Logic Foundation of Agents And Its extension

1
Semantic Web Services Composition And
Agent applications for Semantic Web
2

• Motive
• Semantic Web can be obtained by regarding its
  content as behavioral intelligence.
• Services encoded in DAML-S can be viewed
• as specifications for extensions of the user
  agents attempting to exploit the services or
• as independent, collaborative agents that can be
  awakened to assist the user agents.
• Towards a practical architecture for deliberative
  agents for the Semantic Web

3

• Outline
• 1 Definitions Agents and Services
• 2 Bringing Services onto the Semantic Web
• 3 Semantic Web and Software engineering
• 4 Web Services as Agent Behavior
• 5 Practical Reasoning Agents for Semantic Web
• 6 Other Semantic Web Service
• 7 Open issues
• 8 Agent Capabilities

4

• 1 Definitions Agents and Services
• Agent autonomous actor with sets of
• Beliefs, knowledge about world, limited,
  inaccurate
• Desires, goals the agent works to achieve or
  fulfill
• Intentions, goals or subgoals agent is currently
  pursuing
• Behaviors, actions the agent is able to take
• Services Web-accessible program or device
  affecting the world or not,
  free or for sale
• Online book-selling (amazon)
• Web searching engine performing for keywords
  (Google)
• Software company providing program patch or
  update(zero G http//www. Installanywhere.com)
• Composite services
• customizations of reusable, high-level generic
  proceduresto construct and to archive them in
  sharable generic procedures ontologies.
  McIlraith and Son
• Lead directly to the DAML-S process ontology

5

• 2 Bringing Services onto the Semantic Web
• Now keyword-based search, unstructured material,
  natural languages
• Changing the Web, Soon it will be possible to
  access Web resources by content rather than just
  by keywords DAML
• Emerging of DAML for better specifications of
  relationships and ontologies within Web pages to
  facilitate their automated parsing and thus the
  intelligent use of data present on the Web.
• DAML-S is the part of DAML and ontology effort
  dedicated to supporting Web services

6

• 2 Bringing Services onto the Semantic Web
• DAML-S
• XML-based language grounded in description logic
• Formal underlying semantics
• Design tool facilitate creation of markup and
  ontology
• Supporting technologies, automatic ontologies
  mapping
• Focus on activity-based websites action
• Commercial world
• WSDL ( Web Services Description Language)
• UDDI ( Universal Description, Discovery and
  Integration )
• WSFL ( Web Services Flow Language)
• WSIF, PBML, PBEL4WS, XIANG, etc.

7

• 2 Bringing Services onto the Semantic Web
• Comparison
• Commerce effort on
• Standardization on registration and look-up
  mechanisms
• platform-independent interoperability
• interchange of syntactically well-defined
  document types
• Academes are concerned with
• providing greater expressiveness in describing
  the characteristics of services in a way that can
  be reasoned about
• in support of more fully automated service
  discovery, selection, invocation, composition,
  and monitoring

8

• 2 Bringing Services onto the Semantic Web
• DAML-S
• profile describes what the service does
• process model tells how the service works
• grounding tells how the service is used
• DAML-S profile
• advertising, discovery, and matchmaking for
  service
• for a service-seeking agent to determine whether
  the service meets its needs
• organized into ontology-based taxonomies
• Description logic approach for matchmaking

9

• 2 Bringing Services onto the Semantic Web
• DAML-S process model
• Designed to support composite services
• Inputs, outputs, precon, effects
• Descript complex service beaks down into simpler
  component processes
• For service-seeking agent
• Perform a more in-depth analysis of whether the
  service meets its needs
• Web Service Composition (WSC) description
• During the course of the service enactment, to
  coordinate the activities of the different
  participants
• Monitor the execution of the service

10

• 2 Bringing Services onto the Semantic Web
• DAML-S grounding
• specifies details of how an agent can access
  service
• communications protocol( RPC, HTTP-FORM, CORBA,
  SOAP,RMI, KQML)
• service-specific details, such as port numbers
• exchanging data type with the service
  (serialization techniques)
• Summary
• Profile and process are abstract specifications
• Grounding provides concrete specification of
  implementing details

11

• 2 Bringing Services onto the Semantic Web
• DAML-S Processes
• Provide a basis for specifying the behavior of a
  wide range of services and draws on the following
  work
• AI on standardization of planning languages
  (PDDL)
• programming languages distributed systems (Mobile
  Agent)
• emerging standards in process modeling Work flow
  technology (PSL,WFMC)
• Modeling verb semantics and event structure
  (reasoning about action)
• previous work on action-inspired Web Service
  markup
• AI on modeling complex actions
• Agent communication languages (KQML, OAA) and
  Multi-Agent infrastructure (Coordination)

12

• 2 Bringing Services onto the Semantic Web
• DAML-S Processes
• Atomic processes
• Units of invocation
• Execute and return in a single step with a
  grounding
• Simple processes
• Like atomic processes, single-step executions
• Unlike atomic processes, no grounding
• Provide a means of abstraction, black box
• Realized By an atomic process
• Expand to a composite process, provide a
  simplified representation for planning and
  reasoning engines that dont need the full
  details of decomposition
• Composite processes
• Constructed from atomic, simple, composite
  processes
• Control constructs specify the structure, glass
  box
• Composed of conditions and process components

