FABLES: A Functional Agent-Based Language for Simulations

1
FABLES A Functional Agent-Based Language
for Simulations

• László Gulyás
• AITIA International, Inc
• lgulyas_at_aitia.ai
• with Sándor Bartha
• Simulation Center, Cooperative Research Center
• Faculty of Informatics, Lorand Eotvos University
• Supported by the GVOP-3.2.2-2004-07-0005/3.0
  grant of the Hungarian National Office for
  Research and Technology

2
(No Transcript)
3

• This page was intentionally left blank.
• That is, my presentation will have an approach
  that is slightly unusual in this community.

4
Motivation

• Software is not as it used to be.
• Traditional methodologies are aimed at a single,
  monolithic program with well-defined and
  controllable input streams.
• Todays software is almost always situated in a
  dynamic environments.
• Computers are networked, but even on a single
  computer, many programs are running
  simultaneously.
• The software designer/engineer can no longer
  enumerate or control the state(s) of the
  environment.
• More importantly, the expected behavior of the
  software is most often not independent of the
  non-controlled components.
• For example, the success of an autonomous agent
  negotiating a deal on an auction site clearly
  depends on the performance of other similar
  agents, programmed by unknown parties.
• We need methods, techniques and tools for
  engineering emergent complex (software) systems.

5
Outline

• Agent-Based Social Simulation
• FABLES (Functional Agent-Based LanguagE for
  Simulation)
• Goals, Concepts, Examples
• Context MASS (Multi-Agent Simulation Suite)
• ABM and participatory simulation
• Demo prototype

6
Agent-based social simulation (ABSS)

• Scientific investigation of complex social
  systems, where
• many individuals (agents) interact,
• each pursuing its own agenda.
• Seeks to explain global phenomena from (often
  relatively simple) local (individual-level)
  behavior.
• The practice of agent-based social simulation
  requires
• both programming skills and
• social science education.
• Often carried out by collaborating teams of
• social scientists and
• software engineers.

7
Agent-Based Modeling and Simulation I.

• Computational models to create artificial
  societies to study emergent social phenomena.
• Agent-based modeling (ABM) and multi-agent
  simulation (MAS) has gained popularity in many
  disciplines, especially in the social sciences.
• They provide a powerful and unique means to
  experiment with and to understand the processes
  underlying many complex phenomena in systems
  involving many interacting individuals.

8
Agent-Based Modeling and Simulation II.

• Agent-based models are computer programs where
  agents represent the interacting individuals of
  the system.
• Agents here are independent behavioral engines
  that satisfy the so-called weak notion of
  agency.
• However,
• they are often not pro-active.
• their behavior is typically fairly simple and
  synchronized,
• thus they are usually very different from
  AI-agents.
• The goal of ABM/social simulation is to
  understand and to explain (the complex behavior
  of the system given the local rules)
• Therefore, the agents are usually in lack of
  sophisticated knowledge representation and
  communication.

9
Practices of ABSSREPLICATION above everything

• Scientific experiments (tests and replicas)
• True (uncontrolled) parallelism is ruled out.
• Probabilistic models
• Pseudo RNGs
• Control over the seed
• Independent variables, Separate RNGs
• Full specification
• E.g. Standard practice of random choice among
  equal maxima.

10
Practices of ABSSGenerating and Handling of
Results

• Statistical nature of results
• One go is no go.
• Sensitivity Analysis and Confidence Intervals.
• Parameter Sweep
• Non-Linear Dependencies
• Tricks like Active Non-Linear Tests (ANTs)

11
Practices of ABSSSeparating Model and Observer(s)

• Basic idea in science,
• but in computational practice its only been
  around since Swarm (1994)
• Several observers
• GUI
• Batch1
• Batch2
• Independence of the Observers RNGs from the
  Models RNGs.

12
Tools for ABM

• Swarm/MAML (1994-1998)
• RePast (2000)
• Ascape (1999-2001)
• MASON (2003-2004)
• MASS/FABLES (2005)

13
FABLES Design Goals

• Source should be easily readable for readers
  familiar with the basic mathematical formalism.
• Should have a precise semantics and the source
  should be the exact specification of the model.
• The FABLES source should be as close to a
  publishable model description as possible.
• FABLES models should be executable.
• The model description should focus on the nature
  of the model and leave implementation to the
  compiler.
• The language should be general enough to
  accommodate possibly any agent-based model, but
  should focus on the common techniques and methods.

