CPSC 609 Foundations of Multi-Agent Systems Winter 2019 Jrg - - PowerPoint PPT Presentation

▶
cpsc 609 foundations of multi agent systems winter 2019
SMART_READER_LITE
LIVE PREVIEW

CPSC 609 Foundations of Multi-Agent Systems Winter 2019 Jrg - - PowerPoint PPT Presentation

Apr 21, 2023 381 likes •593 views

CPSC 609 Foundations of Multi-Agent Systems Winter 2019 Jrg Denzinger ICT 752 denzinge@cpsc.ucalgary.ca http://www.cpsc.ucalgary.ca/~denzinge/ 1. Introduction 1.1. Organization n Lectures: MoWe 15:30 n

slide-1
SLIDE 1

CPSC 609 Foundations of Multi-Agent Systems Winter 2019

  • Jörg Denzinger

ICT 752 denzinge@cpsc.ucalgary.ca http://www.cpsc.ucalgary.ca/~denzinge/

slide-2
SLIDE 2

Multi-Agent Systems

Jörg Denzinger

  • 1. Introduction

1.1. Organization

n Lectures: MoWe 15:30 n Assignments:

  • Midterm paper on a cooperation concept
  • Final paper describing a team of agents that work

together rescuing people buried under rubble after a disaster struck a city (agents would be acting within the ARES system)

slide-3
SLIDE 3

Multi-Agent Systems

Jörg Denzinger

Determining your grade

  • n Midterm paper
  • 30%

n Final paper

  • 30%

n Oral exam

  • 40%
slide-4
SLIDE 4

Multi-Agent Systems

Jörg Denzinger

More Information

n Course webside: http://pages.cpsc.ucalgary.ca/~denzinge/

  • courses/cs609-winter2019.html

n Internet n Recommended papers n Talk to me, ask questions, send me email.

slide-5
SLIDE 5

Multi-Agent Systems

Jörg Denzinger

1.2 Motivation

Avatar

Desire

Teamwork

Competition

?

Robo-Cup

Market mechanisms

Cooperation

Intentions Belief

negotiations

Blackboard Contract-net

Vickrey auction

Commitment Game theory Adversary

slide-6
SLIDE 6

Multi-Agent Systems

Jörg Denzinger

MAS and related areas (in Computer Science)

n Artificial Intelligence

  • Planning
  • Learning
  • Knowledge representation / logic

n Distributed Systems n Efficient Algorithms n Security n Robotic n (graphical) user interfaces

slide-7
SLIDE 7

Multi-Agent Systems

Jörg Denzinger

MAS and related areas (outside Computer Science)

n Business Science / organizational theory / market studies n Game theory n Psychology n Sociology n Arts n All kinds of Simulations n Soccer n ...

slide-8
SLIDE 8

Multi-Agent Systems

Jörg Denzinger

Goals of MAS (I)

In general: higher efficiency, faster answers, better solutions Especially: Synergy (the whole is more than the sum of its parts)

  • Additional goals:

n Adequately modeling given distributions (functional, spatial, etc.) n Improving adaptability, extendability, reliability of systems (F software engineering)

slide-9
SLIDE 9

Multi-Agent Systems

Jörg Denzinger

Goals of MAS (II)

n Easy integration of existing systems, achieving open systems (F software engineering) n More cost efficiency by specialization and distribution of tasks n More security and safety by autonomy and isolation n Usage of computer networks for AI applications n Simulating human approach to hard tasks: F computer (program) as colleague or part of a colleague

slide-10
SLIDE 10

Multi-Agent Systems

Jörg Denzinger

Basic problems (I)

n Modeling agents

  • degree of autonomy
  • degree of willingness to collaborate
  • degree of intelligence
  • complexity

n Construction of a community (society, organization)

  • f agents
  • communication
  • control
slide-11
SLIDE 11

Multi-Agent Systems

Jörg Denzinger

Basic problems (II)

  • different roles
  • division and assignment of tasks
  • distribution/assignment of resources

n Utilizing a community of agents

  • environment
  • constraints/conditions to keep fulfilled
  • efficiency
  • reuse of components
slide-12
SLIDE 12

