IEEE Aerospace and Electronic Systems Magazine - August 2020 - 21

Cataldo et al.
vector and its parameters. As shown in Figures 4 and 5, the
whole tracking process is performed into three main steps:
Local tracking to merge detections into tracklets;
Multilateration to estimate the RSO's position coordinates from the range measures;
Global tracking through unscented Kalman filter
(UKF) to estimate velocity vector and covariance
matrix.
Each one of these steps is discussed in the following
sections.

LOCAL TRACKING
As previously stated, in the local tracking step each sensor
carries out a track separation algorithm, in which the single observations are merged to form tracklets. It can be
observed from Figure 7 that each H1 -decision:
Could initialize a new track;

Figure 6.

Could be related to an existing track;

High-level block scheme of the OR detector.

Could be a false alarm.
the question is how to make a detector work optimally in
all possible scenarios.
Even though the optimum number of pulses to integrate is not known a priori, it may be estimated. In particular, knowing the antenna pointing direction and the
velocity of an RSO at a particular altitude, it is possible to
estimate how long the RSO remains in a range resolution
cell and, hence, the number of radar pulses to integrate.
Although the estimate may be affected by an error and,
therefore, the estimated number of pulses may not exactly
correspond to the real one or changes over time, the estimated number of pulses will still be a value close to the
true one and, therefore, also the resulting SNR value will
be next to the optimum one.
Hence, the idea is to use a bank of multipulse detectors
and then decide the detector to use based on the estimated
number of target pulses during the observation time.
In Figure 6, a schematic block of the proposed processing chain is shown. The N blocks integrate a different
number of pulses but with a fixed value of PFA . A generic
detector Detector i-pulse represents a coherent or an incoherent detector that integrates i received pulses. The gray
circle in Figure 6 represents a switch that enables a detector based on the estimated values of radar pulses to integrate N^T . This "Multi-N detector"is applied to any range
cell in parallel.

TRACKING
Once the detection step is performed, a tracking step is
needed in order to estimate the RSO's position and velocity
AUGUST 2020

It can be observed that also a H0 -decision can influence tracklets separation. As stated in [34] and [35], global
nearest neighbor is the most widely used method. This
algorithm handles the input observations sequentially and,
for each one of them, identifies the most likely assignment
of observation to existing tracks.
As Figure 7 shows, when a H1 -decision occurs, a gating operation is performed. Then, the existence of tracks
is verified. If there are not existing tracks a new one is initialized, otherwise the measurement is compared with a
prediction, then the most likely observation is assigned to
the track. Once the assignment is performed, a new prediction is made. If there are more than N consecutive assignments, then the new track is confirmed.
If a H0 -decision occurs for a preexisting track, then a
miss detection (MD) is stated. As a consequence, a track
is declared lost if more than K consecutive MDs occur. On
the contrary, if less than K consecutive MDs are collected,
the prediction is directly assigned to the track and a new
prediction is evaluated. In the event that the track is
declared lost, two alternatives are offered:
1) The track has a suitable number of samples ( > U),
thus it is stored for future processing;
2) The track does not have an adequate number of
samples ( U), hence it is discarded.
Regarding the gating operation, a commonly used
method is the rectangular gating, which consists of identifying a feasible interval where observations related to
a track could be found. The interval is expressed by the
following equation:

IEEE A&E SYSTEMS MAGAZINE

IEEE Aerospace and Electronic Systems Magazine - August 2020

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