INSPIRE Track 1: Is Evolvability Driven By Emergent Modularity? Biomimetic robots, gene inspired information structures, and the evolvability of intelligent agents

Vassar College, Poughkeepsie NY

Investigators

Abstract

This INSPIRE award is partially funded by the Evolutionary Processes program in the Division of Environmental Biology in the Directorate for Biological Sciences, the Behavioral Systems program in the Division of Integrative Organismal Systems in the Directorate for Biological Sciences, and the Information Integration and Informatics program in the Division of Information & Intelligent Systems in the Directorate for Computer & Information Science & Engineering. For millennia, humans have bred organisms to produce better food, clothes, and companionship. Recently, scientists have learned how to breed robots, evolving simulated creatures in virtual worlds, or physical robots in the real world. By combining the evolutionary process with robotic engineering, more complex and novel designs should be possible compared to traditional methods. In spite of the promise, so far evolved robots only do simple things like walk, navigate, or pick up objects. What limits progress is a lack of understanding of "evolvability," the capacity of organisms (or robots) to change and become more complex. Understanding evolvability is the main goal of this project: researchers will borrow ideas from modern genetics so their robots mutate and develop in ways that are similar to how biological creatures do. In theory, this could produce simple robots that evolve into ever more complex, capable and useful robots. Understanding how complexity evolves is central to the study of life, and may enable even non-specialists to automatically and continuously produce diverse kinds of machines. By linking complexity, genetics, and evolution, this project seeks to discover new principles that can be applied in science and industry. To help convert scientific principles into innovation drivers, online software will be created to show how to evolve virtual or physical robots; this will help students learn about engineering, biology, and how to apply both to technology. Finally, evolutionary robotics can be used to solve complex problems in robotic control that defy logical programming solutions, so this research can help companies that manufacture robots.