si 1 swarm intelligence 1

Upload: eka-panuju-quinones

Post on 14-Apr-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 SI 1 Swarm Intelligence 1

    1/8

    1

    Swarm Intelligence

    COMP 5002

    Lecture 1: Introduction

    Overview

    Introductions

    Course

    Logistics

    Process

    Deliverables

    Project

    Lectures

    Assignment

    Outline

    Introductions

    Tony White, Associate Professor

    Office: Herzberg 5354

    Tel: 520-2600 x2208

    Fax: 520-4334 E-mail: [email protected]

    Web: http://www.scs.carleton.ca/~arpwhite

    Course: http://www.scs.carleton.ca/courses/5002

  • 7/29/2019 SI 1 Swarm Intelligence 1

    2/8

  • 7/29/2019 SI 1 Swarm Intelligence 1

    3/8

    3

    Project Deliverables

    Outl ine One paragraph description of project.

    Essentially the abstract for the project paper.

    Due: End of February 2008.

    Project Report Journal-style paper, double column, ~8000 words, format is ACM.

    Final Report due:7th April 2008 (last day of term)

    Implementation Demonstration of software: before7th April 2008.

    Software delivery, including source code, required at time ofdemonstration.

    Plagiarism

    Plagiarism n

    1. A piece of writing that has been copied fromsomeone else and is presented as being yourown work

    2. The act of plagiarizing; taking someone'swords or ideas as if they were your own

    Source: WordNet 1.6, 1997 PrincetonUniversity

    Results of Plagiarism

    If suspected, an oral examination will occur.

    For a first offence:

    If confirmed, student will be given zero marks

    for the piece of work and the incident will bereported to the Director.

    On a second offence:

    If confirmed, the student will be given an Fgrade for the course and asked to withdraw.The Director will be informed.

  • 7/29/2019 SI 1 Swarm Intelligence 1

    4/8

    4

    Materials

    Books: Swarm Intelligence, Bonabeau, Dorigo and Theraulaz, Oxford Press, ISBN 0-19-

    513158-4 (hard), 0-19-513159-2 (paper)

    Swarm Intelligence, Kennedy, Eberhart, Morgan Kaufmann Publishers, ISBN 1-55860-595-9

    Self-Organization in Biological Systems, Camazine, Deneubourg, Franks, Sneyd,Theraulaz, Bonabeau, Princeton Univ. Press, ISBN 0-691-012113

    The Origins of Order, Kauffman, Oxford Press, ISBN 0-19-507951-5

    Emergence, Johnson, Simon and Schuster, ISBN 0-684-86875-X

    Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence,Gerhard Weiss, MIT Press, ISBN 0-262-23203-0

    Web http://www.scs.carleton.ca/~arpwhite/stigmergy-report.pdf

    http://iridia.ulb.ac.be/~mdorigo/ (Ant Colony Optimization)

    http://www.particleswarm.net/papers.html (Particle Swarm Optn)

    http://dsp.jpl.nasa.gov/members/payman/swarm/ (Swarm biby)

    Useful Search Queries

    Swarm intelligence

    Collective intelligence

    Collective Robotics

    Subsumption

    Reactive agent

    Artificial Immune Systems

    Potential Optimization Projects

    Improvements to swarm-based optimizationalgorithms (SBOA):

    Hybrids of Genetic Algorithms (GA), GeneticProgramming (GP) and Ant Colony Optimization

    (ACO) Integrating domain-specific heuristics

    Application of SBOA to practical problems:

    Scheduling, telecommunications, security,

    Contrasting SBOA with other techniques:

    TSP, QAP,

  • 7/29/2019 SI 1 Swarm Intelligence 1

    5/8

    5

    Potential Problem-solving Projects

    Application of swarm-based algorithms to:

    Mobile agents deciding where to go and why!

