Friday, June 25, 2010

Agent-based models: Why Use them?

I have just come across a few movies on YouTube by Miles Parker on "Why Agent-Based modeling?". These short movies offer a good introduction to agent-based models. The first part (below) explains why we might be interested into ABM and what they are.

The second movie (below) follows on from the first, looking at a specific application ABM, that of modeling H1N1 and contrasts ABM with a more traditional equation-based approaches to looking at disease transmission.

More movies by Miles on YouTube can be found here, while his blog (meta beta) is here. For more details about his H1N1 agent-based models click here.

Thursday, June 24, 2010

Avatars in OpenSim

OpenSim (OpenSimulator) "is a 3D Application Server. It can be used to create a virtual environment (or world) which can be accessed through a variety of clients, on multiple protocols." In a sense its similar to Second Life (SL) but one can host environment on ones own web server (Click here to see some of ourt previous work with SL).

One of the problems with using SL is that one has to own/rent the land in which the environment is created and housed. While with OpenSim one only needs a web server which in a sense is highly appealing for creating permanent environments. Another problem we encountered when using SL was the time it took to execute our agent-based models due to the SL servers. So when I was shown by Chris Rouly this latest test on the ScienceSim Beta Grid by Nebadon Izumi showing 1000 Avatars of mostly bots but some human controlled avatars in real time (as shown below), this got me thinking again about agent-based modeling in virtual worlds.

It would be nice to see Avatars (agents) populating real buildings such as the Ajax FC stadium below. Where the agents are programed to react to changing conditions such as evacuation in times of crisis.

The Ajax FC stadium (Source: KZero Worldswide)

One of our initial tests on using SL for the evacuation of buildings can be seen below. But in our example we only use a maximum of 100 agents but using 1000s of agents with real anthropomorphic dimensions in a 3D space and in real time is highly appealing.

Hurricane Landfall Game

With the hurricane season starting I was looking for models which explored the impacts of hurricanes when they make landfall and came across a game from University Corporation for Atmospheric Research (UCAR) called "Disaster Dynamics: Hurricane Landfall: an experiential learning game." To quote from the website:
The game..."teaches players about interactions between natural hazards and human decisions in a Gulf Coast barrier island community. It is a strategy and negotiation computer game intended for use in undergraduate classes, and is designed for four players who are connected to one another via the Internet."

More information about the game can be found here:

Thursday, June 17, 2010

GMU Fiarfax Campus created with CityScape

I have just been exploring the demo version of CityScape from PixelActive and I am quite impressed. CityScape is an urban modeling tool that allows users to build both custom and real-world environments quickly and easily. The focus of the modeler is on allowing developers to concentrate on design rather than the labor involved in creation.

The movie below show my initial attempt of creating a agent-based traffic model around the Fairfax Campus of George Mason University. The model utilizes a real world elevation data downloaded from the U.S. Geological Survey's (USGS) National Map, the road data comes from OpenStreetMap and the building footprints from the Geospatial Data Collection at Mason.

Some readers might find the movie below from an earlier post created in Repast Simphony an interesting comparison. Basically in this model, agents are pedestrians and they choose the shortest route between two buildings.

Wednesday, June 16, 2010

Crime and Slums

This project focuses on the emergence of criminal activity due to the unmet human needs of those living in Rio's favelas. An agent-based model is developed to explore how human needs, environmental factors, and individual attributes impact state-level behaviors. The emergence of organized crime is observed as "common" criminals turn into gang members. The prevention of conflict requires policies that anticipate responses and avoid conflict. By "re-creating" the current environment, we have the ability to potentially predict the onset of violence where it does not yet exist or understand the source of conflict in those areas already in the midst of violence.

Selected outputs from this research: 

Pint, B., Crooks, A. T., and Geller, A. (2010), An Agent-based Model of Organized Crime: Favelas and the Drug Trade. 2nd Brazilian Workshop on Social Simulation, Sao Bernardo do Campo, Brazil. (pdf)

50 Facts about Agent-Based Computing

I just found out about an interesting pamphlet entitled " 50 Facts about Agent-Based Computing" by Michael Luck and thought it was worth sharing. The pdf can be downloaded from here.

Monday, June 14, 2010

Urban Modelling: Algorithms, Calibrations, Predictions

Mike Batty of CASA "Urban Modelling: Algorithms, Calibrations, Predictions" which was originally published in 1976 has been republished by popular demand!

The book covers a plethora of topics, introducing the reader to simulation models and the need for such methods. For example, "simulation methods are used to derive the behaviour of the system when the system is too complex to be modelled using the more direct analytic approach (Batty, 1976)."

The book provides a summary of the first generation of urban models referring to the key authors and models such as Lowry (1964) model and it successor including the Pittsburgh Time-Oriented Metropolitan Model (TOMM), the Projective Land Use Model (PLUM) for the San Francisco area, and a wide variety of Activity Allocation and Stocks-Activities models. The book presents how such models were mainly developed for practical planning situations through metropolitan planning agencies or consultants in North America and in several European cities. How at first, these models where developed with the aim of solving land-use and transportation questions, later being employed to address a wider range of urban problems.

Anyone interested in urban modelling and spatial interaction models is recommended to explore this book.

Lowry, I.S. (1964), A Model of Metropolis, Rand Corporation, Santa Monica, CA.