You are on page 1of 15

Thales WATCHKEEPER WK450 (2010)

The Thales Watchkeeper System is an unmanned aerial vehicle (UAV)


for all-weather, intelligence, surveillance, target acquisition, and
reconnaissance (ISTAR) delivered to the British Army.

48

Emerging Radar
Trends

The bat species Pipistrellus pygmaeus emits


acoustic signals whose ambiguity functions
are varied with time in a cognitive manner,
according to its perception of the target scene.
Future radar systems will mimic this behavior.

48.1 Introduction
To date, much of the emphasis in improving radar performance
has been on enhancing sensitivity and resolution. Sensitivity provides for better detection range, while resolution enables more
detail to be observed, such as for target classification. These
have been achieved, as we have seen throughout this book,
with considerable success. Long-range detection and high-resolution imaging are now routine. However, as a consequence,
discrimination between targets of interest and other objects is
becoming more important than the problem of target detection.
At the same time, advances in digital technology continue to be
made such that almost every radar parameter, particularly the
radar waveform, can be altered on a pulse-by-pulse basis. This
is the basis of the subject of waveform diversity, and is creating
many new and exciting possibilities not only for improving radar
performance but also in opening up new applications.
Here we touch on just a few trends that are emerging from
research to development. Specifically, we explore some of the
technology advances that are already beginning to have an
impact on radar design. We then go on to see how they facilitate what may well become a revolution in radar through the
application of closed loop processing, much of which takes its
cue from natural echolocation systems. Indeed, echolocation is
a technique employed in the natural world to great effect by
mammals such as the bat, whale, and dolphin. Although their
sensing is rooted in acoustics, these mammals can effectively
see with sound and exhibit many of the characteristics that
would be highly desirable in radar systems.

48.2 Technology Trends


Perhaps the most significant of all technology trends that is
impacting and will continue to impact the design of radar
683

684

PART IX: Special Topics and Advanced Concepts

systems is that of digital technology. The trends in digital technology are as follows:
1. Increasing dynamic range and higher speed analogue to
digital convertors
2. Increasing capability for processing at faster and faster rates
3. Increasing memory capacity and speed of accessibility
4. Reducing costs
The effects of improvements in digital technology are already
evident in the design of both transmit and receive radar
sub-systems.
For example, on transmit, waveforms can be programmed with
almost limitless freedoms. The frequency, modulation, amplitude, bandwidth, and PRF can all be chosen with high accuracy
and programmed into a digital waveform generator. Furthermore,
these design parameters can be changed dynamically so that
each pulse emitted can have a quite different specification.
On receive, the location of the analog-to-digital (A/D) converters is becoming progressively closer and closer to the antenna.
It is now almost normal for digitization to take place at intermediate frequency (IF) at rather than at baseband, as was the
case previously. Thus, the much greater flexibility that comes
with digital processing is being used to overcome limitations of
analog circuitry. This is already resulting in both greater versatility and overall improved system performance.
Research systems have begun to appear that are termed software defined, By software defined it is meant that the radar
operating modes can be programmed in software and, at least
in principle, can be varied nearly instantaneously, such as on
a pulse-by-pulse basis). This trend towards digital control and
parameter programmability is likely to lead to the advent of
all-digital radar. Lower operating frequency systems are again
being reported in the literature, and higher frequency systems
seem likely to follow.
However, A/D converter dynamic range is still often insufficient for many radar applications. At wide bandwidths high
A/D converter speeds are required. It is more difficult to
obtain high A/D converter dynamic range at fast A/D converter speeds. High range resolution demands very fast A/D
converter speeds, and dynamic range requirements are less
easy to meet. This necessity for both high dynamic range and
high speed A/D converters is likely to slow the onset of alldigital, high frequency, wideband radar.
Another technology trend, yet one that is nearly as old as the
invention of radar, is electronic scanning. Despite its longevity, electronic scanning is far from universal, most usually
only appearing in complex and expensive military systems.
Electronic scanning has been employed successfully to create
beam patterns with very low sidelobes. Electronic scanning
also enables adaptive beam forming that can simultaneously
view targets with high gain while rejecting sources of interference. Indeed, multiple modes of operation such as search and

