"Push": a Language for Evolutionary Computation Integrated With breve

This chapter describes genetic programming, and describes how to use genetic programming in breve simulations using an evolvable language called "Push".

Many artificial life simulations are designed around the concept of evolving agents: agents in a simulated world use evolution in order to find better and better solutions to some task such as locomotion, survival or some sort of "intelligent" behavior.

Evolution of agent behaviors is typically accomplished in one of two ways: optimizing a set of parameters, a technique known as "genetic algorithms" (GA); or by developing evolved computer programs, a technique known as "genetic programming".

Genetic programming involves random crossover and mutation of computer programs. Most computer languages designed for people are picky about syntax and variable types and are not well suited to evolution—most random crossovers and mutations would produce code that simply does not compile (like x = y + =;), or code that does compile, but does not make sense semantically (like x = sqrt( "puppydog" );).

The "Push" language is designed specifically for genetic programming and other evolutionary computational applications. Push is designed to avoid most of the complications that can arise when evolving code. In fact, Push code is almost never written by hand. This chapter is intended as a short introduction to using Push in breve simulations and is not a definitive overview of Push or its features. More information about Push and how it is being used can be found on this page.