IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO.

1, MARCH 2017 1

Radar-Communications Convergence: Coexistence,

Cooperation, and Co-Design
Alex R. Chiriyath, Student Member, IEEE, Bryan Paul, Student Member, IEEE,
and Daniel W. Bliss, Fellow, IEEE

Abstract—In this paper, we introduce a radar information In this work, we delineate solutions to spectral conver-
metric, the estimation rate, that allows the radar user to be gence using three categories: coexistence, cooperation, and
considered in a multiple-access channel enabling performance co-design. We define coexistence methods as those that bur-
bounds for joint radar-communications coexistence to be derived.
Traditionally, the two systems were isolated in one or multiple den radar and communications transceivers to treat one another
dimensions. We categorize new attempts at spectrum-space-time as interferers. For these methods, any information required to
convergence as either coexistence, cooperation, or co-design. The mitigate the other system’s interference is not shared, and must
meaning and interpretation of the estimation rate and what it be estimated.
means to alter it are discussed. Additionally, we introduce and Cooperative techniques are techniques where some knowl-
elaborate on the concept of “not all bits are equal,” which states
that communications rate bits and estimation rate bits do not edge is shared between systems in order to more effectively
have equal value. Finally, results for joint radar-communications mitigate interference relative to one another. In this regime,
information bounds and their accompanying weighted spectral the systems may not significantly alter their core operation, but
efficiency measures are presented. willingly exchange information necessary to mutually mitigate
Index Terms—Joint radar-communications, SSPARC, radar interference.
information theory, performance bounds. Co-design we define as the paradigm shift of considering
communications and radar jointly when designing new systems
I. I NTRODUCTION to maximize their joint performance. Co-designed systems are
jointly designed from the ground up, and now have the oppor-
ADAR and communications have typically been devel-
R oped in isolation. A growing interest of electromagnetic
radio frequency (RF) convergence is driving the future growth
tunity to improve their performance over isolated operation.
For example, communications users can use codes that are
invariant or even beneficial to communications operation, but
and operation of both class of systems [1]. Recently, with also benefit radar-like operation for known training sequences.
growing spectral congestion concerns, researchers have begun Simultaneously, radar processing can improve channel esti-
investigating methods of spectrum-space-time harmony. We mation to assist in equalization for communications systems.
define radar communications RF convergence to be the operat- Future users will find it advantageous to consider co-designing
ing point at which a given bandwidth allocation is used jointly systems to handle complicated RF convergence scenarios such
for radar and communications to mutual benefit. as the one shown in Figure 1.
Achieving RF convergence for joint radar-communications
coexistence is incredibly complicated. Even for a simple case
involving a single radar and communications link, one must A. Contributions
consider spatial, spectral, and temporal degrees of freedom. We present the joint radar-communications problem as a
In practice, there are many contributing sources in a given joint information problem. Information is chosen because it
spectrum-space-time, and regulatory restrictions may not ade- forces one to identify uncertainty in the situation and develop
quately protect both users even if isolation is acceptable. An plans to reduce it. Estimation theory and signal process-
example of the type of complicated scenario that is associated ing are often presented with traditional metrics such as the
with achieving RF convergence is shown in Figure 1. Cramér-Rao lower bound (CRLB), minimum mean-squared
Manuscript received April 8, 2016; September 29, 2016 and January 25, error (MMSE), or signal-to-noise ratio (SNR). However, none
2017; accepted January 30, 2017. Date of publication February 14, 2017; date of these address information gained from spectral access.
of current version April 7, 2017. This work was sponsored in part by DARPA When focusing on reducing estimation variance, if the infor-
under the SSPARC program. The associate editor coordinating the review of
this paper and approving it for publication was C. Clancy. mation gained through estimation is minimal, precious spec-
A. R. Chiriyath and D. W. Bliss are with the Bliss Laboratory of trum in a given space-time is being inefficiently utilized.
Information, Signals, and Systems and the Center for Wireless Information The radar information measure to be used in this work is
Systems and Computational Architectures, Arizona State University, Tempe,
AZ 85281 USA (e-mail: [email protected]). denoted estimation rate, which was first defined in [2] and
B. Paul is with the Bliss Laboratory of Information, Signals, and extensively discussed in [3] and [4]. These works defined
Systems and the Center for Wireless Information Systems and Computational the quantity mathematically, and then presented a series of
Architectures, Arizona State University, Tempe, AZ 85281 USA, and also
with the General Dynamics Mission Systems, Scottsdale, AZ 85257 USA. cooperative joint radar-communications inner bounds on per-
Digital Object Identifier 10.1109/TCCN.2017.2666266 formance. The result is a multiple-access information map,
2332-7731  c 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/
redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
2 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

