Spring Term 200203 ismailtk'unilinz'ac'at

1
Spezielle Kapitel aus TelekooperationIntelligent
Agents

• Ismail Khalil Ibrahim
• Summer Semester 2003

2
Course Unit 5

• Intelligent Agents Communication and
  Communication Languages ...

3
Interaction Communication

• Interactions occur because agents exist and act
  in close proximity
• shared environment
• shared resources
• Communication occurs when agents send messages to
  one another with a view to influence beliefs and
  intentions.
• Implementation details are irrelevant
• via communication links
• through shared memory
• because of shared conventions
• signalling vs talking
• body language

4
Communication

• The primary reason for communication among agents
  is to coordinate activities
• Agents may coordinate without communication,
  provided they have models of others behaviour
• Communication involves the dimensions of
• who
• what
• when
• how (resources and protocol)
• why
• To facilitate cooperation, agents often need to
  communicate their intentions, goals, results, and
  state

5
Coordination and cooperation

• Coordination is a property of interaction among
  agents performing some activity in a shared
  state. The degree of coordination is the extent
  to which they
• avoid extraneous activity
• reduce resource contention
• avoid livelock
• avoid deadlock
• maintain safety conditions
• Cooperation is coordination among
  non-antagonistic agents
• Typically, agents must
• maintain models of other agents
• develop model of future interactions

6
Communication vs. computation

• Communication is generally more expensive and
  less reliable
• re-computing is often faster than requesting
  information over a communication channel
• communication can lead to prolonged negotiation
• chains of belief and goal updates caused by
  communication may not terminate
• Communication is qualitatively superior
• information cannot always be reconstructed
  locally
• communication can be avoided only when the agents
  are set up to share all necessary knowledge. This
  is a very limiting assumption which cannot be
  practically achieved in most interesting cases

7
MAS communication protocols

• A MAS protocol is specified by the following
  fields
• sender
• recipient(s)
• language in the protocol
• actions to be taken by the participants at
  various stages
• A MAS protocol is defined above the ISO/OSI
  transport layer
• not about bit patterns
• not about retransmissions or routing
• A MAS protocol is defined at the knowledge level
• involves high-level concepts, such as
• commitments, beliefs, intentions
• permissions, responsibilities, requests

8
Levels of agent communication
Agent Communication Language FIPA-ACL, KQML, ...
Content language XML, KIF, WML, HTML, ...
Agent middleware
Encoding schema Java serialized object, String,
Bytecode
Physical protocols HTTP, IIOP, TCP/IP, SMTP, Fax,
Phone, WAP, ...
9
Classification of messages

• Syntactic
• distinguish messages based on grammatical forms
  in natural language
• Semantic
• distinguish messages based on a notion of
  intrinsic meaning
• e.g., prohibitive is different from directive,
  despite their syntactic similarity
• Use-based
• distinguish messages based on their roles in
  specific classes of protocols
• e.g., assertion is different from acknowledgment

10
Exchanging information

• How can an agent tell another agent something?
• Send the information in a message (message
  passing)
• Write the information in a location where the
  other agent is likely to look (shared memory,
  blackboard)
• Show or demonstrate to the other agent (teaching)
• Insert or program the information directly into
  the other agent (master-slave, controller-controll
  ee)

11
Requesting information or services

• How can one agent obtain information or request a
  service from another agent?
• Ask the other agent (message passing)
• Read a location where the other agent is likely
  to write something (shared memory or blackborad)
• Observe the other agent (learning)
• Access the information directly from the other
  agent (brain surgery)

12
Methods of agent communication

• Blackboard
• information available for all
• no direct communication
• simple architecture

• Message
• Direct exchange
• Common language
• Conversation
• Sequence of messages

Agent
Agent
Blackboard
Agent A (Sender)
Agent B (Receiver)
Message
Agent
Agent
Agent
Agent
13
Speech act theory Austin (1962), How to Do
Things with Words

• Speech act theory, initially meant for natural
  language, views communications as actions.
• Considers three aspects of a message
• Locutionary act
• uttering a syntactically acceptable sentence
• how is the message phrased?
• e.g., it is hot here, or turn on the AC
• Illocutionary act
• communicating of the speaker intentions to the
  hearer
• how is the message meant by the sender or
  understood by the receiver
• e.g., a request to turn on the AC, or an
  assertion about the temperature
• Perlocutionary act
• Bringing about of the effect on the hearer of the
  utterance
• how does the message influence the receivers
  behaviour
• e.g., turn on the AC, opens the window, nods her
  head, ignores the speaker

14
Speech act classification Searle (1969),
Speech Acts, Cambridge University Press

• Representative
• Represent some state of affairs
• Asserting, claiming, describing, informing, etc.,
• Commissive
• Commit the speaker to some future course of
  action
• Promising, agreeing, threading, inviting,
  offering, swearing, volunteering, etc.,
• Directive
• Intention to get the hearer to carry out some
  action
• Commanding, requesting, daring, asking, begging,
  forbidding, advising, etc.,
• Declaration
• Bring about a state of affair
• Arresting, marrying, naming, etc.,
• Expressive
• Indicate the speakers psychological state of
  mental attitude
• Greeting, congratulating, thanking, apologizing,
  etc.,

15
Speech act components

• In general a speech act can be seen to have two
  components
• a performative verb
• e.g. inquire, inform,
• the propositional content
• e.g. shut the door.