14
FABLES Solutions

• Separation of the observer. (Integrated
  Environment)
• Object-Orientation for structure.
• Functional Specifications to summarize
  relationships and behavior.
• Imperative elements to schedule the actual
  dynamics.

15
FABLES Development environment in Eclipse
16
FABLES example Random walk

• model randomwalk
• agentNum 100
• class Agent begin
• pos // is Integer x Integer
• schedule step cyclic 1
• 2 pos pos discreteUniform( -1..1,
  -1..1 )
• end
• schedule init
• 0 seed(0)
• 0 displayload("user.Display3",-20,20,-20,20)
  
• 1 new Agent pos0,0 i is
  1..agentNum
• //////////////////////////////// OBSERVER
  //////////////////////////////////////

17
FABLES example Mousetraps

• model Mousetrap
• worldSize 10 // Model parameter
• // Shorthand for the space
• world 1..worldSize, 1..worldSize
• mousetrapsFired // Counter
• // The agents
• class Mousetrap begin
• pos // is Integer x Integer
• hasFired // is Boolean
• activate (hasFired false) gt
• mousetrapsFired mousetrapsFired
  1,
• printLn("Activated"),
• hasFired true,
• generateTriggers
• otherwise gt

• // Main schedule
• schedule mainSchedule
• 2 discreteUniform(Mousetrap).activate
• generateTriggers addEvent( mainSchedule, 1,
  triggers )
• triggers a.activate a is
  Mousetrap, b is RndPositions when
• a.pos(1)b(1) and
  a.pos(2)b(2)
• where (
• RndPositions discreteUniform(world) a is
  1..2
• )
• //////////////////////// OBSERVER
  ///////////////////////////
• graph // variable to store the graph object
• schedule observer cyclic 1

18
FABLES example Schellings Segregation

• model Schelling
• // Model parameters
• worldSize 10
• agentNum 70
• threshold 0.6
• // Constants
• red 1
• blue 2
• color red, blue
• // Spatial Constructs
• world 1..worldSize, 1..worldSize // The
  world
• occupied a.pos a is Resident // Occupied
  positions
• empty setMinus(world, occupied) // Empty
  positions
• // Helper function to implement a thorus
• norm(x) alt1 gt aworldSize

• // The agents
• class Resident begin
• pos // is world
• c // is color
• t // is 0.0..1.0
• neighbors a is Resident when a.pos
  in neighborous(pos(1),pos(2))
• sameNeighbors a is neighbors when a.c c
• utility try( size(sameNeighbors)/size(neighbor
  s), 1.0 )
• closestEmpty empty(minPlace(d(pos, loc)
  loc is empty))
• schedule Step cyclic 1
• 2 pos ( utility gt t gt pos
• otherwise gt closestEmpty )
• end
• ////////////////////////// OBSERVER
  ///////////////////////////
• display // Variable for the display object

19
FABLES example Conways Life

• model Life
• worldSize30
• world
• norm(x) a lt 1 gt aworldSize
• otherwise a where ( a x mod worldSize
  )
• neighbours(x,y) size ( 1 dx is -1..1,
• dy is -1..1
• when not
  (dxdy0) and
• world(norm(xdx))(no
  rm(ydy))
• )
• step(n,old) n3 or (old and n2)
• newWorld
• step( neighbours(x,y), world(x)(y)
  ) y is 1..worldSize

• schedule Step cyclic 1
• 3 world newWorld
• ////////////////////////// OBSERVER
  ///////////////////////////
• display
• schedule Observer cyclic 1
• 2 display( (x-1,y-1,1) x is
  1..worldSize, y is 1..worldSize
• when world(x)(y) )
• end

20
Summary

• Overview of Agent-Based Modeling Simulation
• As a means to engineer emergent phenomena in
  complex software systems.
• Easy to use formalism for ABM
• Report on an ongoing project.
• Context Multi-Agent Simulation Suite
• Web-enabled simulation environment for
  participatory simulation.
• Safely configurable simulations for educational
  use.

21
Thank you!

• Comments are welcome at
• lgulyas_at_aitia.ai