2006-6-3 AGrIP-SELMAS Zhongzhi Shi 1
Software Engineering for Large-scale Multi-Agent Systems (SELMAS06)
Zhongzhi [email protected]
Institute of Computing TechnologyChinese Academy of Sciences
Agent-Grid Intelligence Platform for Collaborative Working Environment
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 2
Acknowledgement
China Important Basic Research Programme 973 China High-Tech Programme 863 National Natural Science Foundation of ChinaMinistry of Information Industry, ChinaKnowledge Innovative Programme of CAS
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 3
Self-Introduction
Zhongzhi Shi is Professor at the Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, IEEE Senior member.
He is Chair of WG 12.2 of IFIP. He also serves as Vice President of Chinese Association for Artificial Intelligence.
He received the 2nd Grade National Award of Science and Technology Progress in 2002. In 1998 and 2001 he received the 2nd Grade Award of Science and Technology Progress from the Chinese Academy of Sciences.
His research interests include intelligence science, multiagentsystems, Semantic Web, machine learning and data mining.
He published 10 books, edited 11 books and more than 350 technical papers.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 4
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 5
Agent Computing
Agent computing is particularly well suited to the collaborative work. The agent-based computing paradigm has following features:
• Autonomy - agent operate without intervention;
• Social ability – agents interact each other using an agent communication language;
• Goal driven – agent exhibit goal-directed behavior;
• Reactivity – agents perceive and respond to their environment.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 6
Semantic Web
Semantic Web will provide well-defined meaningwhich better enabling computers and people to work in collaboration
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 7
Grid Computing
Flexible, secure, coordinated resource sharing among dynamic collections of individuals, institutions, and resource
From “The Anatomy of the Grid: Enabling Scalable Virtual Organizations”
Grid Forumwww.gridforum.org
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 8
Brain Meet Brawn
Ian Foster, Nicolas R. Jennings, Carl Kasselman. Brain Meet Brawn: Why Grid and Agents need each others. AAMAS’04
The Grid community has historically focused on "brawn": infrastructure, tools, and applications for reliable and secure resource sharing within dynamic and geographically distributed virtual organizations.
The agents community has focused on "brain": autonomous problem solvers that can act flexibly in uncertain and dynamic environments.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 9
Research on MAS
1994:Multiagent Processing Environment MAPE
1996:Agent-based CSCW
1998:Common Agent Request Broker Architecture CARBA,
MAPE2
2000:Multiagent Environment MAGE
2002: Agent Grid Intelligence Platform AGrIP
2003: Dynamic Description Logic DDL
2004: Visual Agent Developing Tool VAStudio
2005: Ontology-based Knowledge Management KMSphere
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 10
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 11
BDI Model
Beliefs
Environment
Beliefs
PlansPlans
FIPA
GoalsGoals
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 12
Description Logic
Concepts and RoleTbox——AssertionsAbox——InstanceReasoning mechanism in terms of Tboxand Abox
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 13
Reasoning in DL
1) Subsumption2) consistency3) satisfiability4) instance checking
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 14
KB
TBox(Scheme)Man = Human ⊓ Male
Happy-father = Human ⊓ ∃ Has-child.Female⊓ …
Abox(Data)John: Happy-father
<John,Mary> : Has-child
Reasoning
Interface
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 15
Dynamic Description Logic
The primitive symbolsConcept name:C1, C2, …;
Role name:R1, R2, …;
Individual constant:a, b, c, …;
Individual variable:x, y, z, …;
Concept operation:¬, ⊓, ⊔, ∃, ∀;Axiom operation:¬, ∧, →∀;Action:A1, A2, …;
Action constraction:;(composition),⋃ (alternation),* (repeat),?(test);Action variable:α,β, …;
Axiom variable:ϕ, ψ, π, …;
State variable:u, v, w, …;
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 16
Dynamic Description Logic
Concepts in DDL are defined as the following:
(1) Primitive concept P, top ⊤ and bottom ⊥are concepts.
(2) ¬C, C⊓D, C⊔D are concepts.
