Distributed Artificial Intelligence I

1
Distributed Artificial Intelligence I
Gerstner Laboratory for Intelligent Decision
Making and Control

• Michal Pechoucek

2
Distributed Artificial Intelligence

• Torsun Distributed Artificial Intelligence is a
  branch of Artificial Intelligence that is
  concerned with cooperative problem solving by a
  decentralized group of collaborating agents.
  Agents are logically independent and autonomous,
  they reason, plan, learn and communicate
• distributed system of a collective architecture
  (domain of distributed problem solving),
  decomposes the problem among collection of
  autonomous, cooperating, knowledge sharing
  agents, case specific
• distributed system of an integration architecture
  (a multi-agent systems) integrates number of
  heterogeneous agents of a different nature,
  complexity reduction

3
Agent

• Jennings "An agent is a computational system,
  situated in some environment, that is capable of
  intelligent, autonomous action in order to meet
  its design objectives."

intelligent agent
collaborative learning agent
cooperate
learn
interface agent
collaborative agent
autonomous
4
Agents Intelligence

• Reactivity ability to provide intelligent
  responses to percepts and agent senses from the
  environment (user interface)
• Proactivity ability to maintain agents long
  term intention, organize its behavior in order to
  meet targeted goals
• Social Intelligence ability to perform
  reasoning about other agents abilities,
  intentions, current status and possible future
  course of actions
• Reactive Agent agent presenting a reactive
  intelligence only
• Cognitive Agent manipulate symbolic model about
  their own status, and capabilities (proactivity)
  and about its collaborative environment

5
Agents Abstract Architecture

• Agents Communication Wrapper
• translation to and from ACL (Agent Communication
  Language)
• physical connection and responsibility delegation
• perception action
• social model

6
Agent is not just a Program

• an agent in the context of Distributed Artificial
  Intelligence is a member of a multi-agent
  community, where its behaviour and logic behind
  reasoning has to bee seen from the multi-agent
  perspective
• freely interact, interaction among agents is
  emergent
• can group into coalitions, teams, they can
  benefit from this
• do not have to be benevolent, have free will, can
  cheat
• can leave/join the community
• can adapt and improve their social role
• however there are also other agents such as
  migrating agent, viruses, information seekers
  who are not members of multi-agent community in
  the above sense

7
Agent is not an Object

• Objects and Agents both
• hide information, allow distributed tractability
• provide interaction mechanisms among system
  components
• Objects unlike Agents
• are passive, do not operate unless activated
• are too grainy structured (ABS System
  Community Agent)
• exchange too primitive imperative message,
  (agents communicate declarative knowledge)
• do not support flexible organisational
  relationship, the systems object model is fixed
• present designed behaviour, while agent are
  expected to form they own patterns behaviour
  emergently (from interaction)

8
Reactive Agents

• Reactive agents are agents that do not contain
  any symbolic knowledge representation (ie no
  state, no representation of the environment, no
  representation of the other agents, ...). Their
  behaviour is simply defined by a set of
  perception-action rules.
• Subsumption architecture System of ordered
  layers of perception-action rules, where the
  lower layers take precedence.
• Example of rock sample collecting robots
  (Steels), inspired by ants
• if detect an obstacle then change direction
• if carrying samples and at the base then drop
  samples
• if carrying samples and not at the base then
  travel up
• if detect a sample then pick sample up
• if true then move randomly

9
Cognitive Agents

• Cognitive Agents are agents with an explicit
  knowledge representation of own capability, other
  agents, the environment, etc.
• There are various models of agents cognitive
  states (differ in purpose, generality, )
• BDI (Belief Desire Intention)
• Joint Intentions Theory
• 3bA (Tri-Base Acquaintance Model)

10
Belief Desire Intention

• Framework for reasoning about formal abstract
  models of mental states
• Contains representations (as objects, data
  structures, or whatever) of
• beliefs, which constitute its knowledge of the
  state of its environment (and perhaps also some
  internal state),
• desires, which determine its motivation what it
  is trying to bring about, maintain, find out,
  etc.,
• intentions, which capture its decisions about how
  to act in order to fulfil its desires
• Intentional Attitudes
• informational attitude knowledge, belief
• pro-atitude desire intention, choice,
  commitment

11
Belief Desire Intention cont

• has a control mechanism which ensures that
• its beliefs change over time in response to
  external events,
• its intentions determine and cause sequences of
  actions to be taken,
• its intentions change over time as a result of
  its beliefs changing, its desires becoming
  fulfilled or failing to be fulfilled, actions
  being taken, and new events being received.
• representation of intentional attitudes
• limitation of propositional logic,
• modal logic, meta-logics, dynamic logic of
  action, temporal logic
• Possible Worlds Semantics, Interpreted Symbolic
  Structures

12
Example of BDI Architecture IRMA
13
Interaction among Agents

• Organization
• an arrangement of relationships between
  individuals or components, division of tasks,
  distribution of roles, and contribution-awards
• Cooperation
• sharing responsibilities in satisfying shared
  goal and generating mutually dependent roles in
  joint activities
• Coordination
• management of agents activities so that they
  coordinate their deeds with each other in order
  to share resources, meet their own interests

14
Interaction among Agents cont

• Negotiation
• information exchange aimed at resolving conflict
  of access to resources, different solutions to
  the same problem or goal conflicts
• Communication
• information, knowledge and request exchange via
  mutually agreed ACL
• Benevolence
• agent are benevolent if they will agree to
  cooperate in asked/required