CHAPTER 48: Emerging Radar Trends

track can be supported simultaneously. However, the adoption of electronically scanned systems has progressed relatively slowly. The main reasons for this are a combination of
complexity, cost, and an inability to easily operate over wide
instantaneous bandwidths. Some of todays operational systems use techniques such as sub-arraying and even mechanical
scanning to help reduce complexity and cost. Nevertheless,
such obstacles are steadily being eroded, and much of the
technology to enable very wide band all-digital arrays with
digitization at the element level already exists. Design drivers
such as the need to use the electromagnetic spectrum resource
more efficiently may eventually play a part in tipping the cost
balance towards electronically scanned systems.
Electronic scanning coupled with advanced digital technology
is moving radar into a new era. Electronic scanning means that
the radar beam can be pointed anywhere at any time. Also
almost all radar design variables can be changed as the beam
moves from one position to another. It is the combination of
both of these that allows electronic scanning to support multiple different tasks leading to true multifunction radar systems. Figure 48-1 illustrates schematically the vast range of
tasks that a multifunction radar might have to carry out.
However, this poses the question as to where and when to point
the radar beam and how to optimize the radar design parameters to best carry out a given task. This simple question raises
significant and fundamental issues that are influencing the direction of radar research that is likely to continue indefinitely.
Electronically scanned radar systems have led to the emerging topic of resource management that has the objective of

Midcourse
Guidance

High Angle
Search/Track

MULTIPLE TARGET
TRACK

Terminal
Homing

Targets

tion

Target Classifica

Jamming

Missile Launch

Cued Se
a

rch

Volu
m

e Se

arch

Jamming
Surface
Picture

SURVEILLANCE

High Speed Horizon Search

Figure 48-1. A multifunction phased array radar can perform a large number of tasks. The resource manager must prioritize the different
tasks so that the most critical are performed first and the radar resource is used in the most efficient manner.

685

686

PART IX: Special Topics and Advanced Concepts

deploying radar resources to most effectively accomplish a


task or group of tasks. This is largely through the design of
waveforms and the direction of beam pointing. Much effort
has been expended on how to schedule radar transmissions
together with making decisions as to which radar resources to
deploy. This notion of the radar making its own decisions is
embedded in electronic scanning and digital processing. In the
future adaptive feedback based on interrogation of received
echoes will inform the design of future transmissions. We
touch upon just a few aspects of electronically scanned digital
radar systems in the next sections.

48.3 Radar Resource Management


Radar resource management is the efficient allocation of finite
radar resources to accomplish a task or set of tasks in an optimal manner. In fact, electronically scanned radars will be able
to fulfill all of their potential only if they are able to optimally
deploy their resources within the finite amount of time that
a radar system has available. In other words, if the radar has
to send a series of pulses in multiple different directions at
the same time it may not have enough time to send pulses
for a tracking task. Thus a decision as to when to look and
where has to be made, and task priorities have to be set. Most
approaches to radar resource management have divided this
topic into the following three separate categories:
1. Adaptive track updating,

Modes in a Multifunction Radar System


Surveillance
Detection
Volume search
Verification
Handover to tracker
Tracking
Initiation, maintenance, termination
Position estimation
Track file updates
Missile guidance
Mid-course correction
Signature interpretation
Detection in clutter
Classification

2. Adaptive search scanning and


3. Scheduling.
More recently, attempts have been made to use measures of
tracking and search performance as a basis for optimization
and hence as a basis for determining and allocating radar
resources.
Radar Resource Management Tasks. Radar resource management plays a central part in carrying out the mission to which
the radar system contributes. This mission might be set at a
high level by defining tasks such as the detection and tracking
of all targets in a designated area or volume. These tasks may
then be further refined as sub-tasks, such as the requirement
that all tracked targets must fall within a set error bound. As
shown in the box to the left, there are a large number of possible tasks that can be carried out by an electronically scanned
radar system.
A mission can have inherent time dependence and may have
to be redefined as a task unfolds. Targets may enter and leave
the coverage area or volume and so the number of targets that
must be detected and tracked will vary with time. Thus, there
is a large number of radar variables, a large number of tasks to
be carried out, and potentially a large number of targets. This
means that the optimization problem faced by radar resource
management is one that is extremely challenging.

CHAPTER 48: Emerging Radar Trends

The previous blue panel illustrates the radar modes that have
a requirement for a waveform design to be transmitted and a
direction in which the waveform is emitted. Once the resource
manager has made these decisions, it must allocate a time slot
at which the waveform is to be transmitted and received. Each
waveform requires different radar parameters such as modulation, power, frequency, bandwidth, and PRF to be selected.
The selection and allocation of waveform parameters as a function of time can be accomplished in a wide variety of different
ways. One approach is to trade off the radar design parameters
as a function of the task to be carried out. For example, the
following steps might be used:
1. Decide the functions to be carried out (e.g., detection, tracking or classification)
2. Ensure that targets can be detected at an adequate signal to
noise ratio
3. Allocate radar resources to functions
4. Check the resource allocations against the available timeline
5. Trade off the allocations and
6. Make sure that nonviable modes are eradicated
However, this doesnt take into account the fact that different tasks may have different priorities and different coverage
regions may be prioritized differently. In fact, these priorities
will most likely dictate the detailed allocation and timing of
radar resources such that the highest priority, time-critical tasks
are carried out first. There can be occasions where so many
tasks have to be carried out that they cannot all be accommodated. In this case the lowest priority, least time critical tasks
would be set aside. Naturally, any tasks that cannot be accomplished must be flagged to the radar operator.
Radar Resource Management Categories. There are three
broad categories by which radar resources are allocated:
1. Rule based, where a pre-determined behavior is programmed into the radar system as a function of a small
range of scenario types and missions. This might pre-define
separate waveforms for search and track but otherwise not
change their specification. It may also predefine time spent
searching versus time spent tracking.
2. Self-organizing, where the radar itself evaluates the scenario
against requirements set by an operator. The waveform
selection and beam pointing parameters, together with their
allocation to the radar timeline, are determined on the fly by
the radar system. Such an approach may continually review
and respecify the radar waveform, the amount of time spent
searching or tracking, etc.
3. Hybrid, where some system parameters are pre-determined
while others are selected on the fly by the radar. Thus the
hybrid approach is a combination of (1) and (2) that is part
rule-based and part self-organizing.

687

688

PART IX: Special Topics and Advanced Concepts

Other Sensor
Information
Force
Requirements

Environmental
Awareness

Command
Requirements

Resource
Management

How to do it?

Hardware and
Signal
Processing

How well did i do?

What needs to be done?

Do it

Task
Selection
Task
Optimisation

Task Scheduler

Figure 48-2. The scheduler allocates each radar task to be


executed at a particular time, making sure that the highest priority
time critical tasks are assigned first. The system then reassesses the
scenario to examine how well the tasks have been, and are being,
accomplished in preparation for the next resource management
cycle.

Most current electronically scanned radar systems are very prescriptive in the way they allocate their time to different modes.
Future systems will learn to adapt their performance as a function of their own understanding of their sensed environment.
They will use this to modify waveform parameters and beam
pointing to optimize performance in a given task. This is a
challenging and vibrant area of electronically scanned radar
research and one that is set to continue for some time into the
future.
Radar Resource Management Activities. Figure 48-2 shows
the flow of activities and information that are a part of radar
resource management. There are a number of inputs that help
set the tasks to be carried out. The job of the resource manager is to turn these into activities that will be carried out by
the radar system. The figure shows that there are many and
varied components that form part of resource management
and consequently just how complex and unwieldy the whole
process is. In addition to sensing the environment, the radar
may receive information from a variety of other sources such
as geographic information systems (GIS), databases, as well as
other sources. This has to be assimilated into as accurate a picture of the scenario as possible so that designated tasks can be
carried out. The resource manager then works out what tasks
to do and how best to do them.
Once the radar tasks have been identified and specified, they
have to be scheduled into a queue. This is the job of the task
scheduler (Figure 48-2).
Role of Radar Resource Manager. Figure 48-3 shows a system architecture for an electronically scanned radar system
highlighting the role of the resource manager. There is a tight

External Data
Radar System

ESA Radar

Signal & Data


Processing

RRM
Command Queue

Data Fusion

Situation &
Threat
Assessment

Sensor
Manager
Operator

Other Sensors (e.g. IRST)

Task Requests Mission System


Weapon System

Figure 48-3. This shows how the radar resource manager (RRM) is set in the context of the overall electronically scanned radar system architecture.