(3) ∃R.C, ∀R.C are concepts.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 17
An action description is the form of),(),...,( 1 AAn EPxxA =
where(1) A is the action name.(2) x1, …, xn are individual variables, which denote the objects the action operate on.(3) PA is the set of preconditions, which must be satisfied before the action is executed.(4) EA is the set of results, which denote the effects of the action.
Dynamic Description Logic
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 18
DDL Semantics
Actions in DDL are defined as the following: Atom action A(a1, …, an) is action.If α and β are actions, then α;β, α⋃β, α* are actions;If ϕ is an assertion formula, then ϕ ? is action.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 19
Metal State ModelMental State: <K, A, G, P, I >,
Where
K beliefA actionG goalP planI intention
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 20
Belief
K = <T, S, B >
T: OntologyS: ConstrainsB: Current belief
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 21
Belief Revision
AddBelief(F, B) {F′ ← Extend(F); Foreach ϕ∈F′ do
If ¬ ϕ ∈B Then B ← B − {¬ ϕ};B′ ← Extend(B∪F);If Consistent(B′) Then Return B′;
Else{Let {ψ,¬ψ} = ConflictSet(B′);If ψ∈B Then Return B′ − {ψ};
Else If ¬ψ∈B Then Return B′− {¬ψ};Else Return error;
}}
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 22
GoalLet A be a set of actions, L be a set of assertions, G can be defined
recursively :
(1) A⊆G,A is basic action;
(2) If ϕ ∈L,then achieve(ϕ )∈G;
(3) If ϕ ∈L,then ϕ?∈G;
(4) If δ1, δ2∈G,
then δ1;δ2∈G,δ1⋃δ2∈G,δ1*∈G;
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 23
Static Plan
δ1 ⇐ ϕ | δ2
δ1∈G, δ2∈G, ϕ ∈L。
δ1: Rule headerδ2: Rule body ϕ: Rule guard condition
go_floor(x)⇐@At(elevator, y) | [y<x?;up(x)∪y>x?;down(x)]
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 24
Dynamic PlanPlan(δ, B) //Goal δ, B Current belief set
{If perform action α B⊢Tδ Then
{ Return B;Enqueue(α, P);}Else {
look for subgoals to reach goal ;computing Eα and δ for each subgoal;computing the order of subgoalsselect optimal subgoal plan;sorting subgoals and form sequence δ1,…,δ n
For π = δ1 to δ n doB = Plan(π,B)
}}
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 25
Petri Net
OWL-S DDL
OWL-S
Interpreter
Petri NetAnalysor
Services
Petri Net
Generator
Incidences matrix+DDL
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 26
Emotion Agent
SensorSensor
EffectorEffector
BeliefBelief
EnvironmentEnvironment
Rational Inf.Rational Inf.
IntentionIntention
Emotion KB Emotion KB
Emotion Inf.Emotion Inf.
