Past projects
Virtual Life course 2004

Endogenous evolution

Walter de Back
Robbie Veldkamp

DOWNLOAD expdefs:

1. Endogenous Co-evolution (zip)

2. Foodchain (zip)

All animal behaviours are solutions to the problems of survival and reproduction. In order to investigate the evolutionary origins by artificial life simulation, the model must include these selective forces. We incorporate these by using endogenous selection scheme in which creatures are born, survive by consuming food, are able to reproduce with conspecifics by colliding, and die when energy is low, or killed. Contrary to traditional artificial evolution, this model contains no predefined fitness functions and selection mechanisms.

This methodology is used in most projects below.

Sympatric Speciation

Anske Luijtelaar
Joachim de Greeff
David Smits
SJoerd Kranendonk

Project proposal (pdf)

Poster presentation (pdf)

Many examples of speciation are found in nature. In some cases this is caused by a physical barrier which forces a population to diverge into two different species (allopatric speciation). However, speciation without physical isolation can also occur in nature (sympatric speciation).
In this research project we will show that speciation within Framsticks can result from one gene pool and one population without a geological barrier. Starting with a population of intermediate individuals varying in artificial trait X (atX), that is not geographically isolated and on whom disruptive selection is working, evolutionary branching into two different biospecies will occur, without the use of explicit fitness rules.

Sexual Selection

Leo Pape
Stefan Kok
Bouke van der Spoel

Poster presentation (pdf)

DOWNLOAD expdef :

- Sexual Selection (zip)

Sexual selection is the theory that competition for mates between individuals of the same sex drives the evolution of certain traits. One of this traits, is the preference of females to mate with older males.

In this research we will model a certain species in an artificial life simulator, Framsticks, and find out whether evolution can lead to such behaviour.

Semiotics

Joost Wegman
Jonas Matser
Rients van Wijngaarden

Poster presentation (pdf)

Natural ecosystems are interlocked by many sign-processes between species and between individuals of the same species. A sign comes to existence when a signal gains value: when it triggers behaviour in another individual. Sign processes are evolved through a process of co-evolution.

In this projects, we investigate the establishment of signs, through the co-evolution of a small predator-prey ecology.

Advanced predator-prey

Marcus Kruizinga
Remco Bonse

Poster presentation (pdf)

A theoretical biological anaylsis of population dependencies in a small predator-prey ecology.

Evolution of flocking

Frederic Zolnet
Sjoerd de Man
Tom Kemper

Reynolds' BOIDS clearly demonstrated that the complex animal behaviour known as flocking is the result of very simple local rules.
What is less clear, however, is how these rules may be implemented in the sensorimotor mechanisms of these animals. Moreover, it is not at all understood what evolutionary pressures (=environmental conditions) lead many species to adopt such strategies.

This project aims to get understanding of the evolutionary history that lead to the emergence of flocking by using a spontaneous evolution scheme.

Endogenous Neuro-evolution of Augmented Neural Net Topologies

Robbie Veldkamp
Walter de Back

Co-evolutionary 'arms' races' are wiedely held responsible for complexification of behaviour. However, complex behaviour should be facilitated by complexification of neural network topologies.

In this project, we incorporate a variety of the NEAT methdology in the Framsticks simulator. Other than NEAT, we show that complexification is not only possible in fitness optimisation evolution, but also emerges from endogenous evolution.

Framsticks Deathmatch

project finished (jan 2005)

Walter de Back

The Deathmatch is an educational tool for learning about artificial life and learning to use Framsticks by competition.

Student groups create a competitor by using all the methods available in Framsticks. The deathmatch is an experiment in which 5 teams of creatures compete for 'world dominance'.

Diploid genetics

project planned for summer 2004

Lars Zwanepol-Klinkmeijer

In normal genetic algorithms (GAs) all information is encoded on one string (haploid). In nature, genetic encoding is pairwise. There are two copies of every gene in the genome. When these pairs are not identical (due to mutation), the decision which gene to express in the phenotype is done by associated genes that describe the dominance / recessivity.

Diploid GAs in (simulated) autonomous robotics promise greater diversity, better peak performance and better and faster adaptivity to chaning environments.