Evolution of Hive Intelligence Using Genetic
Programming
(1999)WWW pages.Аннотация
Executive Summary: Our project deals with the
evolution of hive intelligence using genetic
programming with the classic video game Pacman as our
model environment. Pacman is an arcade game where a
group of "ghosts" try to catch a Pacman as he
attempts to eat all the dots in a maze in order to
progress to the next level. Hive intelligence is the
concept that a group of individual organisms working
together as a cohesive unit can efficiently accomplish
a defined task. In our model of Pacman, the ghosts are
the individual organisms that are assigned the task of
catching Pacman in a maze as quickly as possible. They
work together as a team, communicating with each other
to catch the Pacmen. At the end of each simulation our
program rates them on a fitness scale to determine
their prowess as a team. The ghost team that catches
the most Pacmen in a specified amount of time gets the
highest fitness score. We take the fittest teams and
mix their programs (genes) together using a crossover
algorithm. We then run another series of simulations
and our program tests the fitness of the new generation
of ghost teams. Our results show that genetic
programming is a powerful means of evolving a routine
to be more effective then any human created algorithm.
The applications of such a process are staggering. In
almost any situation in which computer programs are
used to perform a single, definable task in varying
situations, genetic programming can be used to increase
the efficiency of the program. From simulating the
function of organs in the human body to the exploration
of planets, genetic programming is a useful tool in
creating the best routines for the job.