Resolution-Agents migrate to the Advertisement Infrastructure in which CCIS-Agents advertise their services by posting notes on the Bulletin Board, whereas Resolution-Agents consult the Bulletin Board to identify the CCISs that are required to satisfy user needs.
Agent-based architecture
KSCO April 2002 Toulouse p. 21
Advertisement Infrastructure
Agent-based architecture
CCIS-Agent
Advertisement Infrastructure
Bulletin Board
Repositoryof ActiveAgents
Supervisor-Agent
note 1note i
1. Remote request
2. Security check
3. Registration
4. Posting
5. Acknowledgment
KSCO April 2002 Toulouse p. 22
In actions
Agent-based architecture
Initialization MaintenanceAdvertisement Operation
Users
Administrator
Resolution-Agent
Advertisement Infrastructure
Repositoryof ActiveAgents
Supervisor-Agent
1. Help-Agent creation
Service-Agent
2. Transfer
6. Acknowledgment
3. Security check
4. Registration5.Installation
Help
KSCO April 2002 Toulouse p. 23
Satisfying a userAgent-based architecture
User
Function-Agent 11 Function-Agent 1i
CCIS 1Function 11 Function 1i
Interface-Agent 1 CCIS-Agent 1 Resolution-Agent 1
MAS 1
2. CCISs identification
3.a data/4.a results
CCIS-Agent 2
MAS 2
CCIS-Agent 3
MAS 3
9.b/13.b Mobility
10.b Local computing(Route-Agent 1)
11.b Remote computing(Route-Agent)12.b Answers
Advertisement Infrastructure
Bulletin Board
Service-Agent
4.b Queries6.b CCISs
0. Needs
6.a/16.b Answers
1. Request
5.a/15.bAnswers
3.b Request
14.b Answers
7.b Route design8.b Route-Agent
creation
5.b
Not
es
KSCO April 2002 Toulouse p. 24
Preliminary analysis• What information is to be managed, and what are the properties of
this information?– Identify and categorize information items– Specify source, destination, size, update period, comm paths,
security
• Under what circumstances will the information be used and managed?– Define: Context = Goal Stable Conditions Dynamic Conditions
– Specify the Importance (I) of each Context
• What a priori assessments can be made about the value of the different types of information in specific circumstances?– Potential (P): relevance of information for a Context– Quality (Q): how accurate information should be for a Context– Timeliness (T): how recent information should be for a Context
Improving Information Sharing
KSCO April 2002 Toulouse p. 25
Prioritization rule set
Improving Information Sharing
OBJECTIVE: Optimize use of system resources (e.g., BW), and ensure most valuable information is processed first
i = information item = contextI = Importance of w, wP, wQ, wT = weighting factors
Pi = Potential of i for Qi = Quality of i for Ti = Timeliness of i for X = accounts for other factors
(e.g., dynamic conditions)
For example:
qi = measured quality of i i = actual timeliness of i
Priority(i,) = w I · ( wP Pi + wQ Qi + wT Ti + X )
X = f ( Qi , qi , Ti , i )
KSCO April 2002 Toulouse p. 26
Quality assignment rule setOBJECTIVE: Assess the information attributes found in
Coalition databases with the aim of integrating different data sources
• Maintain separate track position and ID quality measures• Account for intrinsic sensor limitations (e.g., range,
environmental conditions) --> the best sensor does not always have the best data
• Provide a systematic and consistent statistical definition of error
• Allow degradation in position quality during DR• Problem: For security reasons, information is often sanitized
or partially stripped (e.g., source) before dissemination, making quality assessment difficult
Improving Information Sharing
KSCO April 2002 Toulouse p. 27
Impact of changes on mission
model-based measures
N
H
N
F
CSIHIH-PHFNU
- commander’s ship- intended hostile target- IH predicted position- hostile- friend- neutral- unknown
CS
FH
F
F
N
NN
N
N
actualtarget
IH
N
HH
N
H H
U
H
IH
N
N
with prediction
without prediction
IH-PIH-P
KSCO April 2002 Toulouse p. 28
Cost of time to discover, deliberate/fuse
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
This axis can be interpreted as a
combination of positional inaccuracies of surface
hostile contacts reported.
Value of the information presented to a commander in hypothetical OTH –T for the surface hostile
contacts reported.
