|
The Integrated Sensing and Processing
(ISP) Program is creating the next paradigm
for application of mathematics to the
design and operation of DoD sensor/exploitation
systems and networks of such systems,
as required to meet computational and
performance challenges posed by next-generation
surveillance, reconnaissance, and strike
scenarios. The program is developing
mathematical tools that enable the design
and global optimization of systems that
interactively combine traditionally independent
functions of sensing, signal processing,
communication, and exploitation. Successful
ISP methodologies will result in replacement
of traditional sensor designs with sensor
system architectures comprising fully
interdependent networks of functional
elements each of which may span the roles
and functions of multiple distinct subsystems
in current generation sensor systems.
ISP considers reductions to key degrees
of freedom in sensing system design and
operation without regard to traditional
subsystem boundaries and interconnect
structures. This is accomplished by applying
recent systematic approaches from physics-based
computational modeling and fast data-adaptive
representations to find and exploit structure
present in the data across all stages
of the sensor system. In many cases,
this procedure permits an automated effective
dimensionality reduction to a tractable
optimization problem that is far more
respectful of the end-to-end structure
of the problem than the conventional
approach.
Past programs within the mathematics
thrust area have successfully applied
a methodology of discovery of physics-based
structure within a sensing problem, from
which it was often possible to determine
and algorithmically exploit efficient
low-dimensional representations of those
problems even though they are originally
posed in high-dimensional settings. Computational
complexity and statistical performance
of fielded algorithms within the DSP
component of sensor systems have both
been substantially improved through this
approach. The aim of ISP is much more
ambitious: to develop and amplify this
concept across all components of an entire
sensor system and then across networks
of sensor systems. Performers in the
ISP Program have developed and demonstrated
schema in which the spatial resolution
of sensors is dynamically adjusted according
to scene content, reducing the number
of pixels that are sampled by a factor
of 10. The same methodology has
been applied to hyperspectral processing,
for which a factor of 100 reduction in
sampled volume of the image cube has
been achieved. ISP researchers
have also demonstrated several schema
for power conserving detection and tracking
in mote fields as well as achieving new
schema for non-myopic scheduling. Recent
scientific activity in the new field
of compressed sensing is a direct result
of ISP funded research. Success
in ISP is expected to yield entirely
new ways of building and operating sensor
systems and networks of sensor systems.
|
