Managing Business Complexity using Multi-Agent Technology: Practical Applications - PowerPoint PPT Presentation

Loading...

PPT – Managing Business Complexity using Multi-Agent Technology: Practical Applications PowerPoint presentation | free to download - id: cb22f-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Managing Business Complexity using Multi-Agent Technology: Practical Applications

Description:

Whenever an Event (new order, failure, delay) occurs they re-negotiate previously agreed deals ... New market conditions in car industry: 3. Complexity. Father ... – PowerPoint PPT presentation

Number of Views:29
Avg rating:3.0/5.0
Slides: 32
Provided by: GeorgeR
Learn more at: http://users.wmin.ac.uk
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Managing Business Complexity using Multi-Agent Technology: Practical Applications


1
Managing Business Complexity using Multi-Agent
Technology Practical Applications
  • George Rzevski
  • Professor Emeritus, Design and Complexity, The
    Open University, UK
  • Founder and Chief Scientist, Magenta Corporation
    Ltd
  • Founder and Chairman, Rzevski Solutions Ltd
  • www.rzevski.net

2
Agenda
  1. Fundamental Concepts
  2. Solving Complex Business Problems
  3. Complexity
  4. Multi-Agent Technology

3
1. Fundamental Concepts
Stephen Hawking, The Newton Professor of Physics
at Cambridge
  •  "I think the next century will be the century of
    complexity"

4
What is Complexity?
  • A situation is complex if
  • It consists of a large number of diverse
    components engaged in unpredictable interaction
    (Variety and Uncertainty)
  • Its global behaviour emerges from the interaction
    of local behaviours of its components (Emergence)
  • It self-organises to accommodate unpredictable
    external or internal events and therefore its
    global behaviour is far from equilibrium or at
    the edge of chaos (Adaptability and Resilience)
  • It co-evolves with its environment
    (Irreversibility)

5
Examples of Complex Systems
  • Molecules of air subjected to a heat input
    autocatalytic chemical processes
    self-reproduction of cells brain
  • Colonies of ants swarms of bees ecology
  • Civilisations human communities epidemics
    terrorist networks language
  • Free market global economy supply chains
    teams adaptive business processes
  • Artificial Complex Systems multi-agent systems
    the Internet

6
Modelling of Complex Situations
  • Complex situations change during the modelling
    process
  • Therefore tools for modelling complex situations
    must be complex (adaptable) - they must be able
    to build a model in a stepwise manner,
    accommodating every change as it occurs with a
    minimum disturbance to the unaffected parts of
    the model

7
Multi-Agent Software
  • An Agent is a small computer object capable of
    composing, sending, receiving and interpreting
    messages. Agents operate in Swarms
  • Agents consult problem domain knowledge
    (Ontology) before acting
  • An Agent is assigned to every Demand and Resource
    with a task to negotiate the best possible deal
    for its client
  • Demand Agents and Resource Agents negotiate deals
    until the best possible match between Demands and
    Resources is achieved
  • Whenever an Event (new order, failure, delay)
    occurs they re-negotiate previously agreed deals
  • The allocation of Resources to Demands is
    represented as a current Scene (a
    network where Resources and Demands are nodes and
    their matchings are links)

8
2. Solving Complex Business Problems
  • Reformulate the problem situation as Allocation
    of Resources to Demands
  • Collect relevant problem domain knowledge
    including policies, rules and constraints
    specific to this particular problem situation and
    construct Ontology
  • Assign an Agent to each resource and each demand
    and let them negotiate the allocation attempting
    to maintain balance between goals such as
    maximising profit, minimising risk, and
    delivering a given level of service
  • Whenever an unpredictable Event occurs let Agents
    re-negotiate previously agreed allocations to
    accommodate the new situation

9

Scheduling of Tankers
  • Crude Oil Transportation
  • 40 Ships but huge size 300,000 tons
  • 10 of World Capacity
  • 500 Cargoes per year
  • Voyage costs 1million per 45 day voyage
  • Revenue depends on Spot Market
  • Example 2.6million for voyage to U.S. Gulf

10

A Typical Voyage

11
Airport Logistics A Model of the Delivery of
Meals
Resource Agents
Demand Agents

Meal 1, 2, n
Meal order
Flight Agent
Loading bay 1, 2, n
Loading bay order
Food Order Agent
Truck 1, 2, n
Truck order
Trolley 1, 2, n
Trolley order
Route 1, 2, n
Route order
Luggage Order Agent
12
Road Transportation Logistics
  • The Problem
  • 4000 transportation instructions
  • 200 trucks of different capacities and
    with/without trailers
  • operating over the whole UK business network
  • primary and secondary deliveries between 600
    locations
  • 3 cross-docks
  • 4 secure trailer swap locations
  • A considerable number of very small orders
  • dynamic routing, cross-docking
  • handling location availability windows and driver
    breaks
  • frequent occurrence of events such as arrival of
    new orders, change of orders, truck failures,
    road closures, non-show of loaders and drivers
    etc.
  • The Solution
  • For 4000 orders with dynamical routing through 3
    cross docks it took a multi-agent scheduler about
    4 hours to build a schedule
  • The schedule shows strong consolidation of small
    orders onto trucks
  • It schedules new orders in real time (a few
    seconds for a new order)

