Agent Based
Modeling
What is ABM?
• Easiest way to describe it is to demo building one
• Agent Based Modeling is a modeling technique
• Made up of autonomous decision making entities
called agents
• A collection of interacting agents make up a
system
• When we run the system we should see emergent
properties. Things that happen because of the
interactions between agents.
Simple agent rules, can result in different sorts of
complex and interesting behavior in the system.
Lets build a simple ABM
1. The Environment
Field
Grass
Mud
Creating a simple
ecosystem
2. Agents
Agent. An agent is an autonomous, dynamic rule-based entity within a defined environment.
Predator Agents
Prey Agents
Creating a simple
ecosystem
3. Create Rules
Environment Rules
• Grass turns to mud if eaten
• Grass grows back after certain amount of time
Predator Rules
• Will chase prey?
• Need to eat prey for energy
• Can reproduce
• With no energy the agent will die
Prey Rules
• Needs grass for energy
• Can Reproduce if near other prey
• Walk around
• With no energy the agent will die
http://modelingcommons.org/browse/one_model/240
Running the Model
Construct Virtual
World
World develops
through agent
interaction
Virtual world events
driven by agent
interactions.
(Emergent Behavior)
Creation Stages
Go Back and edit the
Virtual World
Predator – Prey game
“This model simulates a predator-prey relationship. The
population consists of wolf packs (predators) and
sheep herds (prey), some controlled by students via
HubNet clients and some androids controlled by the
computer. The wolves gain energy from consuming
sheep, and the sheep gain energy from consuming
grass (a primary producer). The model allows students
to examine simple population dynamics like those
modeled through the Lotka-Volterra equations in a
participatory way.”
Models can be participatory
Other Models
• SKIN Model (Simulating Knowledge Dynamics in
Innovation Networks)
o http://cress.soc.surrey.ac.uk/SKIN/
o Agents are firms, universities, consumers, suppliers
o Agents have to produce an initiative product to survive
• Supply chain modeling
o http://www.sciencedirect.com/science/article/pii/S0925527308002016
o http://egon.cheme.cmu.edu/ewocp/docs/GonzaloEWO-GG-11Dec.pdf
• Consumer decision making
o http://ccl.northwestern.edu/netlogo/models/community/customerBehavi
or -based on customer questionnaires
Key Points to ABM
• Captures Emergent Phenomena
As the components of a system interact with each
other, and influence each other through these
interactions, the system as a whole exhibits emergent
behavior (Roetzheim)
• Flexibility
Can easily be adapted to new constraints – new rules,
agents, changes to environment. Agile?
• Lots of ABM software available
Personal Opinion
• ABM has is good at different things depending on
what you are trying to model:
o Explanatory models
o Exploratory models
o Predictive purposes
ABM Platforms
• Netlogo (very easy, designed to be like Logo)
• Repast Symphony ( Java , Eclipse based )
• Repast HPC ( C++ based for cluster/super
computer simulations )
Useful sites
• Netlogo: https://ccl.northwestern.edu/netlogo/
• Journal of Artificial Societies and Social Simulation:
http://jasss.soc.surrey.ac.uk/
• Modeling Commons http://modelingcommons.org/

Agent Based Models

  • 1.
  • 2.
    What is ABM? •Easiest way to describe it is to demo building one • Agent Based Modeling is a modeling technique • Made up of autonomous decision making entities called agents • A collection of interacting agents make up a system • When we run the system we should see emergent properties. Things that happen because of the interactions between agents. Simple agent rules, can result in different sorts of complex and interesting behavior in the system.
  • 3.
    Lets build asimple ABM 1. The Environment Field Grass Mud
  • 4.
    Creating a simple ecosystem 2.Agents Agent. An agent is an autonomous, dynamic rule-based entity within a defined environment. Predator Agents Prey Agents
  • 5.
    Creating a simple ecosystem 3.Create Rules Environment Rules • Grass turns to mud if eaten • Grass grows back after certain amount of time Predator Rules • Will chase prey? • Need to eat prey for energy • Can reproduce • With no energy the agent will die Prey Rules • Needs grass for energy • Can Reproduce if near other prey • Walk around • With no energy the agent will die
  • 6.
  • 7.
    Construct Virtual World World develops throughagent interaction Virtual world events driven by agent interactions. (Emergent Behavior) Creation Stages Go Back and edit the Virtual World
  • 8.
    Predator – Preygame “This model simulates a predator-prey relationship. The population consists of wolf packs (predators) and sheep herds (prey), some controlled by students via HubNet clients and some androids controlled by the computer. The wolves gain energy from consuming sheep, and the sheep gain energy from consuming grass (a primary producer). The model allows students to examine simple population dynamics like those modeled through the Lotka-Volterra equations in a participatory way.” Models can be participatory
  • 9.
    Other Models • SKINModel (Simulating Knowledge Dynamics in Innovation Networks) o http://cress.soc.surrey.ac.uk/SKIN/ o Agents are firms, universities, consumers, suppliers o Agents have to produce an initiative product to survive • Supply chain modeling o http://www.sciencedirect.com/science/article/pii/S0925527308002016 o http://egon.cheme.cmu.edu/ewocp/docs/GonzaloEWO-GG-11Dec.pdf • Consumer decision making o http://ccl.northwestern.edu/netlogo/models/community/customerBehavi or -based on customer questionnaires
  • 10.
    Key Points toABM • Captures Emergent Phenomena As the components of a system interact with each other, and influence each other through these interactions, the system as a whole exhibits emergent behavior (Roetzheim) • Flexibility Can easily be adapted to new constraints – new rules, agents, changes to environment. Agile? • Lots of ABM software available
  • 11.
    Personal Opinion • ABMhas is good at different things depending on what you are trying to model: o Explanatory models o Exploratory models o Predictive purposes
  • 12.
    ABM Platforms • Netlogo(very easy, designed to be like Logo) • Repast Symphony ( Java , Eclipse based ) • Repast HPC ( C++ based for cluster/super computer simulations )
  • 13.
    Useful sites • Netlogo:https://ccl.northwestern.edu/netlogo/ • Journal of Artificial Societies and Social Simulation: http://jasss.soc.surrey.ac.uk/ • Modeling Commons http://modelingcommons.org/