Agent-Based Models within spatial information science

1
Agent-Based Models within spatial information
science possible applications and methods

• Charlotte Bruun

2
Object-oriented programming

• Et objekt er en software enhed der rummer
  attributter metoder
• Objekter "kommunikerer med hinanden gn. metoder
• Agent-baseret modellering hænger uløseligt sammen
  med objekt-orienteret programmering
• Distributed software crises
• Computing hardware and networks get smaller,
  faster and cheaper, yet distributed software gets
  larger, slower and more expensive to develop!
  (fayad og Schmidt)
• Genbrug af kode OG design (gn. Frameworks)

3
Terminologi

• Class definitionen af et objekt
• Superclass class som et objekt arver adfærd og
  variable fra
• Subclass en class som arver adfærd og variable
  fra superclass
• Instance et objekt er et instance af en class som
  er blevet skabt i hukommelsen
• Instance variable en variable som er tilgængelig
  for alle funktioner i et objekt
• Method en funktion kaldes gennem objektet
• Attributes data variable

4
3 hovedprincipper

• Encapsulation
• Objekter gemmer deres funktioner (methods) og
  data. Begrænset brug af globale variable. Gør det
  lettere at udskifte dele, og teste enkeltdele.
  Begrænser utilsigtede ændringer af variable.
• Inheritance
• Hver subclass arver alle variable og metoder fra
  sin superclass.
• Polimorphism
• Multiple instances af samme class. Kopierne deler
  adfærd, men ikke state eller hukommelse.

5
Frameworks

• Beskriver arkitekturen af et objektorienteret
  system. Typer af objekter og hvordan de
  interagerer
• Fokuserer på design genbrug (modsat class library
  m. componenter.
• Et framework er et skelet som tilpasses
• Abstrakt klasse er en superclass m. virtuelle
  (tomme) metoder. Bruges til udformning af
  subclasses IKKE instances. (huskeseddel!)
• Genbrugsdesignet er et set af abstrakte klasser
  metoder til interaktion af instances (virtuelle).

6
adfærdsbeskrivelser

• Fra dumme til superintelligente agenter afhængig
  af konteksten.
• Typisk agenter der i en eller anden forstand
  lærer.
• Agenter typisk begrænset i tid og rum - også hvad
  angår informationer
• Goals?? Hvad er formålet med adfærden?
• Metoder til adfærdsbeskrivelse
• Genetiske algoritmer
• Neurale netværk
• If then beslutnings regler

7
Genetiske algoritmer

• Randomly generate initial population M(0)
• Compute and save the fittness u(m) for each
  individual m in M(t)
• Fitness function!!!
• Define selection probabilities p(m) for each
  individual so that p(m) is proportional to u(m)
• Generate M(t1) by probabilistically selecting
  individuals from M(t) to produce offspring via
  genetic operators
• Crossover (recombination)
• mutation
• Repeat step 2 until satisfying result is obtained

8
An agent-based architecture for the simulation of
social reality in a cadastre - S. Bittner

• Environment
• Agents, land, system of documentation (cadastral
  system)
• Agent
• Inbox - messages sent to the agent
• Internal state - goals (duty objective) and
  beliefs
• Outbox - messages sent by the agent
• Decision rules
• Update internal state based on inbox
• Decide on actions to perform (duty (tax)
  objective (buy/sell))
• Update beliefs based on outbox
• Send outbox content to inbox of reciver

9
ABLOoM Location behaviour, spatial patterns, and
agent-based modelling - Otter, Veen og Vriend

• Environment
• Land use layer (land, natural area, sea), fixed
• Attraction layer (agglomeration effects),
  non-fixed
• Different for each type of agent
• Agents households and firms
• Households have Preference for employment,
  neighbours service levels and environment.
• Firms industry, manufacturing, service -gt
  requirement for inputs
• Rules
• Agents search the grid for optimal location
  (local or global)

10

• Example of houshold rules (low-income)
• Search for location with higest attraction
• Set this value as target attraction
• Search for employment opportunities
• Choose location with target attraction closest to
  employment
• If chosen location is vacant, move there - else
  nearest vacant
• Update attraction of chosen location
• Example of firm rules (heavy industry - natural
  ressource)
• Search for location nearest to nature
• If more locations - choose randomly (OBS!
  RANDOM)
• If chosen location is vacant, move there - else
  nearest vacant
• Update attraction of chosen location

11
Litteratur

• Bittner, Steffen (2001), An agent-based
  architecture for the simulation of social reality
  in a cadastra, 4th AGILE conference.
• Otter, H.S, A. van der Veen and H.J. de Vriend
  (2001), ABLOoM location behaviour, spatial
  patterns and agent-based modelling, JASS vol.4
  no.4
• Teran, O, J. Alvaraz, M. Ablan and M. Jaimes
  (2007), characterising emergence of landowners in
  a forest reserve, JASSS vol. 10 no.3
• Dibble, C. and P.G.Feldman (2004), The geoGraph
  3D computational laboratory network and terrain
  landscapes for RePast, JASSS vol 7 no.1
• The spatial dimension and social simulations a
  review of three books, JASSS vol. 9 no. 4
• Hodgson, G. and T. Knudsen (forthcoming), The
  emergence of proporty rights enforcement in early
  trade a behavioural model without reputational
  effects, Journal of economic behavior and
  organization.
• Obs JASS Journal of artificial societies and
  social simulation http//jasss.soc.surrey.ac.uk