15
Agent Communication Language

• In order to ensure agent interoperability,
  mutually agreed communication protocol (ACL) must
  be provided
• Mutual knowledge understanding
• translating from one knowledge representation
  language into another
• sharing of semantics (and often pragmatics)
• Inter Agent Communication
• transport protocol (e.g. TCP/IP, SMTP, HTTP, )
• communication language
• interaction protocol

16
Knowledge Sharing Effort

• is an initiative addressing agent
  interoperability and knowledge sharing
• Ontolingua - a software tool and methodology for
  - means for sharing semantics (and often
  pragmatics) of represented knowledge
• Knowledge Interchange Format (KIF) - an
  inter-lingua for translations between different
  knowledge representations
• Knowledge Query and Manipulation Language (KQML)
  - a language for communicating attitudes about
  the shared knowledge.
• Ontologies-KIF-KQML is sometimes denoted as ACL.
  An ACL message is a KQML expression where
  arguments are KIF sentences formed from terms
  from appropriate ontologies.
• Foundation for Intelligent Physical Agents
  Organization (FIPA).

17
KQML Message
communication layer

• Message Struture
• (
• performative evaluate
• sender pma-243
• content (operation d021 (deadine 22)
  (costlimit 22)
• receiver pa-685
• reply-with (id 8567)
• language KIF
• ontology transmitter-production
• )

message layer
content layer
18
Selected KQML Performatives
transport-adderss announcement of an agents
changing address register agents
registration ask querry for agents
knowledge achieve request for change of agents
knowledge unachieve undo achieve subscribe
subscription for monitoring change of an
information advertise advertisement of a
service evaluate request for evaluation tell
general announcement ask-all request
broadcast ask-if querry for effects of a
hypotehtical request reply, sorry reply to
performative
19
Basic Models of Communication

• Broadcasting of a task announcement
• autonomous communication
• communication intensive
• Central Communication Agent
• well organized, saves communication
• central, fragile, communication bottleneck
• Acquaintance Models
• model of the environment in an abstract sense

20
Broadcasting a Task Announcement
21
Central Communication Agent
22
Utilisation of an Acquaintance Model
23
Basic Models of Communication

• Broadcasting of a task announcement
• autonomous communication
• communication intensive
• Central Communication Agent
• well organized, saves communication
• central, fragile, communication bottleneck
• Acquaintance Models
• model of the environment in an abstract sense

24
Tri-Base Acquaintances Model

• the motivation economize communication traffic
  thru agents mutual awareness
• 3bA is a knowledge model located in agent
  wrappers models an agents social neighborhood
• the methodology contains
• knowledge structures and
• knowledge maintenance mechanisms

25
3bA Model Knowledge Structures
26
3bA Model Maintenance Mechanisms

• task base
• plan construction
• plan maintenance
• plan execution
• state base
• subscribe-advertise
• post-planning back-propagation mechanism
• communication shift
• cooperator base
• Meta-agent

TB(A) ? ?PRS(A), PLS(A)?, PRS(A) ? ?(T,A)
and PLSt(A) ? áT, á?s, B?s? S,O,C,
Trust(T)ñ?T? ?t(A)
SB(A) ? ?AS(A), TS(A)?, where AS(A) ? ?B,
Cap(B), Load(B), Trust(B)?B? ?t(A)
CB(A) ? ?B, Addr(B), Lang(B), ?(B)?B? ?(A)
27
3bA Model Parameters

• agent A cooperation neighborhood in time t
• set of currently subscribed agents
• ?t(A) ? ?(A) ? ?
• meta-agent
• agents scope of reasoning
• number of preprepared plans in the PLS
• ?t(A) ? ?(A) ? S
• meta-agent learning,
• various planning phases

28
ProPlanT Project

• PROPLANT (Production Planning Tool) multi-agent
  system for project driven-production planning
• initiated by Eureka 1439 PVS98 project
• open architecture combining C platform with
  heterogeneous software Prolog, Java, CLP, Lisp,
  Clips,
• communication via sockets - TCP/IP protocols,
  KQML, KIF, FIPA standards

29
Taxonomy of ProPlanT Agents

30
ProPlanT Architecture
project planning agent
project managing agent
production agent
31
Meta-Agent

• an independent agent observing the community
• unlike information brokers, facilitators it is
  not central
• passive role visualisation (community
  structure, workflow, distributed plans, etc.)
• active role affects community operation
  (updates contents of agent bases)
• provides efficiency consideration, maintenance of
  parametrs

32
Active Role of the Meta-Agent

• agent's termination
• record retraction from the cooperator base
• agent's loss/acquiring of its capability
• plan section redesign, replannig
• record retraction from the cooperator base
• agents creation
• registering agent, autonomous replanning
• agents properties change
• plan section redesign, replannig

33
ProPlanT in Industry

• Tested in Tesla-TV (within Eureka project)
• ExPlanTech (EC IST Trial Project)
• Pattern Shop Liaz
• Chatzapulous Packaging
• Future Internet Integration, ERP integration

34
Experiments and Testing of the 3bA Model

• communication savings in a production planning
  system

35
Experiments and Testing cont

• request frequency requirements
• 1/f ? tc ti,
• degree of freedom

?(?) number of messages number of messages number of messages
?(?) m(tri-base) m(broadcast) m(maint)
0 21 21 12
2 21 32 12
7 21 47 12