DesireDesire
PlanningPlanning
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 27
Agent Architecture
Agent kernel
SensorFunctionModuleInterface
Resource Database
Engine
Communicator
Scheduling
Function Component
Plug-INs
Reasoning
Negotiation
coOperation
others
Plug-inManager
TaskDatabase
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 28
Agent Life Cycle
Waiting Suspended
Transit Initiated
Active
Wake Up
Move
ExecuteInvoke
ResumeSuspend
Destroy
Quit
Create
Dead
Wait
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 29
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 30
Multiagent Environment MAGE
Req
uirem
ent A
nalysis
System
Design
System Development
Beh
aviour
Lib
rary
Agen
t Lib
rary
Agen
t Society
System
Dep
loymen
t
VAStudio MAGE Running Support
AUMP
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 31
Multiagent System Design Procedure
Use Cases Model
FunctionDescription
Activity Model
Reaction Rule Model
Inference Model
BehaviorDescription State model Interactive
ModelPlan Model
Organization Model
Agent ModelAgent Description
Ontology Model
OntologyDescription
Platform ModelPlatformDescription
一
二
三
四
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 32
AUML—Use Cases
actor
use case:goal
system
<<include>>
association
generalization
include system boundary
<<extend>>
actor use case
extend
Actor1
Use Case 5:Goal5
<<Extend>>
<<Include>>
System1
Actor2
Use Case1:Goal1
Use Case 4:Goal4
Use Case 3:Goal3
Use Case 2:Goal2
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 33
AUML—Activity Diagram
activity(:goal)
activity
start end decision
inputoutput
output event input event
fork or jointransition role
role
Symbols in Activity Diagram
(condition)
Activity1
Condition1Output1 Input1
Activity2:Goal1
Activity3
Condition2
Activity4
Role1 Role2
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 34
AUML—Interactive Diagram
role
role lifeline and split and combination or split or combination xor split xor combination
× ×
timelimit split
T
thread asynchronous message synchronous message
com-act: content com-act: content
and split and combination or split or combination decision split decision combination
× ×
terminate of thread
Role1 Role2 Role3
TX
InformIf4 Cancel
QueryRef2QueryRef1
InformRef2
InformRef1
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 35
AUML—State Diagram
initial state finial state transition
state
state
event/action
State1Event1/action1
State1Event4/action4
Event2/action2 Event3/action3
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 36
AUML—Reaction Diagram
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 37
AUML—Inference Model
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 38
AUML—Plan Model
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 39
AUML—Agent Class Diagram
agent classgoal
role: modelontology-classreaction-rule-libaction-librational-rule-lib
activity:Goal1
agent class
activity
association
role1,role2
Agent-class1Goal1
Role1: Activity1Ontology-class1
Agent-class2Goal2
Activity1: Role1,Role2Activity2: Role1
Ontology-class1Action-lib1
Activity1:Goal1
Role1 Role1,Role2
Activity2:Goal2
Role1
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 40
AUML—Organization Diagram
organization
agent
control relationship
control
agent
organization
peer relationship
peer
benevolence relationship
benevolence
benevolence
dependency relationship
dependency
dependency
Organization1
Agent1
Agent2
Agent3
Agent4
Agent5
peer
control
control control
control
benevolence
dependency
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 41
AUML—Ontology Diagram
association generalization
name
attr1: datatypeattr2: datatype
ontology class aggregate
role1 association-name role2
multiplicity1 multiplicity2
multiplicity1 multiplicity2
Class1
attr1: datatypeattr2: datatype
Class3
attr1: datatypeattr2: datatype
Class2
attr1: datatypeattr2: datatype
Class6
attr1: datatypeattr2: datatype
Class4
attr1: datatypeattr2: datatype
Class5
attr1: datatypeattr2: datatype
role1 association1 role2
0..1 *
1
1 *
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 42
AUML—Platform Diagram
communicationagent platform
agent platform
agent agent movement
moveagent:agent class
goal1,goal2
agent clone
clone
move
Agent platform1
Agent platform2
TCP/IP
Agent1:Agentclass1
Goal1
Agent1:Agentclass1
Goal1
Agent2..5:Agentclass1
Goal2
Agent6:Agentclass2
Goal3
Agent7:Agentclass2
Goal3clone
move
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 43
AUMP —Source Code
MAGE的行为类结构
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 44
Compare AUML with UML
ModelElement
Classifier
Feature
StructuralFeature
BehavioralFeature
Attribute
Operation Method
Class
Relationship
Generalization
Association
AssociationEnd
AgentClass
Organizational
Relationship
Control Peer
Benevolence
Dependency
Goal
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 45
