Projects

SCL v0.05.11 [Updated to Swarm 1.3] - Barry McMullin is studying artificial chemistries and the origins of life. He has two papers available at SFI and many more accessible via his home page. The first one, Computational Autopoiesis: The Original Algorithm ( postscript -- 548207 bytes), is about the particular artificial chemistry that inspired him to develop the SCL model. The second one, SCL: An Artificial Chemistry in Swarm (postscript -- 205293 bytes), is about the SCL model. (The file scl-data00.tar.gz contains the data for bmcm-ecal97.)

Gecko - Gecko is an individual-based simulator for modeling ecosystem dynamics being developed at the Center for Computational Ecology at the Yale Institute for Biospheric Studies. It is being developed by Ginger Booth.

BacSim - BacSim is an individual-based simulator of bacterial growth developed by Jan Kreft at Cardiff University, Wales, UK. It is based on Gecko which is being developed by Ginger Booth. We have screen shots of BacSim and movies of Escherichia coli colonies growing at low (0.1 g glucose per liter), medium (1 g glucose per liter), and high (10 g glucose per liter) substrate concentrations. The embryo of a web page is in its early stages of development.

J.J. Merelo is interested in artificial life, genetic algorithms, neural nets, and other techniques, especially applied to optimization. See the neurolib and libga libs he's contributed to Swarm.

Bugverse is project to look at evolution of neural complexity in neural networks with a recursively modular architectures (i.e. different "scales" of organisation). These networks are embedded within creatures that inhabit a simple artificial ecology. The team is currently Alex Lancaster now at the Santa Fe Institute and David Alexander of the Mental Health Research Institute in Melbourne, Australia.

BeeSim - David Sumpter at the University of Manchester Institute of Science & Technology is studying the behaviour of honey bees and is progressing on the simulation of such using Swarm.

Metabolizing Agents (MAG) - Peter Zvirinsky has developed a project with Swarm called Metabolizing Agents (MAG). MAG simulates a bacteria-like population in an environment with different resource types. MAGs search for the food, must move to the food, burn energy to make movements and to survive, grow as they eat and split in two as they reach reproduction size. Each agent has its genetic code determining food he consume and food he produce and some of its other abilities. The code is also subject of mutation to enable agents adaptation by evolution. Resources flux through the system is also controlled. Monitored parameters are overall population, agents qualities, resources amount and 2D world view itself.

Evo (pronounced EE-vo) is a software framework that allows researchers to study complex systems of independent agents interacting with one another and with their environment. Evo employs biological operators such as genetic recombination and mutation to evolve the behavior of agents so that they are more adapted to their environment.

Brian Enquist at the Santa Fe Institute and Jay Pringle of the UNM Computer Science Department are working on a Swarm model of plant ecosystems incorporating allometric scaling relations.

Gerard Weisbuch of Ecole Normale Supérieure is using Swarm to study adaptive agents in interaction with their physical or biological environments, exploiting renewable resources such as fisheries or polluting their environment and in interaction with other agents in markets

Vladen Babovic and Thomas Gudmunsson from The Danish Hydraulic Institute are studying Individual Based Modelling of Aquatic Ecosystems and Adaptive Numerical Grids.

Melissa Savage (UCLA), and Manor Askenazi have built a model of forest dynamics whose goal is to examine the role of fire on species diversity. The source code for Arborgames, as the app is called, is provided in our anarchy directory. The next generation version of the application is called Arborscapes.

Matt Hare at Macaulay Land Use Research Institute (MLURI) is building a model of the socio-economic & ecological domain of Red Grouse population dynamics called Weaver using Swarm. Matt is also interested in the larger issues of the dependence of a model on implementation and how to avoid such or ensure the integrity of a model when translated.

Claudia Pahl-Wostl's group at the Swiss Federal Institute of Environmental Science and Technology, Duebendorf (EAWAG) is studying ecological and socio-economic networks. A current Swarm model is being used to generate ecological networks, which will allow the investigation of the effects of a network's structural organization and of the properties of the individual network elements on system performance.

A team team affiliated with the Department of Mathematics, Humboldt State University, under the direction of Dr. Roland Lamberson is working on individual-based fish modeling, including development of (a) individual-based modeling theory and (b) modeling software. They are developing animal movement methods that cause realistic complex behaviors to emerge from simple rules as habitat conditions and food supply change. Their California Individual-based Fish Simulation System is a Swarm-based platform for rapidly building, testing, and running experiments with individual-based fish population models. The prototype was developed by Glen Ropella; the code is now being developed by Steve Railsback and Steve Jackson. Their presentations from SwarmFest '99 should be online from the 4th of April onwards.

