Juyang Weng
Embodied Intelligence
Laboratory
Department of Computer
Science and Engineering
East Lansing, MI 48824, USA
http://www.cse.msu.edu/~weng/
Abstract
It is
desirable for humans to control robots through high-level commands, and it is
tedious for humans to issue detailed commands to direct actions for every
fraction of a second. However, it is
extremely challenging for humans to program a robot to execute such high-level
commands in unknown human environments.
How can we address this fundamental challenge in robotics? Can a robot develop its skills autonomously
as a human infant or a higher animal does?
What is the fundamental difference between the paradigm of mental
development and that of traditional engineering?
Autonomous
mental development by robots is a new field that has drawn increasing attention
in robotics and artificial intelligence.
Recent advances in neuroscience have cast serious doubt on the notion
that the brain structure and its representation are very much pre-determined by
the human genes. The developmental
program in the human gene seems more of a general-purpose nature that many have
thought, and it enables a human to develop its mind from infancy to adulthood
through real-time experience. For
example, recent results have showed that if visual signal is rewired to
auditory cortex early in life, the primate can develop to perform visual tasks
using its rewired auditory cortex. In this
tutorial, we will present computational models that enable robots to develop
its mental skills autonomously through online interactions with its
environment. Such robots are called
developmental robots. This requires that the internal architecture and
representation of the robot be generated automatically and incrementally
through the developmental process. The
goal of this new research field is to enable humans to “raise” developmental
robots “mentally” through online, interactive “robot sitting” and “robot
classes.” We will also describe some
experimental robots built at MSU and elsewhere as early prototypes of such a
new kind of robot along with some video demonstrations. Finally, the potential social and economical
impact of developmental robots will be discussed.
Tutorial
topics:
Length:
3 hours
1.
Muddiness
of tasks
2.
Overview
of approaches --- knowledge-based, learning-based, behavior-based, evolutional
and the new developmental approach
3.
Human
mental development, results from neuroscience and developmental psychology
4.
Review
of animal learning theories
5.
Supervised,
reinforcement and communicative learning
6.
Architectures
for automatic mental development
7.
Sensory
mapping: representation, development and computation
8.
Cognitive
mapping: representation, development and computation
9.
Motor
mapping: representation, development and computation
10.
Value
system
11.
Integration
of mental capabilities: audition, touch, language, reasoning, decision making,
planning, object manipulation and navigation
12.
Thinking
by a developmental robot
13.
Examples
of developmental robot
14.
Theoretical
completeness and performance metrics
15.
Applications
and the future of developmental robots
Prerequisites: general programming experience, basic knowledge about vector and
matrix operations. Students in
neuroscience and psychology can understand most of the material.
Primary
audience: researchers and graduate students working on
robots, machine intelligence, neuroscience and psychology.
Secondary
audience: system developers for autonomous robots
(e.g., entertainment robots and service robots), educators, philosophers,
investors, and government policy makers.
Handout: Some written tutorial material will be provided.
Biographical
sketch of the lecturer:
Juyang
Weng is an associate professor at the Department of Computer Science and
Engineering, Michigan State University, East Lansing, Michigan, USA. His research interests include autonomous
learning robots, computer vision, autonomous navigation, human-machine
multimodal interface using vision, audition, speech, gesture and actions. He is the author of over one hundred
research articles and book chapters. He
is a coauthor (with T. S. Huang and N.
Ahuja) of the book Motion and Structure from Image Sequences
(Springer-Verlag, 1993). He is the
program co-chair of the NSF/DARPA Workshop on Development and Learning (WDL),
held April, 5-7, 2000 at Michigan State University (MSU), East Lansing, MI
(http://www.cse.msu.edu/dl/), and a program co-chair of International
Conference on Development and Learning 2002 (ICDL’02), to be held at
Massachusetts Institute of Technology (MIT), Cambridge, MA 2002. He initiated and supervised the SAIL
(Self-organizing Autonomous Incremental Learner) project in which he and his
coworkers have designed and custom built their SAIL robot for autonomous mental
development. He and his coworkers are
currently working on the next generation developmental humanoid robot:
Dav. More detail is available on-line at
http://www.cse.msu.edu/~weng/.