13

• 2 Bringing Services onto the Semantic Web

14

• 3 Semantic Web and Software engineering
• Semantic Web Service V.S software component
• Black box
• Interface
• Action selection V.S component composition
• Commit base
• temporal logic
• statechart

15

• 4 Web Services as Agent Behavior
• Service are behavior modules
• Service modules V.S agent action
• Service composition V.S high level action

16

• 5 Practical Reasoning Agents Architecture Nuin
• 5.1 Background and Motivations
• practical architecture for deliberative agents
  for the Semantic Web
• commonly studied architecture for deliberative
  agents is the belief-desire-intention (BDI) model
• build agents for the Semantic Web for
• key tools and techniques being developed under
  Semantic Web research
• design tools, ontology reasoners and persistent
  RDF stores
• Semantic Web will make knowledge for agents
• drive for Semantic Web pull agent tools and
  applications into closer collaboration
• Design the architecture to interoperate with the
  emerging standards and tools of Semantic Web.

17

• 5 Practical Reasoning Agents Architecture Nuin
• 5.2 Foundations of Approach
• BDI Foundations
• Explicit mental attitudes manage agents internal
  structures and external environment and achieve
  balance between optimal behavior and resource
  limitations
• Create practical agents balance
• Build upon a formal, but non-computable logical
  model
• Implementation approach without well-formed
  properties
• AgentSpeak(L) abstraction of reactive planning
  model
• Agent a set of first-order terms formed beliefs
  and a set of plans
• Plan e b1 ? ? bm ? h1 hn
• external and internal events
• Proof theory referred to papers of Rao

18

• 5 Practical Reasoning Agents Architecture Nuin
• 5.2 Foundations of Approach
• BDI Foundations
• Limitations of AgentSpeak(L)
• Intentions are commitment to a particular plan to
  handle an exogenous event. There is no sense of
  committing to achieve a goal held by, or shared
  with, another agent, nor of committing to
  maintain a state of the world.
• All choices in the AgentSpeak(L) interpreter are
  non-reversible. If a plan body fails, the whole
  plan (and in a naïve implementation the whole
  interpreter) simply fails.
• choice functions -- agent comes down to which
  event to focus on, and which plan to intend to
  handle that event. (for example functions SO and
  SI, in ASL) not discussed in Raos paper.

19

• 5 Practical Reasoning Agents Architecture Nuin
• 5.2 Foundations of Approach
• Architectural Foundations
• Re-use existing components (platform)
• FIPA, OAA
• Design pattern interface-driven design
• Factory pattern, abstract key platform functions
  into interface
• Semantic Web Foundations
• RDF triples are the basic representation of
  knowledge
• ontological information is encoded in DAMLOIL or
  OWL which extends the representational capability
  of RDF
• knowledge sources are openly available and
  decentralized, HTTP(SOAP,KQML,RMI,RPC)

20

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture the only start-up
  parameter that an agent requires is a URL from
  which it can retrieve its configuration.

21

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Knowledge-representation model
• Representation vocabulary
• Logical sentences are formed from the usual first
  order connectives. integers, reals, Booleans,
  strings
• Many syntaxes may be parsed using Java interfaces
• S-expressions (KIF-like)
• a Prolog-like syntax with infix operators
• binary predicates are translated from RDF sources
• Knowledge-source and reasoners
• Knowledge-source(KS)
• Store logical sentences, in-memory or databases
• Associated with at least one reasoner
• Reasoner support services
• Core services matching, identification,
  query,get meta
• Backward chain, forward chain, updateable --
  optional

22

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Knowledge-representation model
• Knowledge-source and reasoners
• Meta-data on KS is RDF model
• Predefined nuin configuration ontology
• Provide means to agent reasoning its own
  capabilities
• Compose multiple KS when forward-chaining
  inferences
• Dispatcher route query to different reasoners
  Frank
• Context added to query interface encapsulates
  strategy for delegating queries to other KS
• Resource-bounded Reasoning
• Predicate logic powerful, computationally
  undecidable
• Some queries will never terminate, consume
  resources
• representation systems are weaker than full
  first-order logic, computationally more
  tractable.
• description logics ,widely used in ontology
  research

23

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Agent mental states
• Beliefs
• Not sufficient for complex compositions of modal
  operators
• Folding of modalities into the KS label is
  sufficient form many practical applications
• Desires
• First-order sentences available to the choice
  functions
• Model social attitudes of agents in future
• Intentions
• Commitment to a course of action
• Triggered by various conditions
• Events
• Sensing of the world
• Recent-event history

24

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Interpreter

25

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Operational details
• Agent configuration
• semantic web technology RDF model
• expressed as an RDF document with a resolvable
  URL
• defined an ontology to represent configuration
  options
• Only need the URL of model to start up flexible
  apaptable
• Services
• Commit to FIPA abstract architecture (FAA)
• Define an abstract service (message sending)
• Interoperation with agent middleware
• Define service adapters abstract service lt-gt
  platform
• Web service deployment binding to abstract
  service

