|
The University of Arizona 1230 E. Speedway Blvd., Rms 241/241a Tucson, AZ 85721-0104 Phone: (520) 621-4296 Fax: (520) 621-2478 |
|
Research in the laboratory of
Professor Charles Higgins
is in the areas of computational neuroscience (focusing on dipteran visual
motion processing), biologically-inspired
engineering systems, and Neuromorphic Engineering,
particularly in the areas of VLSI vision and motor control systems
applied to problems in autonomous robotics.
The laboratory is currently supported by the
National Institutes of Health
and by the Air Force Office of Scientific Research.
|
Engineers have a lot to gain from studying biology. The study of
biological neural systems alone provides numerous examples of
computational systems far more complex than any man-made system that
perform real-time sensory and motor tasks in a manner that humbles the
most advanced artificial systems. Despite the evolutionary genesis of
these systems, there are common design strategies employed by biological
systems that span taxa, and engineers would do well to emulate these
strategies. However, continuous-time parallel biologically-inspired
computational architectures do not map well onto conventional
discrete-time serial processors. Rather, an implementation technology
that is capable of directly realizing the layered parallel structure and
nonlinear elements employed by neurobiology is required for power and
space -efficient implementation. Custom neuromorphic VLSI hardware
meets these criteria, and yields small, light, low power dedicated sensory
systems that are ideal for autonomous robot applications.
|
|
Higgins lab research as a whole attempts to address the question of how
engineers can best learn
from the representations and computational architectures demonstrated by
neurobiology. A new paradigm for design is developing which
abstracts away the specific organism to reveal strategies that span taxa.
Projects range from neuromorphic VLSI design to computational emulations of
biological systems at levels from the biophysical to the neural system.
|
|
|
Standard serial processors
cannot be used for efficient implementation of
neurobiologically-inspired architectures and representations.
The field of Neuromorphic VLSI was created by
VLSI pioneer Carver Mead at Caltech in the late 1980's as a way of using our
rapidly growing knowledge of neurobiology as an inspiration for more efficient
engineering designs.
Neuromorphic VLSI chips incorporate architectures and representations
inspired by neurobiology in
a mix of analog and digital circuitry, often using MOSFETs operated in the
subthreshold (weak inversion) regime.
Such systems are
unclocked, highly parallel, small in size and weight, and operate at very low
average power consumption.
Past projects have focused on visual motion, with particular interest in
extraction of egomotion information and sensitivity to complex patterns of
optical flow.
|
|
These VLSI sensory systems are ideal for small, inexpensive
autonomous robots, whether for underwater, flight, or land-based applications.
Because a sensory system can be most efficiently utilized if tightly coupled
to a motor system, laboratory projects currently concentrate on building VLSI
systems which incorporate both sensory and motor control systems. Projects
also involve integration of these systems with robotic platforms of various
types, from Lego robots to custom biomorphic robots.
Such VLSI systems can not only be efficient engineering
solutions for sensory/motor applications, but also serve as real-time hardware
models of complex nonlinear neurobiological systems.
|
|
|
Some pictures from the lab are below.
|
|
|
| Layout of the lab | Testing a vision chip |
This page updated on 7/11/07.