Aerospace and Electronic Systems Magazine October 2017 - 32

Knowledge-Aided Processing for MER

Figure 9.

Clutter dispersion due to multipath. (a) Overhead view of a building and
reflection zone corresponding to a single wall. (b) Spreading of clutter
away from the classic clutter ridge due to single-bounce clutter from a
single wall.

resulting localization is driven by the intersection of the range estimates, which is a localization process known as trilateration.

DATA ARTIFACTS AND MITIGATION
In this section, the artifacts that were observed in the MER Phase
I data set are illustrated and mitigating approaches are described.
The term "artifact" is used to denote any of a collection of issues
with the MER Phase I data that required postcollection mitigation
before the desired multipath features described above could be
reliably detected. As illustrated in Figure 10, the four major artifacts observed in the data were modulation caused by vibration and
automatic gain control (AGC) (Figure 10a), range shift spreading
32

(Figure 10b), range-Doppler spurs (Figure 10c), and intermodulation products (Figure 10d). Although a subset of these artifacts was
likely unique to the radar used, there is value in documenting the
artifacts to aid future users of the data set. In addition, certain artifacts, such as nonlinear effects in the presence of an extreme CNR
associated with urban clutter, are expected to be a concern in any
future MER measurement campaign. It is recognized that no data
set is perfect and radar engineers frequently must peel the onion to
detect and overcome a series of nonideal behaviors. It is shown in
the conclusion of this article that excellent tracking results were
obtained despite the artifacts described here.
In Figure 10a, range-Doppler maps are shown before and after
postcollection mitigation of modulation introduced by vibration
and AGC. Recall that a helicopter was employed to host the radar
in order to provide a slow platform velocity that would in turn minimize clutter spread in Doppler. (The null-to-null clutter spread in
Doppler can be approximated with 2vtθn/λ, where vt is the platform
velocity, θn is the null-to-null beamwidth, and λ is the wavelength.)
A challenge associated with operation of radar on helicopters is the
significant vibration associated with its engines, gears, and rotors.
This vibration is manifested as high-frequency modulation to the
received phase. Additional modulation introduced to the received
data is due to the AGC. In the radar used for the data collection, the
AGC was permitted to change attenuator settings during data collection intervals, thereby introducing an undesired low-frequency
amplitude modulation. Unfortunately, the exact sizes and times of
the attenuator changes were not recorded. In order to highlight the
impact of the undesired phase and amplitude modulation, STAP
was not applied when generating the two plots in Figure 10a. (The
application of STAP alone is not a solution to this issue, because
conducting STAP in the presence of the undesired modulation
would generate a wider null than would otherwise be necessary.)
The red line in the middle of the two range-Doppler maps thus corresponds to the clutter return. In the absence of a mitigating strategy, the clutter is dramatically spread in Doppler and the resulting
sidelobes obscure vehicle low-velocity LOS and multipath returns.
The range-Doppler map after the mitigation of vibration and AGC
modulation exhibits the desired, narrow clutter line.
In Figure 10b, range-Doppler maps are shown before and after postcollection mitigation of range shifts. Again, STAP is not
applied in order to illustrate the artifact. Recall that the radar selected for MER collection incorporated stretch processing. This
architecture is frequently utilized when analog-to-digital converters with reasonable performance and cost cannot be obtained for
the bandwidth of interest. More relevant to MER collection, this
architecture was consistent with the desire for a high duty cycle
and a relatively small range window of 800 m. The high duty cycle
enabled a sufficient SNR with a radar less expensive than otherwise required. The range window associated with this architecture
is smaller than would be used by a fielded system but sufficient for
testing purposes. However, because the airborne platform executed
approximately linear trajectories of 2-min durations, the center of
the range window could not be maintained precisely at a single
point in space. To compensate for the sliding of the range window,
the location of the range window center was estimated (internal
to the radar), and aperiodically adjusted. This was manifested as

IEEE A&E SYSTEMS MAGAZINE

OCTOBER 2017

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