Compare AUML with UML
Use Cases Model
Function
Activity Model
Reaction Model Inference Model
BehaviorState Model Interactive
ModelPlan Model
Organization Model
Agent ModelAgent
Ontology Model
Ontology
Platform ModelPlatform
一
二
三
四
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 46
Agent Unified Modeling Platform —AUMP
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 47
Computation Model for Agent
AgentClient/Service
AgentClient/Service
Soft Bus
Agent Service
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 48
Common Object Request Broker Architecture
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 49
Common Agent Request Broker Architecture
ApplicationPattern
ApplicationPattern
ApplicationFacilities
ApplicationFacilities
Agent Request BrokerAgent Request Broker
AgentServices
AgentServices
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 50
VAStudio ArchitectureAgent Society
Agents
Behaviours
Editor
Agent Library
Behaviour Library
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 51
VAStudio Architecture
VAStudio Design PlatformDesign GUI
TemplateLibrary
Composition
AgentClone FSM
Process Design
VAStudio Develop PlatformDevelop GUI
Edit Compile Debug Test
Process Design
VAStudio Run PlatformRun Monitor
AgentWorkList
ADL/BDLLoader
ReasonerLoader
ProcessEntity
MAGE
LibraryInterface
WSInterface
OntologyInterface
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 52
Behavior Library
• Data Package
package intsci.ace.data
• Neural Network Package
package intsci.ace.neural
• Learning Packagepackage intsci.ace.learning
• Data Mining Package
package intsci.ace.mining
• Language Processing Package
package intsci.ace.language
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 53
Behavior Library• Graphic Package
package intsci.ace.graphics
• Image Package
package intsci. ace .image
• Search Package
package intsci.ace.search
• Expert System Package
package intsci.ace.expert
• Model Package
package intsci.ace.model
• Decision Making Packagepackage intsci.ace.decision
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 54
Screenshot of VAStudio
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 55
VAStudio:Visual Agent Studio
Select agent society
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 56
VAStudio:Visual Agent Studio
Select behavior from behavior lib.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 57
Create Agent by VAStudio
Add behavior
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 58
Create Agent by VAStudio
Fill out behavior parameters
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 59
Create Agent by VAStudio
Build agent (code generation)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 60
Build Multiagent System by VAStudio
Auction protocol using FSM
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 61
Build Multiagent System by VAStudio
Seller and buyer using clone
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 62
Build Multiagent System by VAStudio
Compile to generate the auction system
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 63
Build Multiagent System by VAStudio
Run the auction
system in MAGE
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 64
Build Multiagent System by VAStudio
Display the agent running
procedure
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 65
Build Multiagent System by VAStudio
Seller agent running
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 66
Run-Time Platform
MAGE
主体
Agent Management
System
DirectoryFacilitator
主体Agent
Message Transport System (MTS)
Software
AgentLibrary
FunctionComponent
Other Agent Platforms
Message Transport System (MTS)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 67
AgentCities
Beijing!
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 68
MAGE ComparisonPhase
Agent Platform AgentBuilder Jack Zeus MAGE
Analysis
Completeness: ★★★ ★★★★ ★★★★★
Applicability: ★★★ ★★★ ★★★★
Complexity: ★★★★★ ★★★★★ ★★★★
Reusability ★★★ ★★★ ★★★★★
Design
Completeness: ★★★ ★ ★★★★ ★★★★★
Applicability: ★★★ ★ ★★★ ★★★★
Complexity: ★★★★ ★ ★★★ ★★
Reusability ★★★★
★★★ ★★★★
Development
Completeness: ★★★★★ ★★★★ ★★★★★ ★★★★★
Applicability: ★★★ ★★★★★ ★★★ ★★★★
Complexity: ★★★★ ★★ ★★★★ ★★★
Reusability ★★ ★★★★★★ ★★ ★★★
Deployment
Completeness: ★★★ ★★★ ★★★★ ★★★★★
Applicability: ★★★ ★★★★ ★★★ ★★★
Complexity: ★★★★ ★★★ ★★★★ ★★★★★
Reusability ★
★★★
★★★
★★★
★★★
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 69
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 70
Agent Collaboration
ACLWorking Flow Ontology-based Knowledge ManagementPolicy DrivenPlanning
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 71
Ontology
In philosophy, an ontology is a theory about the nature of existence.
An ontology is a document or file that formally defines the relations among terms.
An ontology is a formal, explicit specification of a shared conceptualization.
The most typical kind of ontology for the Web has a taxonomy and a set of inference rules.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 72
Semantic Web Layer Cake
10 Feb 2004
W3C
Recommendation
OWL
by Tim Berners-Lee
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 73
Collaborative Architecture
Agent B
Goal Instance
Agent A
Goal Instance
has(1,n)
compatible goals = collaboration partners
automated collaborationexecution
KMSphere
uses(1,n)
has(1,n)
uses(1,n)
Collaboration Service
Web Service
Goal Template Repository
ServiceRepository
Collaboration Service
Web Service
OwnerOwnergoal assignment
service discovery
owns
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 74
KMSphere
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 75
KMSphere
Email Document File Image Video Web
Ontology Acquisition
Knowledge organization
Knowledge Distribution
KnowledgeApplication
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 76
KMSphere
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 77
KMSphere Demo
Create ontology by hand
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 78
KMSphere Demo
Ontology acquisition
from databases
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 79
KMSphere Demo
Ontology acquisition from text
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 80
KMSphere DemoEdit
ontology
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 81
KMSphere Demo
Ontology consistency
check
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 82
KMSphere Demo
RDQL (RDF Data
Query Language)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 83
Semantic Web ServicesDefine exhaustive description frameworks for describing Web Services and related aspects (Web Service Description Ontologies)
Support ontologies as underlying data model to allow machine supported data interpretation (Semantic Web aspect)
Define semantically driven technologies for automation of the Web Service usage process (Web Service aspect)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 84
OWL-S
Service
ServiceModel
ServiceGrounding
ServiceProfile
Presents(What it does)
Support
How to access it
Described by
How it works
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 85
OWL-S Context
Service Resource User Provides
Described by
Service Model
U-Context
Show U-Context
Service Grounding
SupportsPresents
Uses
Service Profile
S-Context
R-Context
Show R-Context
Show S-Context
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 86
Service Description Language SDLSIN
<asdl-descr>::=(ctype:service-name name:context context-name+:types (type-name = <modifier> type)+:isa name+:inputs (variable: <modifier> put-type-name)+:outputs (variable: <modifier> put-type-name)+:input-constraints (constraint)+:output-constrains (constraint)+:io-constrains (constraint)+:concept-description (ontology-name = ontology-body)+:state-language name:concept-language name:attributes (attributes-name : attributes-value)+:text-description name
)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 87
Policy Driven
p S A S U=< >trigger goal, , ,
Strigger: set of states for policy executionA: set of actionsSgoal : set of goalsU: set of utility functions
Goal stateoptimizationAction
policyGoal policy Utility
policy
Modelingplanning
Modelingoptimization
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 88
Action Policy
<ActionPolicy> :: =<Name><Parent><Performative><Type><Subject><Object><Action><Precondition><ConstraintLanauge>
<Duration><Priority>
Access Control PolicyObligation Policy
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 89
Action Policy
Gold-Silver Action Policies
G: IF (RTG > 100 msec) THEN (Increase CPUG by 5%)
S: IF (RTS > 200 msec)THEN (Increase CPUS by 5%)
Overlapping Action Policies
Conflict if CPU (almost) fully utilized!
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 90
Goal Policy
<GoalPolicy>::= <Name><Performative><Subject><Object><Precondition><Postcondition><ConstraintLanauge><Duration><Priority>
( :name goalpolicy1:performative “Achieve”:subject ftpagent:precondition between(clienthost, 192.168.0.0,
192.168.0.255):postcondition greater(bandwidth(clienthost), 1M)
)
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 91
Optimization
<UtilityPolicy>::= <Name><Performative><Subject><Object><Precondition><UtilityFunction><ConstraintLanauge><Duration><Priority>
( :name utilitypolicy1:performative “Optimize”:subject httpagent:utilityfunction
)∑∑ +=
ii
ii clientmeresponsetiwclientbandwidthwf )()( 21
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 92
Plan Representation
A plan Π is a triple <SO, OO, CS>
SO: a set of action-steps, OO: a set of ordering constraints on the actions in SO.CS: a set of variable-binding constraints between the
variables of the action-steps in SO and other variables or constants.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 93
Distributed Multiagent Plan Algorithm
Input: a set of action-templates OPS, a set of capabilities of agents CaS , initial state I , and goal G
Output: Plan Π
),,,( GICaSOPSPlanMultiAgentdDistribute −−
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 94
Distributed Multiagent Plan AlgorithmProcedure:1. Inform all agents that I will start planning, and wait for them
finishing current actions and entering plan state.2. Initialize: Build an empty plan where
, , , . Decompose goalinto a set of sub-goals through adding to
according to every literal .3. Invoke recursive function .4. Taking effect: For each action executed by other agent, request
it to add the action. For each constraint related to the actionsexecuted by other agents, request it them to add the constraints.
