Aerospace and Electronic Systems Magazine October 2017 - 30

Knowledge-Aided Processing for MER
cess. In this modified approach, the
direct conversion from a detection
(described in measurement space) to
Cartesian space (i.e., bypassing the
measurement-space tracker) occurred
only for certain situations. In all other
cases, the measurement-space tracker
served to collect and preserve the data
until confidence was obtained that
the relevant ambiguities could be resolved.

STAP in Urban Environments

Figure 7.

Illustration of the multipath naming convention: side view of paths corresponding to LOS (green), single
bounce (blue), and double bounce (red).

To eliminate the geometric ambiguities described above, an
approach using building signatures was developed and successfully demonstrated. In this concept, it is recognized that the utility of vehicle multipath returns is more than simply providing the
ability to detect a vehicle, through reflections, when the LOS is
blocked. Buildings also modify reflections in a consistent, repeatable manner, giving additional information about vehicle location. Collectively, these reflections or sequences of reflections are
called building signatures. As the effort was conducted, evidence
supporting the approach was obtained. For example, in some situations, processing of LOS and multipath returns can provide nearly
instantaneous geolocation of vehicles. A signature that supports
this analysis is the triplet, which is a term used to describe the
simultaneous presence of a vehicle's returns corresponding to the
LOS, single-bounce, and double-bounce paths. The single-bounce
energy may result from two similar paths: (1) from radar, to the
building, to the vehicle, and back to the radar and (2) from radar,
to the vehicle, to the building, and back to the radar. The singlebounce path is compared to that of LOS and double bounce in Figure 7. Two examples of triplets from measured data are provided
in Figure 8, and analysis of the corresponding localization performance is provided below.
It would be expected that MER performance would benefit
from exploiting frequent observations of the same urban area
(and hence multipath) from identical (or similar) observation
points. However, during Phase II, only a limited set of test points
with vehicles traversing only a portion of the illuminated urban
area was available. The overall processing architecture was thus
constructed to exploit this information, when available, but to
not be dependent on its existence. This approach was found to be
robust, even in the presence of severe radar artifacts, as is shown
below.
In the initial architecture suggested for MER data processing,
the ray tracer was used to immediately convert measurement-space
parameters for each detection into a set of feasible Cartesian state
vectors, i.e., without the use of a measurement-space tracker. This
approach is worthy of additional investigation, but in the work discussed here, the numerous ambiguities inherent in this approach
led to many misassociations and resulted in poor performance. It
is believed that the presence of the measurement-space tracker in
the revised approach played a significant role in its ultimate suc30

In any application of STAP [9]-[13],
the presence of clutter discretes, signals from moving objects in training sets, and clutter heterogeneity is of concern. This is the case to an even greater degree in an
urban environment, where numerous large flat surfaces generate
many strong clutter discretes, vehicle multipath returns may fall
into range bins used for clutter training, and clutter characteristics may vary over short range intervals. Application of STAP in
such an environment will benefit greatly from KA approaches
[14] that exploit the three-dimensional city model utilized by
MER. For example, the city model could be used to predict likely
locations of clutter discretes and range bins in which vehicle multipath associated with a cell under test will fall into clutter. As
described in the next section, operation of radar in an urban environment also increases the potential for clutter intermodulation
products due to a high clutter-to-noise ratio (CNR) and limited
receiver linearity.
The behavior of clutter in the presence of multipath is also
potentially of interest. Although double-bounce clutter maps (in
the absence of crab) back onto the classic clutter ridge (at a range
different from that of the corresponding LOS clutter return), simple analysis shows that a single-bounce clutter return does not
map to the classic clutter ridge. The spreading (i.e., dispersion) of
the clutter ridge due to this effect is illustrated for the multipath
associated with a single building in Figure 9. At the left of this
overhead view, the portion of the ground that is illuminated by
energy reflecting from the building's right wall is denoted the reflection zone. A similar zone exists for the bottom wall but is not
shown. The range-Doppler region corresponding to the dispersion of single-bounce clutter is indicated by the red and magenta
regions in Figure 9b. The return due to clutter immediately at
the base of the wall, such as Point A, falls on the classic ridge
indicated by the blue line, but for all other points in the reflection
zone, two paths exist for the single-bounce returns. For example,
the energy corresponding to Point B appears at the two locations
in range-Doppler shown in Figure 9b. For the parameters associated with MER Phase I data collection, it is not expected that
this source of dispersion is a major concern, although it could be
significant in other applications of MER. Finally, it is suggested
here that dual polarization be considered for future urban GMTI
radars to aid STAP training, as well as improve localization and
association. Only a single polarization was employed in the MER
Phase I data collection.

IEEE A&E SYSTEMS MAGAZINE

OCTOBER 2017

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine October 2017