EPSRC Network on Evolvability in Biology & Software Systems

Evolvability, Genetics & Development in Natural and Constructed Systems: Abstracts of the EPSRC Evolvability Network Symposium

Tewin Bury Farm Hotel, Hertfordshire, England, UK
26-28 August 2003


University of Hertfordshire Computer Science Technical Report 389
C. L. Nehaniv, P. J. Bentley & S. Kumar (Editors)

Why Genetic Regulatory Networks are Useful for Evolving Robots

Why Genetic Regulatory Networks are Useful for Evolving Robots

JOSHUA C. BONGARD

Computational Synthesis Lab
Sibley School of Mechanical and Aerospace Engineering
Cornell University, 191 Grumman Hall
Ithaca, NY 14850
U.S.A.

jb382@cornell.edu
http://www.people.cornell.edu/pages/jb382/


Evolutionary robotics is a rapidly maturing sub-field of Artificial Intelligence and Artificial Life. This technique uses computer algorithms, which act as models of biological evolution, to automatically generate behaviours for simulated or real robots. In this talk I will describe an algorithm that models genetic regulatory networks (GRNs), and evolves these GRNs in order to 'grow' both the body and brains of robots in simulation to perform a desired task. I will show that such an ontogenetically-based model has several advantages for the automatic design of robots: first, the length of the genome specifying a robot phenotype does not necessarily lengthen as the phenotype becomes more complex over evolutionary time; second, the model allows evolution to design phenotypic modules, which can be repeated at different locations on or within the robot's body; and third, environmental influence of the growth process can easily be incorporated and exploited by evolution. During the talk I will provide details regarding the algorithm, a brief description of other evolutionary robotics approaches to this problem, observed trends in the evolved GRNs, images of videos of the evolved robots, and future prospects.