CHAPTER 48: Emerging Radar Trends

689

coupling between the radar hardware, signal processing, and


the resource manager that reflects the rate at which radar
parameters can be re-assigned. There is a looser coupling with
external data sources and the setting of requirements because
they occur on a slower timescale.
Overall, resource management is an ongoing area of research,
and there remains much to be done before optimal systems
can be said to truly exist. Perhaps the most significant challenge is for the radar to correctly and accurately interpret
echoes so that it has the best possible information for re-setting
the parameters for future emissions. This can only be accomplished by the radar system itself as the radar parameters can
be changed in times of a millisecond (i.e., well inside human
decision-making and action). Accurate and full interpretation
of radar echoes provides the motivation for the adaptive and
cognitive sensing that we shall examine in the remainder of
this chapter.

48.4 Echolocation in Nature


The bat, whale, and dolphin are well known for their ability to
use echolocation as everyday activities key to their very survival as a species. Indeed, they have been successfully using
echolocation for over fifty million years and have honed the
technique to carry out a remarkable range of tasks. Perhaps
less well known is the fact that humans can also echolocate,
and there are some blind people who have become extremely
expert. Bats are able to forage for food in dense clutter environments; some of their targets are stationary and some are
moving. Their acrobatics are nothing short of extraordinary,
often taking place in complete darkness. Dolphins are able
to use echolocation underwater with such acuity that they
have been recruited by the US Navy to search for mines.
Although they are using acoustic rather than electromagnetic
signals, their ability to achieve tasks beyond radar makes them
a suitable subject for study with the aim of inspiring new and
improved forms of concepts and processing. For these reasons
bio-inspired radar research is a very active area and may well
yield new insights that influence the design of future radar
systems. Here we examine some of the characteristics of the
echolocating bat to illustrate the techniques used.

One technique routinely employed that makes great sense


from a radar perspective is to adjust the PRF such that range is
always unambiguous. This is shown in Figure 48-4, in which
the black dots are the position of a bat flying close to a hedge.
The circles surrounding the black dots represent the extent of

3
Height / m

Waveforms. Bats, dolphins, whales, and humans all use some


form of tongue-click to generate their signals. The waveforms
are generally pulsed with pulse lengths around 1 ms. The
pulse repetition frequency is of the order of 100 Hz. The emitted waveforms are used as a tool for both navigation and target
classification (i.e., selection and acquisition of prey).

SOZ (Y = 2m)
4

Hedge

1
X/m

Figure 48-4. This shows how the bat dynamically adjusts the pulse
repetition frequency of its calls so that the signal overlap zone
(SOZ) is no more than the range to the hedge.

690

PART IX: Special Topics and Advanced Concepts

the first range zone that is adjusted to be no more than the


range to an obstruction, in this case a hedge along which the
bat is flying. The bat will also adjust its PRF to the wing beat
of flying insects as a means of identification. This continuous
adjustment of PRF as a function of the prevailing environment
is not something routinely used in radar. Instead, either one of
a small number of PRFs is selected or the unambiguous range
is extended by using multiple PRFs.

Frequency (kHz)

100
20

80
60

40

40
60

20
0

0.005 0.01
Time (sec)

80

Figure 48-5. A spectrogram plot of the acoustic signal produced


by a bat. The spectrogram shows three harmonically related
components each exhibiting a hyperbolic component and a
constant frequency component.

Frequency (kHz)

100
20

80
60

40

40
60

20
0
0

0.005 0.01
Time (sec)

80

Figure 48-6. A spectrogram plot of the acoustic signal produced


by a bat, showing three harmonically related components in the
final phase of attacking prey. Note the absence of the constant
frequency component that is evident in the latter part of the pulse
shown in Figure 48-5.

Bats modulate their waveforms but with a structure not used in


radar. Furthermore, the modulation is varied as a function of
time throughout an activity, such as selection and acquisition
of prey. The modulation itself varies considerably from one bat
species to another, and within a species according to the particular task that the bat is undertaking. Bats will emit anything
from a pure tone of a near constant frequency to a structure
that has a time frequency profile close to a hyperbolic modulation. Even more elaborately, bats often emit, within a single
pulse, a series of harmonically related modulations. These can
be a group of tones or a group of hyperbolic modulations or
a group combination of tones and hyperbolic modulations, as
shown in Figure 48-5. Another feature of these waveforms is
that the power emitted in the different components is not the
same. The reasons for these waveform structures are far from
clear and are the subject of current research. However, since
they are the result of evolutionary optimization over millions
of years, there must be some reason for them, and it therefore
seems likely the waveforms have these structures in order to
aid the bat in its navigation and recognition tasks.
Another feature of these waveforms is that the different harmonic components are sufficiently separated in frequency that
they will be emitted with different beam widths. The higher
frequencies have narrower beam widths (shorter wavelength
divided by a fixed aperture size). Perhaps this helps to distinguish between targets that have different reflective properties
at the different frequencies? Alternatively, it might be a way of
distinguishing between targets at different angles? We can only
speculate, but nevertheless it shows how analysis of biological
systems can help stimulate new thinking, thereby generating
ideas for future radar research.
The waveform parameters used by bats are continually being
changed as they fly their mission. For example, the PRF for
general surveillance is typically around three or four times
lower than that used to intercept prey. Furthermore, the modulation that the bat applies to the waveform is continually being
changed. In the final phase just prior to intercepting and capturing prey, the constant frequency component (Figure 48-5)
is all but gone, leaving behind three much more steeply raked
hyperbolic harmonics as shown in Figure 48-6.
Figure 48-6 also shows that the duration of the waveform has
greatly reduced. One reason for this is that the bat expends
energy when emitting a waveform. Thus, as it closes in on a

