Evolving Robots Evolve Robotics:  Robo-inspired Animal Models Test Biological Hypotheses

John H. Long, Jr

Vassar College

While animals provide inspiration for roboticists, robots, in turn, inspire biologists. A growing number of biologists use robots to represent, as model simulations, biological mechanisms, behaviors, and evolutionary processes.   For example, embodied robotic animals have been built and evolved to explore a fundamental question in biology:  why is morphospace (the morphological hyperspace that contains all forms) clumped?  In other words, given all physically-possible forms, why has evolution produced only a subset?  Moreover, why do only a fraction of the evolved forms persist?  The use of robots to address these questions has been influenced by the bio-robotic and evolutionary-robotic approaches of Ijspeert, Floreano, Lipson, Nofli, and Webb, among others.  Useful robotic simulations, when subjected to selection pressures or behavioral experiments, produce experimental data that test the predictions of specific hypotheses.   Populations of robots that evolve also produce a library of animal models that are extinct, rare, or never existed.  We can then explore both the evolved and the unevolved regions of morphospace.  This expansion of our knowledge of morphospace, originally for biological purposes, provides roboticists with new ideas for novel designs.  This feedback between biology and robotics is iterative.

Bio

John Long holds two appointments at Vassar College:  Professor of Biology and Professor of Cognitive Science.  In addition, he is co-founder and former Director of Vassar's Interdisciplinary Robotics Research Laboratory.  He received his Ph.D. from Duke University's Zoology Department, specializing in Biomechanics.  His research combines locomotor biomechanics, biomimetics, bioinspired robotics, and evolutionary biorobotics.