    Network routing; e.g. multi-priority and QoSintegration

    Real supply chain management

    Automatic programming (variation on GP)

    Novel problems involving clustering: Document classification

    Communications network design e.g. ring

    Alarm correlation and fault diagnosis

    Intrusion detection

    Simulation Projects

    Implementing (learning) agents for:

    Soccer (look for RoboCup)

    Economic systems (look for Kephart)

    Social simulations

    (http://www.biz.uiowa.edu/class/6K299_menczer/social.html)

    Game playing; e.g. tic-tac-toe, go

    Layered problem solving; e.g. subsumption

    Social networking problems/systems

    Theoretical Work

    Analysis of simple swarm algorithms for:

    Complexity

    Asymptotic performance bounds

    Contrast Ant Colony optimization with:

    Reinforcement Learning (RL)

    Neural Networks (NN)

  • 7/29/2019 SI 1 Swarm Intelligence 1

    6/8

    6

    Implementation

    Create or extend a mobile code framework

    that facilitates the generation of swarm

    systems.

    Extend Repast: http://repast.sourceforge.net/

    Students selecting this will lecture on the repast

    framework.

    Repast

    University of Chicago's Social Science Research

    Computing's Repast is a software framework for

    creating agent based simulations using the Java

    Provides library of classes for creating, running,

    displaying and collecting data from an agent-

    based simulation.

    Repast can take snapshots of running simulations,

    and create quicktime movies of simulations.

    Review

    Document state of the art in:

    Swarm Engineering

    Particle Swarm Optimization

    Swarm-based robotics; e.g. Swarm bots

    IMPORTANT:

    Reviews arent description, theyre analytical

  • 7/29/2019 SI 1 Swarm Intelligence 1

    7/8

    7

    Focus and Goal

    Course will have a software agent focus:

    How can simple, reactive agents solve complex

    problems?

    Background in multi-agent systems will be provided.

    Background in GA, GP and RL will be provided.

    Course has as a goal:

    To provide students with an ability to understand and

    exploit biological metaphors with a view to applying

    them to problems in their own domain.

    Course Outline

    The course will cover the topics selected from the following list:

    Introduction to agent systems, and multiagent systems. Describe the variouscommunication mechanisms employed and architectures exploited.

    Introduction to Swarm Intelligence, collective computation, and collectiveaction.

    Natural examples of swarm intelligence: social insects - ants, bees, wasps,termites; emergent control of collective movement - bird flocks, grazing herds,fish schools.

    Ant based algorithms for combinatorial optimization problems, andtelecommunications routing.

    Division of labour, task allocation, task switching, and task sequencing.

    Clustering, brood sorting, data analysis, and graph partitioning.

    Course Outline

    The course will cover the topics from the following list:

    Nest building, and self-assembling.

    Cooperative transport by insects and robots.

    Learning mechanisms for software agents: GA, GP, RL and NN.

    Introduction to the mobile agent, robots, and control methods.

    Projects on mobile agents and simulators applying swarm intelligenceprinciples.

    Software agent architectures for swarm-based problem solving.

    Emergent behaviour in cellular automata.

    Emergent behaviour in social systems

    Reaction diffusion systems

    Self-organized criticality

    Artificial Immune Systems

  • 7/29/2019 SI 1 Swarm Intelligence 1

    8/8

    8

    Overview

    Swarm Intelligence is a new computational andbehavioural metaphor for solving distributedproblems

    Based on the principles underlying the behaviourof natural systems consisting of many agents.

    Technique inspired by the biological examplesprovided by social insects - bees, wasps, ants, andtermites - and by swarming, flocking, herding, andshoaling phenomena in vertebrates.

    Emphasizes distributed solutions to problems,direct or indirect interactions among relativelysimple agents, flexibility, and robustness.

    Overview

    Swarm Intelligence provides a new way to control multipleagent systems - the emergent strategy

    local interactions between simple agents mediated by environment

    self-organize in such a way as to achieve the required task.

    Systems appear to transcend the abilities of the constituentindividual agents

    Emergence of high level control has been found to be mediated bynothing more than a small set of simple low level interactions

    between individuals, and between individuals and theenvironment.

    Overview

    Applications include optimization algorithms,communications networks, and robotics ...

    In this course we study natural systems exhibitingswarm intelligence, and apply the principles to thecontrol of simulated, distributed mobile agentsystems.