About breve

All models are wrong, some are useful. (George Box)

A simulation is a model, and every model leaves things out. breve is built around that: it is a place to build a model quickly, see what it does, change what you think is wrong, and compare the versions, rather than a place to find the one right answer.

Principles

What is in it

Free agents (flocks, ants, herds), cellular grids and patch worlds, networks, and articulated bodies with physics; neural networks and language models as controllers; sweeps and searches over parameters; and an assistant that proposes checked changes in plain language. It runs in the browser, and a headless mode runs the same scenes from the command line.

Where it comes from

breve began as a C++ toolkit for 3D simulations of multi-agent systems and artificial life, with behaviors written in Python or a small scripting language called steve, plus physics, collision detection and an OpenGL display. It was written to support research on artificial life and evolutionary computation, and went quiet around 2009. It was described in breve: a 3D simulation environment for the simulation of decentralized systems and artificial life (Artificial Life VIII, MIT Press, 2002) and, later, in the chapter 3D Multi-Agent Simulations in the breve Simulation Environment (Klein & Spector, Artificial Life Models in Software, 2nd ed., Springer, 2009). This is a rebuild for the web: it runs in the page, with no install.

The author

Hi, I’m Jon Klein, a software architect at LinkSquares in Boston. Before that I did AI research on evolutionary computation, multi-agent simulation and artificial life (publications), and breve is the simulation toolkit I wrote for that work. Twenty papers and chapters came out of that work, several of them with Lee Spector, including Emergence of Collective Behavior in Evolving Populations of Flying Agents, which won the Best Paper Award at GECCO-2003. The full list, and what I make now, is at kleininen.com.

Source

The code is on GitHub.