IEEE Aerospace and Electronic Systems Magazine - June 2020 - 50

Fast Fully Adaptive Signalling for Target Matching

Algorithm 1. FFAST algorithm outline
1: while Mission Active do
2: Input Rn matrix of measurements from receiver
"PAC1-Environment Sensing PAC
3: in
Estimate Interference(Rn )
4: sINT
Design Interrogation Waveform(in )
5: while Interference Estimate OK do
6:
Input Rn target sINT "PAC2-Target Interrogation PAC
7:
tn
Estimate Target Response(Rn )
8:
sTMI
Design TMI Waveform(tn ; in )
9:
while Target Estimate OK do
10:
Input Rn target sTMI "Exploitation Cycle
11:
Calculate Metrics(Rn )
12:
end while
13: end while
14: end while

The ESPAC exists to sense, and adapt to, the background environment. The perception element of the PAC
consists of collecting signals from the radar receiver over
a period of a CPI and creating an estimated occupancy
mask of the frequencies at which the PU is currently transmitting. The data collection is carried out with the local
radar transmission inactive to avoid RF contamination.
The action component of this PAC is the design of an initial target interrogation waveform sINT containing notches
at the locations of estimated PU occupancy, to avoid causing interference to the PU, for the use in the second PAC.
Ideally the environmental sensing process is repeated at
intervals matched to the dynamic behavior of the PU such
that the required response to PU occupancy changes is
achieved. In this work constant update intervals are used.
The second algorithm phase, called TIPAC, instigates
the transmission of a series of pulses of the sINT waveform, collecting radar returns from the target, plus PU
transmissions, and noise over a CPI. The perception
gained from this process consists of a description of the
target estimated frequency response, outside of current PU
occupied frequencies. The action is to design a waveform
sTMI , which both avoids the PU occupancy and matches
the signal to the target characteristics by combining the
previously estimated PU occupancy with the estimated
target response to form a composite frequency mask. The
TIPAC should repeat at intervals in some way matched to
the target response dynamic behavior. In practice, we use
constant update intervals.
The final, non-PAC, loop is the EC, which exploits the
waveform designed from the previous phases, over multiple CPIs, to provide improved SINR for the target, compared with a nonadaptive system, while still avoiding the
PU occupancy. No further adaptation currently takes place
in this phase.
As noted, the nature of the dynamics of the PU frequency occupancy and target response defines the required
50

repetition interval for the PACs. In the current framework,
constant repeat intervals are chosen, although adaptive
control based on performance metrics could be included.
The algorithms employed in the estimation of the RF
environment and the waveform design process are described
in the following sections.

PU ENVIRONMENT ESTIMATION
Under the assumption of Gaussian distributed signals, estimation of the spectral occupancy of the PU is achieved by
making the use of the minimum description length (MDL)
algorithm [57] to evaluate the number of interference signals present in the sampled data.
The environmental estimate enables the creation of a
spectral or Fourier transform mask (FTM), which provides
the waveform design constraints in the Fourier domain.
From [57], the MDL is evaluated as
0Q
1 1ðpÀkÞN
p
pÀk
l
i¼kþ1 i A
MDLðkÞ ¼ À log@ P
p
i¼1 li

(11)

1
þ kð2p À kÞlogðNÞ
2
where N is the number of observations, p is the number of
samples in each observation, k is the number of signals
being modeled, and l represents ordered estimates of the
eigenvalues of the sample covariance matrix, formed from
the sampled interference and noise data.
The MDLðkÞ value given by (11) is evaluated for each
value of k. The estimate for the number of signals contained in the sampled waveform is given by the value of k
for which the MDLðkÞ is minimized. The k largest values
from the average estimated power spectral density (PSD)
calculated over N observations are taken as the frequency
components containing the interference signals. The
allowable spectral occupation for the radar waveform can
then be formulated as the FTM based on these components, q ¼ ½q1 ; q2 . . .qk , where qi 2 ½0; 1, 0 and 1 representing estimated PU-occupied frequency bins and noiseonly frequency bins, respectively.
An example interference PSD and the associated estimated binary signal mask, created by the application of
(11), are shown in Figure 1. The plot represents an
instantaneous snapshot of the dynamic environment. In
this example, the two interference peaks are defined with
equal bandwidths. The second peak and, therefore, the
associated masked region is seen to be wider than the
first peak, this is due to the leakage of energy into adjacent frequency bins in the FFT processing. The use of
longer, zero filled, FFTs removes the issue. Alternative
rank estimation techniques could be applied in place of
the MDL [58], [59].

IEEE A&E SYSTEMS MAGAZINE

JUNE 2020

IEEE Aerospace and Electronic Systems Magazine - June 2020

Table of Contents for the Digital Edition of IEEE Aerospace and Electronic Systems Magazine - June 2020