5. Inform all agents the planning finishes, thus they can continue their plans.
6. If the invoking succeeds, then return success, otherwise return fault.
⟩⟨=Ψ CLSCSOOSO ,,,},{ GoalInitSO = φ=OO φ=CS φ=CLS G
GS Goalg, GS
Gg∈),( GSStepNextGenerate Ψ−−=ϕ
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 95
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 96
Agent Grid
Resource-specific implementations of basic servicesE.g., Transport protocols, name servers, differentiated services, CPU schedulers, public keyinfrastructure, site accounting, directory service, OS bypass
Resource-independent and application-independent servicesauthentication, authorization, resource location, resource allocation, events, accounting,
remote data access, information, policy, fault detection
DistributedComputing
Toolkit
CommonResources
AgentEnvironment
DevelopingToolkits
Data-Intensive
ApplicationsToolkit
CollaborativeApplications
Toolkit
RemoteVisualizationApplications
Toolkit
ProblemSolving
ApplicationsToolkit
RemoteInstrumentation
ApplicationsToolkit
ApplicationsService
IMS
Biology
Power Supply
E-Business
Environment
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 97
Agent Grid Intelligence Platform AGrIP
Information Information SoursesSourses
WebWeb
DatabasesDatabases
Stream Stream MediaMedia
MiddelwareMiddelwareMiddelware
DSSDSSDSS
MSMinerMSMinerMSMiner MIRESMIRESMIRES
MAGEMAGE
KMSphereKMSphereKMSphere
GHuntGHuntGHunt OKPSOKPSOKPS
ApplicationsApplications
E-BEE--BB E-GEE--GG IEIEIE IBIBIB SimulSimulSimul CorlCorlCorlDiag.DiagDiag..
CADCADCAD
GISGISCBRCBR
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 98
Data Mining Platform MSMiner
MSMiner Architecture
Data resources
数据仓库
Mete D
ata Managem
ent
Extract Transform Load
Topic2Topic1 Topicn
OLAP
Data Mining
...
MSDM
MSMetadata
MSETL
Data Warehouse
MSOlap
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 99用 户 接 口
信 息 定 制 文 本 信 息 检 索基 于 内 容 的图 像 检 索
数 据 表 格 检 索
U R L 数 据库
信 息 收 集S p i d e r
文 本 数 据 库 图 像 数 据 库
倒 排 文件 索 引
表 格 数 据 库 库
全文检索
目录检索
概念语义
检索
文 本 目录 索 引
语 义索 引
专题检索
专 题索 引
网 页 内 容 解 析
切词倒排
文本聚类
文本分类
信息抽取
概 念空 间
共现分析
图像分类
专 题索 引
文 本 摘要 数 据
库
文本摘要
图像标注
图 像索 引
目录检索
图像语义
检索
网址解析
URL
管理器
表格分类
表 格索 引
图像语义
分析
图 像语 义索 引
I n t e r n e t
词 库 语 料 库
用 户 导 航
网 络 日 志O L A P信 息 检 索 管 理
检索请求
检索结果
Search Enginer GHunt
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 100
Expert System OKPS
KnowledgeAcquisition
Knowledge Base
Inference EngineData Mining
InterfaceData Warehouse Visualization
OLAP
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 101
Outline
IntroductionAgent ModelMultiagent Environment— MAGEAgent CollaborationAgent Grid Intelligence PlatformApplicationsConclusions
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 102
Emergency Interactive Systsem GEIS
Emergency
Signal
Emergency Center
Emergency Case Base
Retriv
al Case Base
Solutions
Decision Making in Terms of Reasoning
EmergencyExecution
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 103
GEIS Interface
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 104
Receive Crime Interface
