Evolved look-up tables for simulated DNA controlled robots

Abstract

We describe our efforts to convert (short) DNA sequences obtained from the NCBI library into control sequences for simulated robots by simultaneously evolving both a look up table to assign codons to robot commands and a look up table to assign codons to numerical values that serve as arguments to those commands. Our simulated robot is loosely modeled after the Khepera robot. When the robot's sensing capabilities are disabled, we are provided with a sophisticated turtle graphics platform. We formulate a fitness function for evaluating the drawings obtained from codon look up tables and we make inter-DNA and intra-DNA comparisons using our evolved tables. Our results suggest that information content can only be weakly extracted from DNA in this way. © 2008 Springer Berlin Heidelberg.