CHAPTER 48: Emerging Radar Trends

target, the signal strength required for detection reduces, and


consequently a shorter pulse can be used. A second reason
is that the bat avoids eclipsing losses. As we have seen in
Chapter 45, this is very desirable in airborne radar. Modulation
of pulse width offers an additional variable that could be used
to minimize the effects of eclipsing.
Thus the bat is adjusting its PRF and waveform modulation
as it goes about the selection and acquisition of prey. In addition the bat is also continuously adjusting its trajectory. Rarely
does a bat go straight towards its prey. More typically, it will
circle the target, appearing to probe for information by using a
number of pulses emitted at different orientations. Could it be
gleaning additional information to confirm that this is truly a
source of nourishment? Further research is needed to answer
such questions, but it is clear that the bat is using echolocation
to see the world and can do so with great effectiveness.
Overall, the bat is constantly adjusting many, if not all, of the
waveform parameters under its control, including the orientation between it and the object being interrogated. It would be
reasonable to assume it does this on the basis of information
extracted from predecessor pulses. In other words, it is interrogating its environment and making changes to improve its
ability to carry out its task within that environment. In this way
the bat is an excellent example of an adaptive and cognitive
echolocating system. Research is now beginning to examine
cognition in echolocation mammals that may provide further
important clues as to how radar systems can be improved and
even extended into new applications.

48.5 Fully Adaptive Radar


The technology now exists for all radar parameters to be varied
on a pulse-by-pulse basis. We have seen in the previous section
that this is something done in natural echolocation systems to
great effect. It seems intuitive that there are better and perhaps
optimal parameters that should be collectively selected for a
given radar function. Indeed, we have seen cases throughout
this book where this is done for a particular application. One
example where this occurs is in the setting of the radar PRF.
This might be set to low, medium, or high depending on the
application. However, this is fixing one set of parameters so
that they are as close as optimal for one application. The likelihood of the selected parameters being optimal in all cases, let
alone multiple applications, is very low. However, an adaptive
approach, in which the parameters can be varied as a function
of the information gleaned from predecessor pulses, provides
a basis for continually iterating towards a more optimal parameter set.
Hence, the question immediately arises about the selection
(and reselection) of the best parameters. This also has some
quite fundamental implications. First, this has to be done
by the radar itself, since the PRFs are too high for humans

691

692

PART IX: Special Topics and Advanced Concepts

to intervene on a pulse-by-pulse basis. Secondly, it implies


the radar system itself has to have knowledge of the scene
it is illuminating as well as a clear understanding of the task
or function to be carried. Only armed with this information
can the radar choose the best parameters. Continuous optimization of the radar parameters is the objective for fully
adaptive radar.
Consider an example based upon a single target tracking. The
task of the radar system is to track an already detected target. It does this by continually pointing the peak gain of the
illuminating radar beam onto the target in order to maintain
the most accurate track. A maximum difference between the
location as measured by the radar and the predicted location
can be set. Radar measurement errors are largely determined
by detection performance, which in turn is largely determined
by the signal to noise ratio. Thus, adaptive radar might have
a control loop (Figure 48-7) where the radar parameters are
adjusted to maintain a desired signal-to-noise ratio (i.e., target
detection performance). So if the target echo were to begin to
fade, the radar system would interpret this as a reduction in
signal to noise ratio. It can then react to automatically redress
this, perhaps by increasing transmitter power (or pulse length
or PRF or some combination of parameters). There may also be
constraints that have to be observed. Constraints may be due
to limits on transmitter duty cycle or the need for unambiguous ranging, and these also have to be built into the adaptive
feedback concept.
Adaptive radar might then have a control loop (Figure 48-7)
where the radar parameters are adjusted to maintain a desired
signal-to-noise ratio (i.e., target detection performance).
A fully adaptive approach also allows parameters such as
the PRF to be reduced when, for example, a target is flying a

