Misc3025

A model for developing behavioral patterns on multi-robot organisms using concepts of natural evolution

Lukas König

Veröffentlichung: 2007 Oktober
Art der Veröffentlichung: Diploma thesis
BibTeX

Kurzfassung
In this thesis an evolutionary model for robot swarms is developed, interpreting the behavioral program of a single robot as an individual, which has to survive in a population of robot behaviors. The model is applicable to most kinds of robots (including robots, that are capable of building connected multi-robot organisms). The model is experimentally evaluated on real Jasmine-III robots. A byte sequence encoding for robot programs is proposed, which can be manipulated by evolutionary operators. Additionally, it can be exchanged between different robot and simulation platforms if they have a simple interpreter implemented. Therefore, a program has to be developed only once to work on all platforms.

Especially, an interpreter is implemented on the robot platforms Jasmine-III and Jasmine-V, and on the simulation platform BREVE. Evolutionary operators are implemented on the robot platforms Jasmine-III and Jasmine-V. Furthermore, a program for visually constructing genetic sequences and simulating them in an artificial environment is implemented in Java.

Verknüpfte Tools

Organic Computing Learning Robots Arena

Forschungsgruppe

Effiziente Algorithmen

Forschungsgebiet