13
Knowledge Discovery
  • An Agent is assigned to each new data record
    agents invite other agents to join a cluster
    agents negotiate clustering conditions
  • Record Agents of similar records form clusters
    an agent is assigned to each new cluster Cluster
    Agents invite Record Agents of suitable records
    to join
  • As a new record arrives (an Event) it may cause
    re-clustering in dynamic situations clustering
    is a perpetual process
  • Clusters may be represented as rules that
    describe a pattern in dynamic situations
    patterns have a transient character
  • Agents are capable of autonomously discovering
    all inherent patterns (without user intervention)

14
Semantic Search
  • An Agent is assigned to each word in a sentence
    Word Agents through a process of negotiation
    construct
  • A Syntactic Descriptor of the text, which is a
    network of words linked by syntactic relations
    representing a grammatically correct sentence.
  •  
  • A Semantic Descriptor of the text, which is a
    network of grammatically and semantically
    compatible words, which represents a computer
    readable interpretation of the meaning of a text.
    If semantic ontology describes all possible
    meanings of words in a domain, a semantic
    descriptor describes the meaning of a particular
    text.
  • To perform a semantic search Agents compare the
    Semantic Descriptor of the question with Semantic
    Descriptors of candidate answers and select the
    correct answer

15
Other Applications
  • Operating systems
  • Network management systems
  • Support for adaptive enterprises
  • Security
  • Decision Support

16
Conflict between Complex Markets and Conventional
Systems
New market conditions in car industry
The frequency of market driven changes is
typically 1-2 hours
Current ERP systems
  • The time required to modify conventional
    production plans to accommodate changes in orders
    is not less than 8-10 hours
  • In global logistics, once a pallet is assigned to
    a pipeline its destination cannot be altered
    until it emerges from the other end of the
    pipeline, which may take several days

17
3. Complexity
  • Father of Complexity Science
  • Nobel prize winner Prigogine
  • Major centre of research
  • Santa Fe Institute

18
A System Classification

Random Systems (total chaos) Complex Systems (far from equilibrium) Steady-state Systems (in stable equilibrium) Algorithms Clocks (pre-programmed)
Total uncertainty no norms Considerable uncertainty flexible norms of behaviour Practically no uncertainty elaborate rules of behaviour Total certainty every step prescribed
Random behaviour Emergent behaviour Planned behaviour Deterministic behaviour
Disorganised Self-organising Organised and controlled Rigidly structured
Random changes Co-evolve with their environment Small deviations possible No changes possible
19
Examples
Random Systems Complex Systems Steady-state Systems Algorithms Clocks
Molecules at uniform temperature Heat transfer Ecology Human brain Language Epidemics Planetary system Atomic particles
Riot Looting Society Global economy Adaptive business processes Management team Command and control management Automated production
Roulette Dice Multi-Agent Systems The Internet Car Aircraft Ship Conventional software clocks


20
Complexity and Evolution
  • There exists compelling evidence that as the
  • evolution of our Universe takes its course,
  • the ecological, social, political, cultural and
  • economic environments within which we live and
  • work increase in Complexity
  • This process is irreversible and manifests itself
    in
  • a higher Diversity of emergent structures and
  • activities and in an increased Uncertainty of
  • outcomes

21
Evolution of Society
Information Society

Industrial Society
Agricultural Society
22



Evolution of Society

key resources
distribution

stages
Agricultural Society
land
village roads
Industrial Society
capital
motorways railways
Information Society
knowledge
digital networks
23
Evolution of English Language

Shakespeare
Chaucer
24
Evolution of Science

Prigogine
Einstein
Newton
25
Complexity is an Opportunity
  • We have to accept that complexity, and therefore
  • uncertainty, is a norm and that attempts to
    simplify complex
  • situations and to eliminate uncertainty, which
    was a useful
  • managerial and technological philosophy in
    industrial society
  • when complexity was manageable and uncertainty
    was
  • small, is now harmful
  • Complexity of markets can be exploited - it
    offers rich
  • opportunities for those who master the mindset,
    skills and
  • tools of adaptation and resilience.

26
4. Multi-Agent Technology
Multi-agent technology provides tools for
building artificial Complex Adaptive Systems
27
Agents Modify the Current Model to Accommodate
Real-life Events

Real World
Current state
The next state
Events
Virtual World
Modified Scene
Current Scene
Ontology Conceptual knowledge
Data
Data Factual knowledge values
28
Recursive Architecture
Ontology scenes
Swarm 1
Engine
Simulator
--------
Interfaces
Swarm n
Engine swarm
Interface swarm
Ontology scenes
Ontology scenes
Engine
Engine

29
Ontology
  • Ontology is conceptual knowledge of a problem
    domain
  • Ontology is structured as a network where classes
    of objects (characterised by attributes, and
    rules of behaviour) are nodes and relations
    between objects are links

30
Scenes
  • A Scene is a current (perpetually changing) model
    of a problem situation
  • A Scene is structured as a network where
    instances of objects (defined as classes in
    ontology) are nodes and relations between them
    are links

31
Engine
  • Engine is a collection of algorithms which
  • Activate and deactivate Agents
  • Allocate roles (Demand or Resource) to Agents
  • Update the current Scene
About PowerShow.com