16
examples

• Informing that the door is shut
• performativeinform
• contentShut the door
• speech act The door is shut!
• Inquiring if the door is shut
• performativeinquire
• contentShut the door
• speech act Is the door shut?

17
Agent Communication

• Speech acts form messages in an Agent
  Communication Language (ACL)
• An ACL is a high-level language whose primitives
  and structures are expressively tailored to
  support the exchange of messages amongst multiple
  agents
• An ACL exists in a logical layer on top of
  existing infrastructures such as TCP/IP, HTTP, or
  IIOP.
• The two prevalent ACLs
• KQML (Knowledge Query and Manipulation Language)
• FIPA ACL (Foundation of Intelligent Physical
  Agents).

18
KQML

• Developed by the DARPA knowledge sharing
  initiative
• based on the idea that communication requires a
  common language
• requires that common language is divided into
  syntax, semantics and pragmatics, using
• KIF - syntax of content
• Ontolingua - semantics of content
• speech acts - pragmatics of communication

19
KQML Syntax

• KIF (Knowledge Interchange Format)
• is based on first-order logic set theory
• encodes declarative knowledge
• Lisp like syntax, e.g.
• see http//logic.stanford.edu/kif/dpans.html for
  more details.

20
KQML Semantics

• Ontology a common vocabulary and agreed upon
  meanings to describe a subject domain.
• Ontolingua is a language for building,
  publishing, and sharing ontologies
• ontologies can be automatically translated into
  other content languages, including KIF, Prolog,
  etc.
• the main language included primitives for
  combining ontologies.
• see www.ksl.stanford.edu/software/ontolingua/ for
  more details.

21
Ontologies

• The philosophical study of the
• nature of being
• Specification of ontological commitments
• Description of concepts
• and relationships used
• to interact

22
Ontologies

• Ontologies define classes, functions, object
  constants, and axioms to constrain meaning
• Allows unambiguous interpretation of logical
  sentences

23
Example Ontology 1
Vessel
Ship
Boat
Sail boat
Motor boat
Hull
Engine
24
Example Ontology 2
25
Translation
How do we map objects from one ontology to
another?
26
KQML Pragmatics

• The pragmatics define the performative verbs, for
  example
• ask-if (is it true that)
• perform (please perform the following action)
• tell (it is true that )
• reply (the answer is )

27
Performatives

• Basic query evaluate, ask-if, ask-about,
  ask-one, ask-all
• Multi-response query stream-about, stream-all,
• Response reply, sorry
• Informational tell, achieve, cancel, untell,
  unachieve
• Generator standby, ready, next, rest, discard,
  generator
• Capability definition advertise, subscribe,
• monitor, import, export
• Networking register, unregister,
• forward, broadcast

28
A KQML Message

• Msg above represents a single KQML speech act
  described by a list of attribute/value pairs e.g.
  content, language, from, in-reply-to.

29
A KQML Dialogue

• Consider agents A and B talking about the
  prices of books bk1 and bk2
• A to B (ask-if (gt (price bk1) (price bk2)))
• B to A (reply true)
• B to A (inform ( (price bk1) 25.50))
• B to A (inform ( (price bk2) 19.99))
• For convenience msg format above is simplified
  and attribute/value pairs for ontology etc. are
  omitted.

30
FIPA - ACL

• FIPA (Foundations of Intelligent Physical Agents)
• its purpose is to promote the success of emerging
  agent-based applications, services and devices.
• its goal is pursued by making available in a
  timely manner, internationally agreed
  specifications for interoperable agent-based
  applications, services and devices.
• see http//www.fipa.org/ for more details.

31
Agent Management in FIPA

• FIPA provides a reference
• model for agent
• creation
• registration
• location
• communication
• migration and
• retirement.

32
FIPA ACL

• Like KQML, FIPA ACL is based on the idea of
  Speech Acts
• agree
• cancel
• cfp
• confirm
• disconfirm
• failure  

• inform
• inform-if
• inform-ref
• not-understood
• propose
• query-if
• query-ref
• refuse
• reject-proposal
• request
• request-when
• request-whenever
• subscribe

33
Ontologies in FIPA
Ontology Agents provide ontology services, such
as translation, to FIPA agents
Ontology servers provide shared ontologies to
agents and other systems
34
FIPA ACL Example
ACL message
inform
(
Note FIPA uses a different content language
from KQML called SL which is based on modal
logic.
sender

agent1
receiver


hpl-auction-server
content

(price (bid good02) 150)
in-reply-to

round-4
reply-with

bid04
language


sl
ontology


hpl-auction
)
35
Example Query-Protocol

• The protocol states that
• IF an agent asks another agent if a proposition
  holds or not
• THEN the participant must reply with answer
  informing the initiator, or may refuse to answer,
  or fail to answer, or failing to understand the
  question.