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 105
Process Crime Interface
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 106
Geographical Information System
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 107
CBR-Based Reserve Plan Against Emergency
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 108
CBR-Based Reserve Plan Against Emergency
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 109
Fish Field Forecast in Eastern Sea of China
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 110
DSS for Oil Pipeline Design
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 111
Conclusions
Proposed Agent Models
MAGE Satisfies the Software Engineering for Large Multiagent Systems
Collaborative Working Agents on Semantic Grid
AGrIP is a Powerful Platform for Constructing Large Complex Systems
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 112
Published Paper List• Zhongzhi Shi, He Huang, Jiewen Luo, Fen Lin, Haijun Zhang. Agent-based Grid Computing, JAMM, 2006•Jiewen Luo, Zhongzhi Shi, Maoguang Wang, Jun Hu: AGrIP: An Agent Grid Intelligent Platform for Distributed System Integration. APWeb Workshops 2006: 590-594•Zhongzhi Shi. Autonomic Semantic Grid. Keynote speaker. IFIP AIAI2005, Sept., 7-9, 2005•Zhongzhi Shi, Mingkai Dong, Yuncheng Jiang, Haijun Zhang. A Logic Foundation for the Semantic Web. Science in China, Series F Information Sciences, 48(2):161-178, 2005•Zhongzhi Shi. Agent-based Semantic Grid. China Agent System, IFMAS, 2004, 12, 13-17•Zhongzhi Shi, Haijun Zhang, Mingkai Dong, MAGE: Multi-Agent Environment, ICCNMC-03, IEEE CS Press, pp.181-188, 2003•Zhongzhi Shi, Mingkai Dong, Haijun Zhang, Qiujian Sheng. Agent-based Grid Computing.Keynote Speech, International Symposium on Distributed Computing and Applications to Business, Engineering and Science, Wuxi, Dec. 16-20, 2002•Zhongzhi Shi, Wenpin Jiao, Qiujian Sheng. Agent-Oriented Software Methodology.CEEMAS2001, Cracow, Poland,2001 •Zhongzhi Shi. Agent-based E-commerce. Invited Speaker, DS-9, Hong Kong, 2001
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 113
Published Paper List• Zhongzhi Shi, Ju Wang, and Hu Cao, The Security Mechanism In Multi-Agent System AOSDE, in: LNAI, Vol. 1733, pp.62-71 PRIMA’99, Dec. 1999.• Zhongzhi Shi, Qijia Tian, Yunfeng Li. RAO Logic for Multiagent Framework. Journal of Comouter Science and Technology, July, 1999• Zhongzhi Shi, Hu Cao, Yunfeng Li, Wenjie Wang, Tao Jiang, A Building Tool for Multiagent Systems: AOSDE. IT \& Knows, IFIP WCC '98, 1998• Zhongzhi Shi, Qijia Tian, Yunfeng Li, RAO Logic for Multiagent Framework, DAIMAS'97, St. Petersburg, Russia, 1997. • Zhongzhi Shi, Tao Wang, An Agent based CSCW System Architecture, Chinese Journal of Electronics, 6(1), 1996. • Wenjie Wang Zhongzhi Shi Qijia Tian Tao Wang. An Agent Belief System for Multi-agent System. IEEE SMC'96, 2065-2069, 1996. • Zhongzhi Shi, etc., MAPE: Multi-Agent Processing Environment, PRICAI-94, 1994.
2006-6-3 AGrIP-SELMAS Zhongzhi Shi 114
THANK YOU!
Intelligence Sciencehttp://www.intsci.ac.cn/