Fig. 1. An example highlighting the difficulties of achieving RF convergence. Fig. 2. The joint radar-communications system ‘basic multiple-access sce-
Future systems must be co-designed to not just mitigate interference, but nario.’ This is a simplified version of the complicated RF convergence scenario
jointly consider each other in their inherent operation. shown previously. However, it provides a point of departure for discussing
future work, and enables tractable, intuitive solutions presented here.

providing systems engineers a region of achievable joint

• Only one radar target is present
operation within a given spectrum-space-time.
• Target range or delay is the only parameter of interest
To provide a tractable solution to achieving RF convergence,
• Target cross-section is well estimated
we define what we call the ‘basic multiple-access scenario.’
• Communications signal is received through an antenna
It is a simple scenario involving a radar and communica-
sidelobe; Antenna gains are not identical.
tions user attempting to use the same spectrum-space-time.
In this paper, we provide a clearer understanding of the esti-
This scenario is instructional, and can easily be scaled to
mation rate metric and also look at the optimality (in terms
more complicated scenarios by using it as a building block
of spectral efficiency) of various joint radar-communications
to construct real world examples. We present a diagram of the
performance bounds. The main goals of this paper are three-
‘basic multiple-access scenario’ in Figure 2. In this scenario,
fold
the joint radar-communications system consists of an active,
• Provide an intuitive understanding of radar estimation rate
mono-static, pulsed radar and a single user communications
and the implications of altering it
system. We consider the joint radar-communications receiver
• Introduce and elaborate on the concept of ‘Not All Bits
to be a radar transmitter/receiver that can act as a communica-
are Equal’
tions receiver. The joint receiver can simultaneously estimate
• Use weighted spectral efficiency to highlight the optimal-
the radar target parameters from the radar return and decode a
ity of joint radar-communications performance bounds.
received communications signal. While the node architecture
can easily be generalized to function as a communications
relay by including a communications transmitter, this is not B. Background
explicitly discussed in this paper. We refer to the scenario Along with the estimation rate, the joint radar-
described in Figure 2 throughout the rest of this paper. communications performance bounds discussed in this
Despite the difficulty in achieving RF convergence in sce- paper were first defined in [2]. The estimation rate was
narios such as the one seen in Figure 1, we do know that some extended in [3] to include Doppler measurement and con-
important assumptions have to be made in order to develop a tinuous signaling radars. In [4], the metric was defined in
tractable solution. Those key assumptions made in this work more detail, along with additional inner bounds on joint-radar
are as follows communications performance. Various extensions to the
• Radar and communications operate in the same frequency joint radar-communications inner bounds were presented
allocation simultaneously in [5]–[9]. New inner bounds were developed in [10] for
• Joint radar-communications receiver is capable of simul- waveforms that were jointly optimized for simultaneous
taneously decoding a communications signal and estimat- radar and communications operation. In [11], we use the
ing a target parameter I-MMSE [12] to develop bounds on the radar estimation rate
• Radar detection and track acquisition have already taken (radar mutual information) from MMSE estimation bounds.
place. These bounds are useful when closed form solutions to the
On top of the assumptions made above, the key assumptions estimation rate do not exist. Reference [13] provides an in
made that apply for the scenario described in Figure 2 are as depth survey of work done in various fields that are relevant
follows to joint sensing and communications with a emphasis on joint
• Radar system is an active, single-input single-output radar-communications.
(SISO), mono-static, and pulsed system Achieving radar-communications RF convergence is com-
• Radar system operates without any maximum unambigu- plicated, and so the solution space tends to be greatly varied.
ous range Nevertheless, certain methods are gaining more traction than
• A single SISO communications transmitter is present others.
CHIRIYATH et al.: RADAR-COMMUNICATIONS CONVERGENCE: COEXISTENCE, COOPERATION, AND CO-DESIGN 3

