IEEE - Aerospace and Electronic Systems - August 2023 - 58

An Overview of Radar Operation in the Presence of Diminishing Spectrum
frequency, power levels, modulation, beam pointing, etc.
In the reinforcement learning approach, the radar learns to
make decisions leading to measurable performance gain,
such as minimizing the number of frequency collisions
between the primary radar user and secondary emitters.
Figure 7 provides a simple, practical view of cognitive
radar. A wideband, RF front end implements both radar
and electronic support (ES) functions. The ES stage
broadly monitors spectrum, whereas the radar operates
within a specific slice of available frequency at a given
time. Measurements include processed inputs from the
radar and ES stage used to track the electronic states of
various emitters competing for spectrum access, whereas
priors include information from databases and other sources
of knowledge (e.g., UTZ, DEM, weather data, computational
modeling, etc.) that provide additional input into
the decision-making process. The reasoner tracks the nonkinematic,
state features of emitters within the radar
aperture's field of view. Generally, the belief states are a
posteriori probabilities of an emitter being in any given
state. Other methods can also be used to reason over, and
characterize, the operating environment. With a current
assessment of the environment, the cognitive radar then
makes choices to optimize its performance; these choices
can be greedy (made in that instance only) or set over a
multistep horizon. Depending on the access logic of emitters
competing for spectrum, the multistep decision policy
holds the potential for greatest performance gain. A
resource manager finalizes parameter selection and
updates configurable radar settings to close the perception/action
cycle.
SUMMARY
This article serves as an overview of radar operation in the
presence ofdiminishing spectrum. A multitude ofcommercial,
civil, and military demands on spectrum requires the
radar community to carefully consider radar design and
implementation methods. A recent U.S. decision to share
100 MHz ofS-band spectrum between 5G and radar is witness
to evolving methods of frequency allocation [13].
While the perceived need to manage radar spectrum resources
is over a decade old [14], advancements in technology
are just now making new approaches to access spectrum
feasible in radar upgrades and new system development.
Radar system design selects spectrum for very specific
reasons, as the section " Frequency Selection and Use in
Radar " highlights. Radar's selection of frequency is neither
arbitrary, nor a simple matter of convenience. Factors
include resolution, propagation, target behavior, system
deployment, and technology availability.
Previously designed radar systems emphasized performance
and cost, operating in a less technologically varying
environment. The emergence ofcommoditized RF hardware,
58
advances in embedded computing, proliferated design tools,
and a commercial mindset fixed on updating or releasing
products on short timelines is creating a new dynamic for the
radar community: development of configurable radar systems,
use ofopen architecture precepts to rapidly insert capability,
and emphasis on algorithmic control of radar data
collection and processing. Configurability implies access to
expanded RF spectrum DoFs, such as waveform parameters,
beam pointing, propagation characteristics, and polarization,
which play a central role in improving radar detection performance
in interference-limited environments. The section
" Degrees of Freedom for Radar Spectrum Utilization " summarizes
different radar DoFs available to enhance radar's use
ofdiminishing spectrum.
Flexible, wideband, and configurable radar systems
provide a foundation for new spectrum engagement strategies.
The section " Spectrum Engagement Strategies for
Radar " highlights a number of plausible approaches available
to manage spectrum use. In the open sharing model,
all spectrum users are peers and must determine approaches
to coexist using a mutually agreed upon framework. This
approach deemphasizes the radar's flexibility, creating a
number of challenges around time-critical performance
and security. From a typical radar mission perspective,
DSA is sensible for use with mobile radar platforms where,
for instance, a shipboard radar may require certain frequencies
when in port, but upon its deployment the corresponding
spectrum is made available to other users. It is possible
to apply DSA at short time scales, but the use cases require
careful consideration and should include factors like radar
propagation times, coherency requirements, and a need to
access extended, contiguous bandwidth for certain radar
modes. Spectrum underlay and overlay provide hierarchical
user access to mitigate some of the significant constraints
of open sharing and DSA, and thus are more
suitable for radar implementation.
Cognitive radar is an emerging area of research with a
number ofapplication pursuits bringing together the aforementioned,
spectrum access approaches. In the context of
frequency selection, cognitive radar provides a decisionsupport
framework to implement the perception-action
cycle and enable the radar to manipulate its DoFs to optimize
its use of spectrum. If we consider the radar systems
and other emitters to be agents transitioning among nonkinematic
operating states based on agent perception and
logic, with radar actions given by system settings (e.g.,
selecting waveform modulation, beam pointing, etc.), then
effectively the radar is an electronic robot, an idea suggested
in [39] nearly two decades ago! This is consistent
with [34], [36], [37], [38], research that leverages a wealth
of POMDP development matured for robotic system
application.
Radar is at the inception ofa new era in its relationship
with spectrum. Much work remains to address the challenges
ahead.
IEEE A&E SYSTEMS MAGAZINE
AUGUST 2023
IEEE - Aerospace and Electronic Systems - August 2023

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - August 2023