Multi-Agent Systems

Jörg Denzinger

Solution approaches (I)

n Adaptation of human forms of organization

  • political structures
  • organization of business companies
  • groups in society

n Negotiation concepts for solving conflicts

  • detection of conflicts
  • determining scope of conflicts
  • protocol for solution process

n Game theoretical concepts, for example utility (payoff) matrixes

slide-13
SLIDE 13

Multi-Agent Systems

Jörg Denzinger

Solution approaches (II)

n Knowledge representation techniques F standards for information exchange, like FIPA initiative or KIF n Different concepts for modeling agents, for example modal logics, state automata, rule based modeling,… n Tools for realizing MAS

  • MAS languages
  • Test beds, like OLEMAS and ARES
  • Groupware

n ...

slide-14
SLIDE 14

Multi-Agent Systems

Jörg Denzinger

(Intended) Structure of the course (I)

1.Introduction

  • 1. Organization
  • 2. Motivation
  • 3. A little history

2.Single-agent systems

  • 1. Formal definitions and properties
  • 2. Modeling of agents

3.Multi-agent systems

  • 1. Formal definitions and properties
slide-15
SLIDE 15

Multi-Agent Systems

Jörg Denzinger

(Intended) Structure of the course (II)

  • 3.2. Interaction and cooperation concepts
  • 3. Competitive agent environments

4.Learning in MAS

  • 1. Learning in general
  • 2. From single-agent to multi-agent learning
  • 3. Examples
  • 5. Testing MAS
  • 6. Larger examples (if time permits)
slide-16
SLIDE 16

Multi-Agent Systems

Jörg Denzinger

1.3. A little history of MAS

1975 Hearsay-II, Blackboard systems (Erman, Lesser)

  • Beings: knowledge as interacting experts (Lenat)

1977 Actor model, message passing (Hewitt) 1980 First workshop on Distributed AI, MIT

  • Contract-net protocol, negotiations as cooperation

concept (Smith) 1981 IEEE Transactions on Systems, Man and Cybernetics special issue on DAI:

  • Organizational View on Distributed Systems (Fox)
  • The Scientific Community Metaphor (Kornfeld,

Hewitt)

slide-17
SLIDE 17

Multi-Agent Systems

Jörg Denzinger

1981 Functionally accurate cooperative distributed systems (Lesser, Corkill) 1982 Synchronization of plans for agents (Rosenschein) 1983 Open Systems (Hewitt, de Jong) 1984 Modeling of Beliefs, Desires and Intentions (Levesque)

  • Actions in plans for multi-agent systems

(Georgeff) 1985 Pursuit Game as testbed (Benda, Jadannathan, Dodhiawala) 1986 Cooperation without Communication, Utilizing game theory (Rosenschein)

  • Classification dimensions of DAI approaches

(Sridharan, ed.)

slide-18
SLIDE 18

Multi-Agent Systems

Jörg Denzinger

1989 First MAAMAW, Cambridge

  • Reinforcement Learning (Watkins)

1990 First MACC

  • Agent-oriented programming (Shoham)
  • Learning in multi-agent systems (Sian)
  • User agents, personal assistant (Hammainen et al.)

(Pan, Tenenbaum) 1991 Knowledge Interchange Format (KIF) (Interlingua Working Group)

  • Teamwork concept (Denzinger)

1993 Founding Workshop of German DAI group, Saarbrücken 1994 Softbots (Etzioni, Weld)

slide-19
SLIDE 19

Multi-Agent Systems

Jörg Denzinger

1995 First ICMAS, San Francisco 1996 First RoboCup, Kyoto Founding of FIPA-Initiative 1997 First Autonomous Agents conference 1998 Agents’ World, Paris Start of the AgentLink program of EU 1999 First IAT, Hong Kong Internet of Things mentioned by Kevin Ashton 2001 First RoboCupRescue competition 2002 First AAMAS in Italy First Trading Agent Competition (TAC) First Workshop on Teaching MAS at AAMAS

slide-20
SLIDE 20

Multi-Agent Systems

Jörg Denzinger

2006 First ART Competition (Using the Agent Reputation and Trust Testbed) Most influential paper award at AAMAS First COMMA (Computational Models of Argument) Conference in Liverpool 2007 First SASO (Self-Adaptive and Self-Organizing Systems) Conference at MIT 2016 IAT rolled into Web Intelligence conference ICA started as alternative In general, trend towards specialized conferences and sub-areas “breaking away” from the field Other disciplines making use of MAS: security, Internet

  • f Things, etc.