26

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Message Passing
• Abstract service, have autonomous control over
  their own behavior
• KQML and FIPA-ACL provide standard encodings
• Abstract messaging model
• Has attributes to, from, ontology, content,
  reply-with, etc.
• Messaging service
• Create, reply, encode, send ,suspend, extract
  content
• Directory service
• Registration and name resolution
• No federate or distributed queries

27

• 5 Practical Reasoning Agents Architecture Nuin
• 5.3 Nuin Agent architecture
• Examples

28

• 5 Practical Reasoning Agents Architecture Nuin
• 5.4 Evaluation
• Nuin
• A practical toolkit for developing deliberative
  agents for Semantic Web applications
• Higher-level agent behavior of agents
• not infrastructures
• Alpha of Nuin is developing
• JAM
• BDI based java implementations of PRS
• Neither address semantic web reasoning,
• or integrate with agent middleware
  platforms
• JADE
• Code agent behaviors using production rules,
  automata
• Useful, but not correspond to an agent theory

29

• 6 Other Semantic Web Service
• ITTALKS
• As part of UMBC's role in the DAML Program
• A web portal offering access to information about
  talks, seminars related to IT
• Utilizes DAML for knowledge base representation,
  reasoning, and agent communication

30

• 6 Other Semantic Web Service
• ITTALKS

31

• 6 Other Semantic Web Service
• Retsina Semantic Web Calendar Agent
• CMU Response to SWWS2001 Challenge
• Import RDF Schedules into Outlook
• Parses RDF Schedules
• Based on iCal Ontology
• Outlook-Agent integration
• RDF Schedules in to Outlook
• using OLE Automation

http//www.daml.ri.cmu.edu/Cal
32

• 6 Other Semantic Web Service
• Retsina
• Allows user to browse SW schedules events
• Displays event, location and attendee information
• Supports additional actions based on available
  information
• E.g. email or visit web page if information is
  available
• Supports serendipitous exploration
• Uses agent discovery (DAML-S) to locate context
  dependent agents
• Imports schedules into MS Outlook
• User can select schedules to import into Outlook
  Calendar
• Can receive KQML requests to autonomously import
  schedule without invoking browser

33
(No Transcript)
34
Importing Schedules into MS Outlook
35

• 7 Open issues
• how to extend the reasoning capabilities of the
  agents
• how to facilitate effective co-operation between
  agents with mental states
• Semantic Web agents is to explore the
  relationships between the strong, but
  theoretically intractable, reasoning of BDI
  agents and the weaker, computationally tractable,
  reasoning embodied in OWL or DAMLOIL in a open,
  changeable environment such as the Semantic Web.

36

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Theories of agency that relate intentions and
  knowledge of agents with their actions
• Knowledge of facts --- Know that
• Knowledge of actions --- know how
• Historical Remarks
• Ryle ltltThe Concept of Mindgtgt
• Stupidity not knowing how
• Ignorance not knowing that
• Know to perform certain complex actions, not know
  that he does a certain specific sequence
• Agre reactive systems in AI
• centrality of planning
• Not address the notion of know how

37

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Historical Remarks
• Moore
• A formal modal logic of knowledge
• Study knowledge required to execute plans
• Execute a conditional plan require knowing
  whether its condition true
• Not address the notion of know how
• philosophical work
• Brown Actions as Exercised Abilities
• Chellas Seeing To It That
• Singh Strategic Know-How
• Segerberg Bringing It About

38

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Motivation
• Actions
• basic actions correspond to the atomic abilities
  of the agent.
• High-level actions can be specified indirectly as
  propositions that an agent can achieve through a
  combination of lower-level basic actions.
• Although basic actions can be performed directly
  if the agent has the corresponding physical
  ability, performing complex high-level actions
  frequently requires not only the physical ability
  to perform the underlying basic actions, but also
  the knowledge to select the appropriate actions
  to perform at each stage of the complex action.
• High-level actions is another view of
  capabilities (know-how, ablities, .etc)

39

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Theory Frameworks
• Know How -- Strategic Know-How
• Munindar P. Singh. Multiagent Systems A
  Theoretical Framework for Intentions, Know-How,
  and Communications. Springer-Verlag, Heidelberg,
  1994.
• uses strategies describe at a high level actions
• Explicit actions of the agents organized into
  trees.
• Foundation Branching time temporal logic

40

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Theory Frameworks
• KARO
• B. van Linder, W. van der Hoek J.-J. Ch. Meyer,
  Formalising Abilities and Opportunities of
  Agents, Fundamenta Informaticae 34(1, 2), 1998
• Knowledge, Abilities, Result, Opportunities
• Can, cannot,
• Foundation Dynamic logic and deontic logic

41

• 8 Agent Capabilities (Know-How, Abilities .etc)
• Theory Frameworks
• BDI C
• Agent Capabilities Extending BDI Theory Lin
  Padgham
• I system in BDI IC system
• Foundation CTL logic Branching time and action