Village - Village is a model being developed to test varying anthropological theories about Anasazi village formation in the Mesa Verde district. It is being developed by Tim Kohler of Washington State University and Carla Van West of Statistical Research, Inc. Eric Carr has done much of the implementation of the model in Swarm. The Village4 source code is up and running with Swarm 1.0.0.

John B. Corliss<corliss@ceu.hu> and László Gulyás-Zana are part of a group at the Central European University that is starting a Systems Laboratory which will study agent-based complex systems models of social systems as collections of individuals.

Carl Lipo and Sarah Sterling at the University of Washington and Emergent Media is working on the Cultural Transmission Project, which is using simulation to model social interaction using an evolutionary approach to social archeology/anthropology.

Drone - The CAR Group at the University of Michigan Program for the Study of Complex Systems has developed a tool called "Drone" that can be used with Swarm or with other simulation packages to do multiple runs of a simulation while varying the inputs automatically. The CAR Group consists of Michael Cohen, Robert Axelrod, and Rick Riolo.

General Electric's Imperishable Networking group used Swarm to model the evolution of complexity on an active network.

Fabrice Chantemargue is a member of the Autonomy Modelisation and Coordination (AMoC) project at the University of Fribourg that has implemented a model of Implicit cooperation and Antagonism in Multi-Agent Systems in Swarm. He has an enhanced application that uses the Vision library to help model this antagonism. (This archive doesn't create a new directory when you untar it.)

MIMD Systems - Jim Clark at McGill University is porting a simulation of an MIMD parallel computer via "processor agents" moving around in a data space to do load balancing on those processors. He has provided a postscript paper entitled "A Model for Natural and Artificial MIMD Systems" describing the model.

Naga Krothapalli at University of Massachusetts at Amherst is interested in the simulation of multi-agent manufacturing systems, ant colonies and supply chain networks.

Benedikt Stefansson is interested in Computational Economics and is specifically using Swarm to study the dynamics of competition in a differentiated product market and an evolutionary model of Principal-Agent organizations.

Sponsored by the Santa Fe Institute, Brandon Weber has written extensive documentation for the two Artificial Stock Market implementations: Swarm and straight Objective C. The documentation is available either as PostScript, converted HTML or in the original Microsoft Word format.

Paul Johnson <pauljohn@ukans.edu> is now maintaining ASM at SourceForge.

The Swarm Web Interface for Experimental Economics (SWIEE) is a project that aims to develop software classes and examples for facilitating experiments with human subjects.

Darren Schreiber has implemented an agent-based model of the formation of political parties. The program unifies four traditional results in political science as emergent consequences of the model: 1) the tendency towards two parties, 2) movement towards the median voter, 3) party realignment with a change in issue salience, and 4) the tendency towards a minimum winning coalition. He is also working with the Empirical Research Group at UCLA Law School on an update of Thomas Schelling's classic model of racial segregation.

Paul Box is an assistant professor of geography and earth resources at Utah State University and assistant director of a very well-equipped remote sensing and GIS lab that is currently part of the geography department at USU. He previously wrote a simulation in Swarm to study recreational boat traffic in Sarasota Bay.

Catherine Dibble is using Swarm model spatial technologies and human settlement patterns and the associated economic, social, and environmental implications of likely changes. She has two relevant papers on the web: Theory in a Complex World: Agent-Based Simulations of Geographic Systems and Geographic Modeling with Computational Laboratories.

Nick Gotts, Alistair Law and Gary Polhill of the Land Use Science group at the Macaulay Land Use Research Institute in Aberdeen, Scotland, are working on an agent-based model of land-use change in Swarm called FEARLUS (Framework for Evaluation and Assessment of Regional Land Use Scenarios).

Fuzzy-Genetic Decision Optimization for Positioning of Military Combat Units by CPT Rob Kewley, U.S. Army. Fuzzy-genetic decision optimization solved a problem of positioning military combat units for optimum performance. It used a Swarm simulation model to evaluate solutions, a fuzzy logic module to map simulation outputs to a single fitness value, and a genetic algorithm to search the terrain for a near-optimal combination of unit positions. In this study, the fuzzy-genetic system outperformed a human expert during a simulated battle.