Waveform design has become a dominant research thread in design using information for statistical scattering targets [53].
joint radar-communications phenomenology. Researchers have Recent results have found connections between information
considered a variety of waveform options including orthogo- theory and estimation theory, equating estimation informa-
nal frequency-division multiplexing (OFDM) [14]–[21]. Most tion and the integrated MMSE [12]. The work presented
of these results are attempting co-designed systems, where in [54] develops an expression for radar capacity (for radar
OFDM waveforms are used for bi-static communications, and systems performing target detection only) which, in combina-
as a mono-static radar. However, results showed conflicting tion with the traditional communications capacity, can be used
cyclic prefix requirements, data-dependent ambiguities, and to measure the total capacity of a joint radar-communications
trouble mitigating peak-to-average power ratio (PAPR) for network. In addition, cognitive radar architectures have been
typical radar power requirements. Similar to OFDM, spread proposed using information to prioritize physical location
spectrum waveforms have been proposed for their attractive, access based on uncertainty [55]. These advances make the
noise-like autocorrelation properties [22]–[24]. MIMO radar joint consideration of radar and communications information
techniques have also been proposed, given that the indepen- interesting when considering co-designed solutions.
dent transmitted waveforms allow more degrees of freedom
for joint radar-communications co-design [25]–[27]. II. C OMMUNICATIONS R ATE
Other researchers looked at spatial mitigation as a means
In this section, we present a brief exposition of commu-
to improve spectral interoperability [28]–[30]. These methods
nications capacity theory to lay groundwork for the sections
can be considered either coexistence or cooperative, depending
to come. The goal is to understand the basic communications
on the amount of information shared between users. However,
phenomenology and to understand dealing with systems in an
this is merely a form of spatial isolation managed by radiation
information theory context. This section serves as a useful
patterns. Another method of isolation utilized polarization for
bridge to discuss radar information theory in the next section,
co-designed systems [31]. Space-time dynamic isolation tech-
and forms the basis of how we consider the joint system.
niques have been proposed, such as communications devices
The communications rate capacity is formally defined as
communicating carefully to avoid spectrum-space-time colli-
the supremum of achievable communications rates for a given
sions with rotating radars [32], [33]. These also varied from
channel model with respect to the input distribution. It tells
coexisting to cooperative systems. An overview of interference
us how much information as a function of time we can com-
mitigation techniques that aid in isolation between WiMax
municate with arbitrarily low bit error rate. This problem was
networks and ground-based radar systems is provided in [34].
solved by Shannon in his seminal work [51].
Employing the existing cellular framework has also been
proposed as a solution to augment the dwindling radar spec-
trum [35], [36]. These approaches range from radar systems A. Communications Rate Capacity for a Single Link
subscribing as cellular users when there is a need for radar illu- For our basic multiple-access scenario, we have a single
mination, to using cellular protocols to prioritize radar tasks. communications user. Here, we present Shannon’s results for
As such, the radar is conforming to the design of the cellular the capacity of this link, assuming the user is operating with no
user, and is subsequently closer to cooperation than co-design. interference. We assume we have a single wireless communi-
Advancements in cognitive radios and radar have been cations link in a continuous memoryless real Gaussian channel
proposed as a natural solution to spectrum congestion prob- with an average power constraint Pcom and fixed bandwidth B
lems [37]–[41]. Cognitive radio has been advancing spectral and subject to receiver thermal noise. The capacity of such a
sharing potential in the communications realm [42]. However, channel was shown by Shannon to be [51]
RF convergence between radar and communications users is  
1 b2 Pcom
largely an open area of research. These two systems, unlike Rcom ≤ log2 1 +
the cognitive radio user base, have vastly different goals, met- 2 Ts kB Ttemp B
rics, and operators. Joint coding techniques, such as robust 1
= log2(1 + SNR), (1)
codes for communications that have desirable radar ambi- 2 Ts
guity properties, as well as codes that trade data rate and
where Ts = 2B 1
is the independent sampling rate of the band-
channel estimation error have been investigated as co-design
limited system, b is the combined gain and communications
solutions [43]–[46]. Research has been done investigating the
propagation loss product, kB is the Boltzmann constant, and
effects of passive and parasitic radar systems that passively
T temp is the absolute temperature.
exploit communications illuminations [47]–[50]. For example,
some systems employ multiple orthogonal radar waveforms
with embedded communications symbols and exploit the dif- B. Altering the Communications Rate
ferential phase between waveforms to extract the parasitic data As we have stated previously, the communications rate is
transmission [47], [48]. simply a measure of the amount of arbitrary information that
Information is well known in communications phenomenol- can be transmitted through the channel given spectrum-space-
ogy, but less so in radar. Perhaps surprisingly, radars were time access. We can increase the communications rate in a
looked at in the context of information theory soon after fixed bandwidth by:
Shannon’s seminal work [51] by Woodward [52]. Interest 1) Changing Source Entropy: The source entropy is dic-
resurged many years later with Bell’s work on waveform tated by the source distribution p(X). The more we increase
4 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

