IEEE - Aerospace and Electronic Systems - April 2022 - 24

Cross-Platform Radar Resource Management for Coordinated Search and Tracking
M is the number of sensors/platforms.
N is the number of tasks.
1 is a vector with 1 s and of size ðN MÞ 1.
0 is a vector with 0 s and of size ðN MÞ 1.
is the elementwise comparison operator.
P is the reward vector of size ðN MÞ 1.
A is the cost matrix of size MðN MÞ.
Equation (1) means that we want to maximize the gain
Figure 2.
Considered scenario with 15 target aircraft in an air volume.
km width, 20 km depth, and 20 km height. In this volume
there are 15 other aircraft flying and their RCS varies
between 2 and 3 m2.
The question that arises is how can the two AESA
radars be controlled such that good situational awareness
for the pilots is achieved. Among other goals, this can
mean the following.
All aircraft are detected in a timely fashion.
Detected aircraft are tracked with a desired accuracy,
depending on track priority.
Resource consumption is balanced among search
and tracking tasks such that a meaningful tradeoff is
achieved.
Sensing tasks are assigned to the best-suited platform
such that resource consumption is minimized
and resource constraints are not violated.
In cases where more resources than the available are
required: decide which tasks need to be dropped or
their accuracy needs to be degraded.
PROBLEM FORMULATION
This problem can be formulated as a generalized assignment
problem (GAP), see [17, Ch. 11]:
find t ¼ argmax P> t

subject to: t is integer
0 t 1
A t 1
ti þ tiþN 1 ;i ¼ 2; ... ;M
where the following hold.
t is the task to aircraft/sensor assignment vector of
size ðN MÞ 1.
24
(1)
(2)
(3)
(4)
(5)
by executing as many important tasks as possible. The elements
of P correspond to task priorities. Constraints (2)
and (3) mean that a task can be assigned to a sensor/platform
or dropped. Constraint (4) means that each sensor/
platform can handle a maximum load, here scaled to 1.
Constraint (5) is valid only for M> 1 and means that
each task can either be assigned to exactly one sensor/platform
or dropped.
The elements ofP are defined to be the task priorities.
The elements of A are defined to be the duty cycle of the
waveform selected for each task. If there is only one platform
with one sensor, i.e., M ¼ 1, then the construction
of reward vector P and cost vector A is trivial.
This GAP can be solved using approaches described in
[17, Ch. 12].
The abovedefined problem is in practice a macromanager
that defines which tasks each sensor on each platform
should execute, as first presented in [18]. In order to do so,
he needs information about the load imposed by each task
to each sensor. In order to determine the task loads, the
corresponding sensor settings (i.e., waveforms in the case
of a radar) need to be first determined by a separate
micromanager.
In practice, the optimal waveform (and the corresponding
resource needed) for tracking each target with a
desired position and velocity accuracy is provided by a
(radar) micromanager. The micromanager needs to run at
a much higher update rate than the sensor macromanager
solving the GAP defined above and is usually located
within a sensor. Such a micromanager has been modeled
extensively in the literature.
For the purposes of our research, we have used a
micromanager that seeks to minimize the radar duty cycle
needed for each task in order to achieve a desired quality,
i.e., track accuracy for tracking tasks and probability of
detection for search tasks.
TASK AND PRIORITIES DEFINITION
In the problem defined above there are two types of tasks.
Search Tasks: searching parts of the defined FAOR;
Tracking Tasks: updating established tracks;
IEEE A&E SYSTEMS MAGAZINE
APRIL 2022
IEEE - Aerospace and Electronic Systems - April 2022

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - April 2022