Tracking and Fusion System


Sensor System
Sensing Hardware:
Received Waveforms

Detection Process:
Data Rate Reduction

Sensor
Data
Sensor
Control

Track Initiation:

Track Processing:

Multiple Frame
Track Extraction

- Track Cancellation
- Object Classification/ID
- Track-to-Track Fusion

A Priori Knowledge:
- Sensor Performance
- Object Characteristics
- Object Environment

Sensor Data to
Track Association

Signal Processing

Track Maintenance:

Parameter Estimation

Prediction, Filtering
Retrodiction

Track File
Storage
Man-Machine Interface:
- Object Representation
- Displaying Functions
- Interaction Facilities

Figure 48-7. This shows a schematic example of the processing for an adaptive radar tracking system in which the radar parameters are dynamically
and continually adjusted to optimize the radar performance, also making use of a priori knowledge of the sensor and the target scene.

CHAPTER 48: Emerging Radar Trends

straight and steady course and the track updates can be less
frequent with little or no loss in performance. In this way radar
resources can be deployed more flexibly. In other words, fully
adaptive radar processing can also play an integral role within
radar resource management. Indeed, there is an increasing
interaction between radar performance and functions facilitated by software control of radar parameters.
Adaptive radar waveform design and optimization is also
beginning to find its way into target classification. The principles are similar to those described in the tracking radar
example except the criterion for success is now classification performance rather than detection. In detection a good
waveform design is one that has an appropriate range and
Doppler resolution with low sidelobes. The goal of the waveform design is to achieve a signal to noise ratio that ensures
confident detection performance. This can be subject to feedback where the design is optimized until the desired detection
performance is reached. In adaptive classification the feedback aims to select the design of waveform that optimizes
classification performance, and this is not necessarily the one
that results in best detection.
In some ways this might be thought of as counterintuitive since
waveforms with high resolution and low sidelobes should provide an HRRP or imagery with best parameters. However, this
assumes that a radar system operating as a coherent sensor
using electro-magnetic radiation in the RF spectrum will see
the world as we are used to seeing and interpreting it in the
visual part of the spectrum using noncoherent imaging. Our
current knowledge of these issues is poor, but conventional
radar thinking is now being challenged, and simple demonstrations showing improved performance have already been
reported. Certainly, the relationship between the way a radar
illuminates a target and the ability to extract information that
characterizes the target requires an improved understanding.
Equally, the way in which received echoes can be processed
to extract information characteristic to target type also has
great scope for improvement. Again, it can be concluded that
there are significant benefits to be gained from an adaptive
approach.
Overall, fully adaptive radar is only just emerging as a research
topic but is an exciting one that has great potential for improving radar performance in many ways and to improve the range
of tasks that a single radar can carry out. Coupled with electronic scanning, the possibilities seem limitless for enhanced
radar capability, but there remains much and challenging
research before this potential can be realized.

48.6 Cognitive Radar Sensing


Cognitive sensing is an emerging strand of radar research
that aims to tackle the challenges of interpreting and exploiting echoes. It builds on the ideas of fully adaptive radar but

