Bijoy K. Ghosh, Ph.D., professor of systems science and mathematics in the School of Engineering and Applied Science, has received a three-year, $706,400 grant from the National Science Foundation (NSF) for a new initiative in Learning and Intelligent Systems (LIS) that seeks to integrate characteristics of biological processes into machines, to program nature into robots.
Ghosh and his Washington University collaborator Alberto Isidori, Ph.D., professor of systems science and mathematics, and partners from Texas Tech University and the University of Chicago will explore biological systems and seek ways of integrating them into machines or robots, making them more flexible in adapting to situations, much as humans and animals are able to use experience and reasoning.
They will model such systems as the visual cortex of the turtle; pattern recognition and visual attention in the primate visual system; and muscle dynamics and head-eye coordination in primates. They then plan to come up with algorithms -- mathematical rules for a prescribed purpose -- that would aid machines in learning from visual clues and predicting the motion of a target in "cluttered" background and would coordinate the motion of the head and eye for "gaze control" and tracking. Gaze control refers to the ability of a system to enter an environment, direct attention and perceive how it must function in that environment.
Humans and animals have the innate capability for learning and adjusting to unstructured environments because of their abilities to observe and move on their own, Ghosh explained. Machines, however, do not.
"The goal is to derive algorithms that would visually estimate the motion parameters in a dynamically changing scene using biologically inspired models of the retina and information-coding processes," said Ghosh. "Engineers would learn from biological systems how robots of the future could integrate visual knowledge, build an internal representation of the knowledge based on neural coding in animals, and be guided by information feedback toward a better machine-human interaction."
Currently, robots are designed and manufactured based on a specific task objective with little emphasis on a feedback control system, which would enable the robot to adjust to changing scenes and environments.
"Our team proposes to introduce and investigate a new feedback paradigm for improved perception, learning, action-planning and control," said Ghosh. "This is an important research problem with a tremendous potential for education."
The LIS umbrella covers a broad range of studies that could lead to rapid and radical advances in how humans perceive the environment through their sensory organs and how they learn to reason and create. Ghosh's grant is one of 28 LIS contracts worth more than $22.8 million issued this fall by NSF.
Interdisciplinary research teams nationwide will undertake projects to help develop a deeper understanding of how learning occurs in humans and animals and in artificial systems -- a robot, for example.
Researchers will try to understand and formulate solutions to such questions as: What kinds of knowledge or skills can actually be learned? How do humans learn? How do other living beings learn? Can artificial systems learn? What kinds of knowledge do they produce?
-- Tony Fitzpatrick
Please send comments and suggestions to: