Agentbased Modeling Geospatial simulation course

1
Agent-based ModelingGeospatial simulation course

• Spring 2008 Sini Ooperi

2
Categorizing Automata

• Geographic cellular automata (GCA)
• cellular automata operating in a geographic space
  (geo-referenced media)
• Geographic automata (GA)
• geographic cellular automata (GCA)
• fixed objects and moving agents
• houses, residents
• buildings, vehicles, pedestrians
• habitat patches, animals

Is the subject of MAS (multi-agent systems) l
3
Multi Agent Systems, a definition

• A multiagent system is one that consists of a
  number of agents, which interact with one-another
• In the most general case, agents will be acting
  on behalf of users with different goals and
  motivations
• To successfully interact, they will require the
  ability to cooperate, coordinate, and negotiate
  with each other, much as people do

4
Multi Agent Systems are inter-disciplinary

• The field of Multi Agent Systems is influenced
  and inspired by many other fields
• Economics
• Philosophy
• Game Theory
• Logic
• Ecology
• Social Sciences
• This can be both a strength (infusing
  well-founded methodologies into the field) and a
  weakness (there are many different views as to
  what the field is about)
• This has analogies with artificial intelligence
  itself

5
What is an agent?

• The main point about agents is that they are
  autonomous capable of acting independently,
  exhibiting control over their internal state
• Thus an agent is capable of autonomous action in
  some environment in order to meet its objectives

AGENT
output
input
ENVIRONMENT
6
General settings of agent-based models

• agents are heterogeneous, not clones
• each agent has its own characteristics,
  "personality"
• characteristics determines the outcome of
  decision-making which in many cases is a movement
  choice
• movement is not restricted to neighboring cells,
  also migration to further away is possible
• agents have their individualistic goals
• agents react differently to their environment and
  to other agents
• agents can react both close- and far-located
  agents or other modeled objects
• Moving objects vehicles in pedestrian
  simulations
• Fixed objects buildings in pedestrian
  simulations

7
General properties of agents

• agents are so-called adaptive autonomous objects
  trying to satisfy their set of goals (fixed or
  time-dependent)
• bounded rationality concerning internal
  decision-making (emotional state and stochastic
  "noise" affect the choice making)
• agents react not only to environment but to other
  agents as well, they can for instance be
  cooperative, they are communicative
• agents are adaptive, they can use their
  experience to continually improve their ability
  to deal with shifting goals and motivations

8
An agent's goals can take on diverse forms

• desired local states
• desired end goals
• selective rewards to be maximized
• internal needs (or motivations) that need to be
  kept within desired bounds

9
Balancing reactive and goal-oriented behavior

• We want our agents to be reactive, responding to
  changing conditions in an appropriate (timely)
  fashion
• We want our agents to systematically work towards
  long-term goals
• These two considerations can be at odds with one
  another
• Designing an agent that can balance the two
  remains an open research problem

10
Social ability

• The real world is a multi-agent environment we
  cannot go around attempting to achieve goals
  without taking others into account
• Some goals can only be achieved with the
  cooperation of others
• Social ability in agents is the ability to
  interact with other agents (and possibly humans)
  via some kind of agent-communication language,
  and perhaps cooperate with others

11
Other properties

• mobility the ability of an agent to move around
• rationality agent will act in order to achieve
  its goals, and will not act in such a way as to
  prevent its goals being achieved at least
  insofar as its beliefs permit
• learning/adaption agents improve performance
  over time

12
Abstract architecture for agents

• Let
• R be the set of all such possible finite
  sequences (over E and Ac)
• RAc be the subset of these that end with an
  action
• RE be the subset of these that end with an
  environment state

• Every agent has
• a set of actions from which to choose and
• a set of internal states which are possible

13
Abstract architecture for agents

• Assume the environment may be in any of a finite
  set E of discrete, instantaneous states
• Agents are assumed to have a repertoire of
  possible actions available to them, which
  transform the state of the environment
• A run, r, of an agent in an environment is a
  sequence of interleaved environment states and
  actions

14
Environments - Static vs. dynamic

• A static environment is one that can be assumed
  to remain unchanged except by the performance of
  actions by the agent
• A dynamic environment is one that has other
  processes operating on it, and which hence
  changes in ways beyond the agents control
• Other processes can interfere with the agents
  actions
• The physical world is a highly dynamic environment

15
State transformer functions

• A state transformer function represents behavior
  of the environment
• Note that environments are
• history dependent
• non-deterministic
• If ?(r)?, then there are no possible successor
  states to r. In this case, we say that the system
  has ended its run
• Formally, we say an environment Env is a triple
  Env ?E,e0,?? where E is a set of environment
  states, e0? E is the initial state, and ? is a
  state transformer function

16
Agents and Objects -gt Emergence

• The three ideas central to agent based models are
  
• social agents as objects
• emergence
• complexity
• Agent based models consist of dynamically
  interacting rule based agents. The systems within
  which they interact can therefore create
  complexity like that which is seen in the real
  world. These agents are
• Intelligent and purposeful, but not so
  intelligent as to reach the cognitive closure
  implied by game theory.
• Situated in space and time.
• They reside in networks and in lattice-like
  neighborhoods.
• The location of the agents and their responsive
  and purposeful behavior are encoded in
  algorithmic form in computer programs.
• The modeling process is best described as
  inductive. The modeler makes those assumptions
  thought most relevant to the situation at hand
  and then watches phenomena emerge from the
  agents' interactions.

