Abstract of my Ph.D. Thesis Proposal
Breaking Out of the Black Box: A New
Approach to Robot Perception
Surprisingly,
the state of the art in avoiding obstacles using only vision--not
sonar or laser rangefinders--is roughly half an hour between
collisions (at 30 cm/s, in an office environment). After review ing
the design and failure modes of several current systems, I compare
psychology's understanding of perception to current computer/robot
perception. There are fundamental differences--which lead to
fundamental limitations with current computer perception.
The key difference is that robot software is built out of "black
boxes", which have very re stricted interactions with each other. In
contrast, the human perceptual system is much more inte grated. The
claim is that a robot that performs any significant task, and does it
as well as a person, can not be created out of "black boxes." In
fact, it would probably be too interconnected to be de signed by
hand--instead, tools will be needed to create such designs.
To illustrate this idea, I propose to create a visual obstacle
avoidence system on the Uranus mobile robot. The system uses a number
of visual depth cues at each pixel, as well as depth cues from
neighbouring pixels and previous depth estimates. Genetic Programming
is used to combine these into a new depth estimate. The system learns
by predicting both sonar readings and the next image. The design of
the system is described, and design decisions are rationalized.
Martin C. Martin: Email: martin at metahuman.org