Meeting Abstract
By Barbara Webb’s codification, biorobots test hypotheses about biological systems. While those systems may be particular organisms, they may also be processes. Evolution, for one, has been modeled in embodied robots to test hypotheses about the origin of early vertebrates. But explicit models of development, the mapping of genotype to phenotype, have been wanting. Thus our work extends evolutionary biorobotics to include development, allowing the two processes to be studied as they interact. Specifically, we test the hypothesis that random errors in transcription feed back to the genotype over generational time to increase genetic variance of the population and alter the evolution of morphological complexity. Key to this process is that random errors in development create masquerading genomes, individuals with indeterminate mapping of genotype to fitness. We digitally simulate populations of autonomous mobile robots in which genomes encode morphological and neural structures, spatial relations, and regulatory elements; the interactions of structures and regulatory elements unfold in an explicitly modeled developmental process. We simulated 11 levels of genetic mutation rate and transcription error rate in 10 populations of 60 robots over 100 generations, with fitness determined by a simple locomotion task. In the presence of directional selection, genetic variation was proportional to the rate of transcription error. Moreover, transcription error and mutation acted independently and in different ways on the evolutionary dynamics of the population. This work was funded by the U.S. National Science Foundation (grant no. 1344227, INSPIRE, Special Projects).