this entropy, the larger the mutual information [56]. While this variance given by nτ,proc 2  = στ,proc
2 . The radar estimation
may appear beneficial, in doing so, we may exceed the aver- rate can be extended to include estimation of different tar-
age power constraint, violating the maximizing terms of the get parameters as seen in [3], in which the estimation rate is
capacity problem. Ignoring the mutual information construct, extended to take into account Doppler estimation.
we can attempt to communicate at a faster rate (rate taking The estimation rate is formally defined as the quantity that
into account redundant and non-redundant information [56]). represents the minimum number of bits needed to encode the
However, exceeding the capacity means an arbitrarily low bit Kalman residual, which is the statistical deviation from the
error rate (BER) is not achievable. As a result, the spectral radar prediction of a target parameter, for a given channel
efficiency in b/s/Hz goes down when considering a channel degradation [4]. The estimation rate tells us how much infor-
with an arbitrarily low BER. mation we stand to gain once we subtract the prediction of
However, if the capacity is not exceeded, we achieve the the target’s parameter, since the predicted target parameter is
maximum spectral efficiency given the problem parameters. already known and does not truly convey any information.
Thus information must be carefully considered as to the root
meaning when trading this parameter, as we see in Section III A. Estimation Rate for a Single Target
when considering radar estimation rate. Considering the radar channel to act as an uncooperative
2) Changing SNR: From Equation (1), we see that by communications channel, the process noise, nτ,proc , is the
increasing the SNR, we get a net gain in information. Sphere information X being transmitted. This transmitted informa-
packing is a good analogy. In an average power-constrained tion X is to be degraded by the addition of some noise N,
channel with fixed bandwidth, this amounts to decreasing the which for target parameter estimation is given by the radar
noise power. As a result, more “levels” can be transmitted estimation error, nτ,est , and a noisy measurement of X is
and resolved on average at the receiver, meaning more entropy received at the radar receiver system. The estimation rate for
states and overall more information. Thus by increasing SNR, our uncooperative channel is therefore given by [56]
we can increase the source entropy level at which an arbitrary
I(X; X + N)
BER is possible. If less throughput is needed, the bandwidth Rest = , (2)
can be reduced (noting the non-linear mapping), or the com- Tpri
munications system can be duty cycled in time. This equates where Tpri = Tpulse /δ is the pulse repetition interval of the
to spectrum-time isolation. radar system, Tpulse is the radar pulse duration, and δ is the
As we see, changing the rate of communication cannot be radar duty factor.
done arbitrarily, as bit error rates may preclude proper system Assuming that the radar estimation error, nτ,est , is Gaussian
operation. Increasing the communications rate through SNR is with variance nτ,est 2  = στ,est
2 , the mutual information can