693

694

PART IX: Special Topics and Advanced Concepts

extends them by drawing on human and other animal forms


of cognition. Indeed, cognitive computing and more general
sensing and signal processing are making rapid advances in
other related disciplines. Radar is but one example where a
cognitive approach has the potential to both enhance performance and also may open up new applications area, especially
those requiring autonomy.
Cognitive sensing has embedded within it the notion of the
perception-action cycle that is at the heart of the cognitive process. By perception we mean the picture of the world
viewed by the radar that is created internal to the radar processor. This is akin to our own artificial perception of the world
that we create in our brain. Thus a radar perception might be
the mapping out of aircraft in the sky.
Once a sufficiently accurate perception has been created,
it can be used as a basis for decision-making. In air traffic
control this might be relocating aircraft so that they remain
at appropriately safe separations. The action part is provided through an instruction that repositions the aircraft to
the desired locations. This perception-action activity is a task
currently performed by air traffic controllers. It is they who
supply the necessary cognition. However, it is a task that, in
principle could be carried out via the radar processor issuing
commands to pilots on an automatic basis. Whether or not a
cognitive component of this type becomes part of air traffic
management systems depends on many factors (not least that
of public acceptability), but it does illustrate that radar systems
are not so far away from being able to operate on a more
autonomous basis.
Cognitive radars sensing goes hand in hand with adaptive
feedback. It includes received echoes providing instructions
for the design of future waveforms. It also includes the repositioning of the radar platform and direction of illumination.
Furthermore, it implies slightly more subtle requirements, such
as the explicit generation and exploitation of memories.
Memory is a fundamental component of cognition, but memory is not generally used in radar signal processing. The generation of memories can occur at many different levels. For
example, predecessor pulses and subsequent detections can
confirm a detection in a surveillance radar. Long-term memories based upon prior experience or missions can be used
to create a database of target signatures to aid classification.
Indeed, such a library could be subject to continuous updating. Perhaps future radar systems will be subject to life-long
learning! Third-party databases, including Internet sources,
can also provide a source of memory. For example, the maps
in a geographical information system (GIS) can be used to plot
an optimal trajectory for airborne radar that maximizes some
aspect of performance.

CHAPTER 48: Emerging Radar Trends

It can also be seen that there is a need for the radar itself to be
able to understand and respond to its perception of its environment. Potentially, it must do this on a timescale down to the
milliseconds at which pulses are transmitted.
Cognitive radar sensing is also taking its cue from observations of cognition in natural systems. The cognitive processing
architecture could look like that shown in Figure 48-8. This
architecture has a number of features not seen in the examples
discussed elsewhere in this book, such as multiple feedback
loops, multiple parallel lines of processing, the dynamic creation and exploitation of memories, the link between sensing
and action (the perception-action cycle), and the requirement
for perpetual training and learning.
Cognitive sensing offers both enormous potential for improved
performance but at the same time presents significant challenges. In the space available here we can only present just an
outline description of cognitive sensing. This barely touches
upon the myriad of options for processing echoes and turning them into a well understood picture for reliable decisionmaking. It remains to be seen how much progress will be
made on this topic, but the potential makes for exciting possibilities. If a radar system has an accurate picture and understanding of its surroundings then, for example, it could lead
to much greater autonomy. Perhaps a miniature aircraft, no
larger than a small bird and guided by a combination of GPS
and radar, can carry out a mission to survey the inside of a
building too damaged to allow human investigation. Perhaps
radar will play a vital role in the goal of collision-less automobiles. Are these and other examples beyond the realms of
possibility? Perhaps, perhaps not.

48.7 Other Trends?


Taking these ideas still further, and combining them with the
distributed radar concepts of Chapter 45, we can imagine a
cognitive radar network using platforms operating multistatically, exploiting ground-based transmitters of opportunity as
well, and configuring itself dynamically according to the target
scene and its mission. There are numerous challenges, including geolocation and synchronization, communication between
the nodes of the network, and the control and management of
such a network.

48.8 Summary
Comparing a modern multifunction airborne radar with the
very first examples, more than seventy-five years ago, we can
see that the sophistication and performance are immeasurably
greater, and we can confidently expect that trend to continue.
Although we are always constrained by the laws of physics, the
only other limit is our imagination.

695

Context

Training and Learning

Event
Memory

Goals

Working
Memory

Episodic
Memory

Internal
Control

Planning

Perception

Internal
Needs

Actions

Sensing

External
Control

Behavior

Reaction

Radar

External
Needs

Effectors

Soma

Adaptation

The World

Figure 48-8. This shows a bio-inspired architecture for cognitive


radar sensing, including the different stages of the processing and
the multiple feedback loops within the overall process.

696

PART IX: Special Topics and Advanced Concepts

Further Reading
S. Haykin, Cognitive Radar: A Way of the Future, IEEE Signal
Processing Magazine, Vol. 23, No. 1, January 2006.
J. Guerci, Cognitive Radar: The Knowledge-Aided Fully Adaptive
Approach, Artech House, 2010.
G. Capraro, A. Farina, H. D. Griffiths and M. C. Wicks,
Knowledge-Based Radar Signal and Data Processing:
A Tutorial Introduction, IEEE Signal Processing Magazine,
Vol. 23, No. 1, January 2006.

You might also like