36
FIPA Interaction Protocols

• Ongoing conversation between agents often fall in
  typical patterns
• typical patterns of message exchange are called
  interaction protocols
• FIPA pre-specifies a number of protocols.
• Advantage ease of implementation (an agent can
  engage in meaningful interactions simply by
  following an interaction protocol).

37
Facilitator broker
ask(x)
Agent 1
Agent 2
tell(x)
Agent 2
Agent 1
1. advertise(ask(x))
5. tell(x)
3. ask(x)
2. broker(ask(x))
Facilitator
4. tell(x)
38
Facilitator matchmaker
4. ask(x)
Agent 2
Agent 1
5. tell(x)
1. advertise(ask(x))
3. reply(x)
Facilitator
2. recommend(ask(x))
39
Facilitator feeder
Agent 2
Agent 1
2. tell(x)
3. tell(x)
Facilitator
1. subscribe(ask(x))
40
KQML examples

• (ask-one
• content (PRICE IBM ?price)
• receiver stock-server
• language standard_prolog
• ontology NYSE-TICKS
• )

41
KQML examples

• (subscribe
• content (stream-all
• content (PRICE IBM ?price))
• receiver stock-server
• language standard_prolog
• ontology NYSE-TICKS
• )

42
KQML Examples

• (monitor
• content (PRICE IBM ?price)
• receiver stock-server
• language standard_prolog
• ontology NYSE-TICKS
• )

43
Discussion questions

• What architectures do the different languages
  support?
• What kinds of communication do you have in your
  project?
• What language to use for your project (if any)?
• What standards do you think are necessary when
  talking about agent communication languages in
  the future?

44
Corporate Agent System

• Employees have a personal agents which gather
  requested files before morning.
• Agents aim is to minimise the cost for the user.
• User agents make the decision on how to achieve
  goal
• Independent. vs Collaborate work.
• Employer wishes to minimise costs
• Users normally request the same files.
• Provides a management agent which negotiates more
  efficient solutions.

45
Employee Agent System

• Agents are given a list of files to download
  before the morning.
• The agent first calculates cost of working
  individually.

Get these documents for me
User A
46
Agents calculate individual costs.
5
4
4
Location and Cost of each file Doc 1
http//www.cs.tcd.ie/ - cost 1 unit. Doc 2
http//www.ianbrown.co.uk/billiejean.wav - cost 1
unit. Doc 3 http//www.deankane.usa/ - cost 2
units. Doc 4 ftp//www.nokia.se/. - cost 2
units.
47
Agents subscribe to Manager

• Message passes the files that this agent needs.
• (subscribe
• sender A
• receiver manager
• content need file 1,2,4
• ontology doc-search
• )
• B need file 2,3,4 C need file 3,4

48
Subscribe Stage
need file 1,2,4
need file 3,4
Management Agent
need file 2,3,4
49
Management Agent System

• Maintains a central black board system.
• Agents submit the files that they need.
• Manager splits work so that no two agents
  retrieve the same file.
• Costs of this service is 1 unit for the user
  agents.
• Manager suggests work to each agent.
• The file that the agent is expected to fetch.
• Negotiates a solution with agents.
• They Accept or Reject solution.

50
Central Black Board System
If an agent fetches a file and places it in the
central store, he can then access all the other
files for free.
51
Unfair manager proposes solution

• Agent A does all the work.
• ( propose
• sender manager
• receiver A
• content get 1,2,3,4 cost 9
• in-reply-to unfair_prop
• )
• B get null, cost 1. C get null, cost 1.

52
Unfair Propose Stage
get 1,2,3,4 cost 9
get null, cost 1
Propose
Management Agent
get null, cost 1
53
User Agents Utility calculation

• Agents calculate the utility of proposed deal.
• Utility local individual cost - management
  agent cost.
• Agents try to maximise their utility.
• Agent A case
• Utility 4 9 -5
• REJECT.

get 1,2,3,4 cost 9
54
Agents accept / reject proposal

• Agent A not happy
• ( reject-proposal
• sender A
• receiver manager
• content no chance
• in-reply-to unfair_prop
• ontology doc-search
• )

55
Unfair Accept / Reject Stage
Utility -5
Reject
Utility 3
Accept
Accept
Management Agent
Utility 4
56
Manager proposes new solution

• Manager propose fair allocation to agents
• ( propose
• sender manager
• receiver A
• content get 1,2 cost 3
• in-reply-to fair_prop
• )
• B get 3, cost 3. C get 4, cost 3.

57
Fair Propose Stage
get 4, cost 3
get 1,2, cost 3
Propose
Management Agent
get 3, cost 3
58
Fair Accept / Reject Stage
Utility 1
Accept
Utility 1
Accept
Accept
Management Agent
Utility 2
59
(No Transcript)