Kirkus Reviews QR Code


What Evolving Robots Can Teach Us About the History of Life and the Future of Technology

by John Long

Pub Date: April 3rd, 2012
ISBN: 978-0-465-02141-3
Publisher: Basic Books

Long (Cognitive Science and Biology/Vassar Coll.) traces his path from a doctoral student studying the evolution of fish vertebrae to his present position as director Vassar's Interdisciplinary Robotics Laboratory.

Although biologists depend on computer modeling to study neural networks, predator/prey relations and virus interactions, the author is frequently asked, “What do robots have to do with biology?” His short answer is that autonomous robots—even the simplest propeller-driven designs with an embedded computer and a sensor—have agency and can move around and interact with their environment. Long explains how a blunder in an early version of his doctoral thesis led to his later work with robots. His hypothesis was that vertebrae strength and flexibility evolved because it enhanced a fish’s ability to compete for food. He developed a computer model to correlate the relationship between the elasticity and flexibility of a marlin backbone to its swimming speed, but was dismayed to realize that he had inadvertently violated the laws of physics. His simplified assumptions had transformed the would-be marlin into a perpetual-motion machine. With a two-dimensional computer model, such an error was possible, but not with a three-dimensional one that actually moved. Long’s first self-propelled robot had a fairly simple design—an embedded minicomputer, one light sensor and a backbone built to mimic varying structural aspects of a marlin vertebrae. Natural selection would be modeled on the ability of a robot to reach a target first in a competition of six robots. His first model failed because his rules deducted points when the robot wobbled, which was accounted as an energy loss. In fact, as Long learned, wobble gave the robot greater flexibility in reaching a target and was a survival advantage. More complex robots allowed him to model predator/prey relationships and target acquisition more realistically, and he was able to consider broader issues such as the relationship between goal-directed behavior and animal intelligence.

Lively and intriguing.