Transcript Using Transactional Workflow Ontology in Agent Cooperation

Using Transactional
Workflow Ontology in
Agent Cooperation
J. Korhonen, L. Pajunen, and
J. Puustjärvi
Background

BRIEFS information extraction system
 Analyses

Transaction research in WorkMan
 Database

news articles for corporate actions
constraint used for workflow control
BriefsAgents – a multi-agent system
 Ontology
used to access information
 Built on top of BRIEFS database
Technology

Multi-agent system
(MAS)
 FIPA agent

platform
Web services
 SOAP,
WSDL, UDDI
 WSFL

DAML+OIL ontology
 DAML-S
BriefsAgents system

Multi-agent system




Ontology communication
Briefs database used for
information



FIPA-OS agent platform
Entity agents, Query
agents and Insert agent
Agent
Entity
Query
Insert
Org
Company
Deals
Deal
Person
Jobs
Entity agents as wrappers
Queries made using
several Entity agents
Problem: changes to
query protocol
Person
Workflow in Web Services
Web service offers operations
 WSDL describes operations

 lack
of expressiveness on workflows
XLANG, WSFL, and WSCL add
capabilities
 Local and global models describe
individual services and composed system

Workflow Ontology 1/2

Directs agents own workflow
 Internal

use
Describe communication protocol
 Multi-agent

Describes how to use offered services
 Dynamic

use
collaboration with unknown agents
Used to compose higher-level services
Workflow Ontology 2/2


Three entities
Message-oriented tasks
 Agents

Control links concurrent
or sequential
 Go,

and web services
fork, join, loop etc.
Data links to/from
blackboard
 Named
data
<<frame>>
Task
Task type
Transaction type
Optional
Commit operation
Abort operation
Protocol type
2
0..*
<<frame>>
Control Link
Link type
Condition
<<frame>>
Data Link
Data name
Translation
2
0..*
Transactional Workflow 1/2
Transactions add reliability to the workflow
 Different transaction models

 For
example nested, option, saga models
 May not be known beforehand for all agents

Tasks have a transaction model attribute
 Optional
does not, model may also be ’none’
 Commit operation, abort operation
Transactional Workflow 2/2

Special tasks
 Start
transaction
 End transaction


Useful for nested
transactions
End transaction
checks failures
 Call
commit/abort
Start
Go
Start
Transaction
Go
Order
Tickets
End
Transaction
Abort
Commit
Commit
Order
Abort
Order
End
Transactional workflow in MAS


Agent internal structure
Agent workflows should
be available




In FIPA terms, a WMA
Web service workflows
should be available
Dynamic negotiation of
protocols
Combined transactions

also web services
Agent Framework
WMA
Agent
Workflow engine
State engine
Web service
interface
Conclusions
Workflow ontology enables agent systems
to collaborate dynamically
 Transactional support needed in agent
systems and web services
 Agent systems integrate with web services