This axis shows the age or delay since sensor time of surface hostile contacts
reported after processing and deliberation.
1 time to discovery
cost
KSCO April 2002 Toulouse p. 29
Discovery/fusion gain (recipe)
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost
KSCO April 2002 Toulouse p. 30
Gain in applying recipe to updates
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost 3 gain with
updates
KSCO April 2002 Toulouse p. 31
Loss in sharing the result
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost 3 gain with
updates
4 cost for sharing
KSCO April 2002 Toulouse p. 32
Sharing recipe preserve gain + capacity
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost3 gain with
updates
4 cost for sharing
5 improvement by sharing recipe instead of fusion results
KSCO April 2002 Toulouse p. 33
Improving interoperability + sharing
• Must be able to measure value of local discovery or fusion.• When a discovery or fusion improves own picture above the
received picture by a given threshold – display this result locally– send own data used in recipe– send recipe with list of ingredients (track# used)– responsible for sending own data for this recipe until found
inadequate locally or remotely.• Eliminate data incest and does not require the sharing of source
identity (avoid loss of information required for appropriate MSDF) but provide an improved confidence in shared information. Provide “track pedigree”.
• First steps in developing agreed information quality schemes.• Important unit and force effectiveness gain for various missions.
KSCO April 2002 Toulouse p. 34
Cost of time to discover, deliberate/fuse
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
This axis can be interpreted as a
combination of positional inaccuracies of surface
hostile contacts reported.
Value of the information presented to a commander in hypothetical OTH –T for the surface hostile
contacts reported.
This axis shows the age or delay since sensor time of surface hostile contacts
reported after processing and deliberation.
1 time to discovery
cost
KSCO April 2002 Toulouse p. 35
Discovery/fusion gain (recipe)
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost
KSCO April 2002 Toulouse p. 36
Gain in applying recipe to updates
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost 3 gain with
updates
KSCO April 2002 Toulouse p. 37
Loss in sharing the result
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost 3 gain with
updates
4 cost for sharing
KSCO April 2002 Toulouse p. 38
Sharing recipe preserve gain + capacity
0 10 20 30 40 50 60
0
4
8
12
16
20
-0.50
-0.30
-0.10
0.10
0.30
0.50
Ave
rag
e V
alu
e
Delay (min)
CUA radius (km)
FOTC - Pertinence of engagement (POE)0.3-0.5
0.1-0.3
-0.1-0.1
-0.3--0.1
-0.5--0.3
2 gain from
discovery
1 time to discovery
cost3 gain with
updates
4 cost for sharing
5 improvement by sharing recipe instead of fusion results
KSCO April 2002 Toulouse p. 39
Improving interoperability + sharing
• Must be able to measure value of local discovery or fusion.• When a discovery or fusion improves own picture above the
received picture by a given threshold – display this result locally– send own data used in recipe– send recipe with list of ingredients (track# used)– responsible for sending own data for this recipe until found
inadequate locally or remotely.• Eliminate data incest and does not require the sharing of source
identity (avoid loss of information required for appropriate MSDF) but provide an improved confidence in shared information. Provide “track pedigree”.
• First steps in developing agreed information quality schemes.• Important unit and force effectiveness gain for various missions.
KSCO April 2002 Toulouse p. 40
•The impact on mission effectiveness of adopting a meeting infrastructure exploiting agent-based architectures for CCISs need to be considered and be accurately assessed.•Presented major characteristics of the MAS interoperability approach and the design of collaborative environments for distributed and heterogeneous CCISs. •Eight types of SAs exist in the architecture proposed for coalition support (Interface-Agent, CCIS-Agent, Resolution-Agent, Control-Agent, Function-Agent, Supervisor-Agent, Help-Agent, Route-Agent) while four stages describe this architecture operating (Initialization, Advertisement, Operation, Maintenance). •Further works need to be done for demonstrating the value of the coalition embedded characteristics of the proposed infrastructure.
Conclusions and recommendations
KSCO April 2002 Toulouse p. 41
Using information value for optimizing end users’ shared awareness is not simple but the potential gains outweigh the effort required, by delivering
increases in mission precision and success rate that guarantee long term benefits and would increase
public support.
An agent-based architecture would also provide cost effective capabilities for future improvements,
measurability, maintainability and support for training and simulation.
Conclusions and recommendations (cont’d)