If it will be successful, will toggle in its "real virtual life" state.
If it will continue to live, eating enough, each 6 hours it will breed and its offspring will receive the same genotype, with the eyes disposal and the neural network weights plus a percentage "error" above each parameter. The error assures that offspring will be a little bit different than the parent and the Natural Selection will do the rest.
Generation after generation Virtual Triops will co-evolve together all others creatures competing for the same food.
At present, after about 240 generations, evolved Virtual Triops are enough fast and precise to compete with the very old Creatures (B).
This experiment is an evidence (if still needed) of Darwin's intuition about the way organisms evolve.
Darwin's law can be adapted to any system or software reproducing itself with error.
"Virtual Triops" are first creatures, in our virtual lab, on Second Life, changing their morphology and behavior as result of Natural Selection. They have 3 proximity sensors "eyes" and a vertical sensor to maintain the horizontal plane. Each eye gives the distance to the next piece of food to a very simple, unsupervised learning, neural network that combines all data coming in one impulse toward the target (the food).
At the very beginning, when a VIRTUAL TRIOP borns the first time, with no parents, it is in "learning mode" and eyes position is randomly generated. The Virtual Triop has only 10 minutes to discover and catch the first piece of food. If it is not able to catch the food it will die. The Virtual Triop will try to find the right weights for its neural network by tentative using a very simple unsupervised learning algorithm.
The Artificial Life Lab is a virtual Lab. You can find us in Second Life © Envirtech Island. Whether you are a Second Life resident or a new user follow the Artificial Life Lab
link.
Artificial Life Resources
Karl Sims
|
Evolving Virtual Creatures
|
PDF
|
Thomas Miconi
|
The Road to Everywhere: Evolution, Complexity and Progress in Natural and Artificial Systems.
|
PDF
|
Gene D. Ruebsamen
|
Evolving intelligent embodied agents within a physically accurate environment
|
PDF
|
Daniel R. Kunkle, Chadd Merrigan
|
Pulsed Neural Networks and their Application
|
PDF
|
Jon Klein
|
BREVE: a 3D Environment for the Simulation of Decentralized Systems and Artificial Life
|
PDF
|
Nicola Chaumont, Richard Egli, Christoph Adami
|
Evolving Virtual Creatures and Catapults
|
PDF
|
Lee Spector, Jon Klein, Chris Perry, Mark Feinstein
|
Emergence of collective behavior in evolving populations
|
PDF
|
Hugo de Garis
|
Genetic Programming - Building Artificial Nervous Systems with Genetically Programmed Neural Network Modules
|
PDF
|
|