Researchers at the Laboratory of Intelligent Systems in the Ecole Polytechnique Federale of Lausanne are applying evolutionary principles to robot development. The robots' operating systems started with a set of basic instructions and some random variations that changed every generation in virtual mutations. After each trial, the code for the most successful robots got passed on to the next generation, while code for the less successful robots was bred out.
The researchers designed hunter robots that pursue prey-bots, maze-running bots, and robots designed to deposit a token in a given area. The predator robots were initially programmed with better eyesight, while the prey-bots had more speed. Over 125 generations, the hunter robots learned to approach the prey from blind spots and to hide against walls, while the prey-bots learned to stay away from walls and retreat with its sensors facing the hunter robots. The maze-running bots learned to navigate the maze without any mistakes in less than 100 generations. The token-depositing robots received points for placing tokens in a marked area, with more points resulting in more offspring.
The robots evolved to help each other, especially those robots from the same code lineage.
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA