We recommend the University of Chicago's Social Science Computing's RePast platform (in Java) for agent-based computational laboratory modeling.

Our general purpose GeoGraph modeling library extends RePast's capabilities by providing both abstract and geographic families of 2D and 3D natural and network landscapes, along with special base classes of GeoGraph agents to move and interact upon such landscapes.

The original 2D version of the GeoGraph library was written by Catherine Dibble to extend the Swarm agent-based modeling platform, but we have retired that version of GeoGraphs and we no longer work with Swarm.

Leigh Tesfatsion also maintains a similar but more exhaustive page than this one of links to software tools for agent-based computational economics. It's also well worth exploring the links on Leigh's home page above to other ACE resources, publications, and researchers.

PGAPack is David Levine's Parallel Genetic Algorithm Package from Argonne National Labs. We have written GAs from scratch in the past, but PGAPack is powerful, flexible, portable, and parallel. We love it, and hope that it will continue to be supported. Alternatively, you may want to try GA packages from The Genetic Algorithms Archive and to check out the GA Group.

LEDA is a C++ Library of Efficient Data Types and Algorithms.   It can be difficult to install, but LEDA is a powerful library that is worth the trouble.   We have used LEDA for graph analysis and for the Cray-based versions of Smallworld simulations.

Finally, geographic landscapes usually have a naturally fixed position for each node. But if you want to visualize the structure of, say, a social network, Tom Sawyer Software provides sophisticated tools for visualizing network structures.