17
Multi-agent system (MAS)

• Is a community of agents situated in an
  environment
• from the bottom up approach means that the
  interplay of agents with their environment and
  with other agents give emergence to global
  behavior of the system
• basic question of the modeler
• " What low-level rules and what kind of
  heterogeneous, autonomous agents do I need in
  order to synthesize the system's observed
  high-level behavior in its environment?"

18
Goals of studying MAS

• Researcher moves the individual agents around,
  change their behavior, and modify the
  environment, for example,
• to find simplest body of rules that are able to
  generate the global phenomenon
• to extract maximum amount of behavioral
  complexity from the least complicated set of
  rules
• to gain novel insights to collective dynamics

19
Suitable phenomena for MAS applications

• Any system whose top-level behavior is a
  consequence of the aggregate behavior of
  lower-level entities
• ecological systems (schooling fish, flocking
  birds)
• social systems (housing patterns in the cities)
• economic systems (domestic markets, stock
  markets)
• transport systems
• traffic (vehicles)
• move of pedestrians

20
Example combat

• soldiers respond to
• the geometry of the terrain (battlefield) like
  rivers and mountains
• changing conditions like changes in own firepower
  or firepower of the pack
• changes in their own state like getting wounded
  or fatally shot
• to the location of the enemies

21
Schematic of three sample rules

• movement rules for advance and retreat depending
  of the position of the others fighting in the
  same pack
• rules for shooting if an enemy is within a
  certain range (depends on the weapon)

22
EINSTein land war combat

• http//www.cna.org/isaac/
• http//www.cna.org/isaac/einstein_avi.htm

23
Example Modeling pedestrians in SimWalk

• Purpose of Pedestrian simulations
• design, safety and egress audit of public spaces
  and buildings (train stations, airports,
  hospitals, public places etc.)
• control and improvement of pedestrian flows in
  urban planning and architecture
• walkability studies
• implementation of human movement in traffic
  scenarios
• every pedestrian is simulated as an autonomous
  agent who follows a certain direction according
  to his goal - e.g. a emergency exit - and is
  constrained by other agents or the architecture
  of the building.

24
Pedestrian flow with SimWalk

• SimWalk is a flexible pedestrian simulation
  software focused on traffic and urban
  planning applications
• Analysis of pedestrian safety and comfort
• SimWalk is a decision support software for
  traffic engineers and urban planners
• SimWalk provides a range of traffic
  related analysis tools like Levels of
  Service (LOS), person countig or space
  utilization analyis

25
Pedestrian algorithm

• pedestrian movement is influenced by
  object and pedestrian pressures
• pedestrians move destination directed
  (shortest path to destination), avoiding
  congestions and other pedestrians and
  decide depending on the actual situation
  (e.g. avoid congested exits etc.)
• http//www.simwalk.com/downloads/thanks.html

P
26
Means of analysis

• density analysis (bottlenecks etc.)
• person counting (e.g. exits or self defined
  in space counting line or area)
• walking speeds (mean and single
  pedestrians)
• travel times (mean or single pedestrians)
• space analysis space utilization
  pedestrians per square meter)
• pedestrian trails
• levels of service (LOS) Fruin, Polus,
  Tanaboriboon etc. or self defined

Density analysis of a train station
27
Example Traffic flow CA models

• CA models use integer variables to describe the
  dynamical properties of the system. The road is
  divided into sections of a certain length ?x and
  the time is discretized to steps of ?t.
• Each road section can either be occupied by a
  vehicle or empty and the dynamics are given by
  update rules of the form
• (the simulation time t is measured in units of
  ?t and the vehicle positions xa in units of ?x).
• the time scale is typically given by the reaction
  time of a human driver, ?t 1s. With ?t fixed,
  the length of the road sections determines the
  granularity of the model. For example, if spatial
  discretization is ?x 1.5m, this leads to a
  smallest acceleration of 1.5m / s2.
• CA models have the ability to reproduce a wide
  range of traffic phenomena. Due to the simplicity
  of the models, they are numerically very
  efficient and can be used to simulate large road
  networks in real time or even faster.

28
Traffic congestion

• is a condition on any network as use increases
  and is characterized by slower speeds, longer
  trip times, and increased queuing
• physical use of roads by vehicles
• occurs when traffic demand is greater than the
  capacity of a road (or of the intersections along
  the road).
• extreme traffic congestion, where vehicles are
  fully stopped for periods of time, is
  colloquially known as a traffic jam
• http//vwisb7.vkw.tudresden.de/treiber/MicroApple
  t/

29
Some MAS web sites

• http//www.swarm.org/wiki/Main_Page
• http//www.red3d.com/cwr/boids/
• http//transims.tsasa.lanl.gov/
• http//tmip.fhwa.dot.gov/transims/
• http//www.cna.org/isaac/
• http//www.savannah-simulations.com/simwalk/index.
  html