acceptable, but requires reduction of the thermal noise floor, be shown to be [56]
or a change in the channel constraints. As we see in the next  
1 στ,proc
2
section, increasing the complementary radar estimation rate Rest ≤ log2 1 + 2 . (3)
must also be done with careful consideration to proper system 2 Tpri στ,est
operation and estimation performance. It should be noted that a Gaussian distribution is used to model
the radar estimation error, nτ,est , and process noise, nτ,est ,
because Gaussian distributions have a closed form solution to
III. R ADAR E STIMATION R ATE
entropy, enabling a closed form solution for the estimation rate
In this section, we present a novel parametrization of radar to exist. In radar estimation problems where a Gaussian dis-
performance, the estimation rate. The estimation rate is a tribution is not appropriate and a closed form solution for the
metric analogous to the communications rate and provides estimation rate does not exist, bounds on the radar estimation
a measure of the information about a target that is gained rate (radar mutual information) exist which can still capture
from radar illumination in radar tracking estimation scenar- a measure of radar information [11]. We leave the estimation
ios. In general, the target has some entropy or information rate as an inequality because typically systems must perform
about itself that is not explicitly being communicated to the certain non-ideal processing steps, such as quantization. As a
radar system by the target. Radar illumination can be viewed result, the data-processing inequality is enforced [56].
as the target unwillingly communicating this target entropy στ,proc
2
Looking at the ratio of variances , we see that
or information to the radar receiver. Thus, the radar chan- στ,est
2

nel can be characterized as an uncooperative communications στ,proc

2 = nτ,proc2 is the power of the transmitted informa-
channel [4]. tion, X2 , and that στ,est
2 = nτ,est 2  is the noise power,
We assume that the radar system has some knowledge of σ2
N2 . Thus, the ratio of variances τ,proc represents the SNR
the target parameter of interest, based on prior observations, στ,est
2

up to some fluctuation. This fluctuation, also called process of the uncooperative communications channel that is used to
noise, is modeled by a Gaussian random variable nparameter,proc characterize the radar channel. This is more evident when com-
with variance given by nparameter,proc 2  = σparameter,proc
2 . paring Equation (3) to Equation (1). Thus, Equation (3) can be
Throughout the rest of the paper, we only consider radar range written as [2]
estimation. In such a case, the process noise for range fluc- 1
Rest ≤ log2(1 + SNR). (4)
tuation is interpreted as a fluctuation in delay, nτ,proc with 2 Tpri
CHIRIYATH et al.: RADAR-COMMUNICATIONS CONVERGENCE: COEXISTENCE, COOPERATION, AND CO-DESIGN 5

If we assume that the radar estimator achieves the CRLB, estimation rate in this manner, enhances the target parameter
the variance of delay estimation, στ,est
2 , is given by the CRLB estimation quality of the radar which is always desirable.
for time delay estimation and Equation (3) can be written as As we have seen above, an increase or decrease in the esti-
1    mation rate is neither strictly good nor bad, rather it is the
Rest ≤ log2 1 + 8π 2 στ,proc
2
B2rms ISNR , manner in which the estimation rate was altered that can be
2 Tpri beneficial or detrimental to the joint radar-communications
1   
= log2 1 + 8π 2 στ,proc
2
γ 2 B2 ISNR , (5) system. If the estimation rate is decreased by lowering the
2 Tpri process noise, then the radar system needs less spectral access
which results in a less congested spectrum (aids in radar-
where Brms is the full RMS bandwidth of the system, ISNR is communications spectrum sharing). However, if the process
T B Prad, received
the integrated SNR given by ISNR = pulsekB Ttemp B where noise is arbitrarily increased by ignoring prior information
Prad, received is the received radar signal power, and γ is the (a physical predictive model, for example), then we gain more
scaling constant between B and Brms that is dependent upon information through measurement, but the estimation perfor-
the shape of the radar waveform’s power spectral density. γ mance is degraded and radar system performance is lowered.
is related to Brms and bandwidth B as follows However, increasing SNR increases both the estimation rate
and estimation parameter performance.
γ 2 B2 = (2π )2 B2rms . (6)
There is a trade-off between reducing radar spectral access
For a flat spectral shape, γ 2 = (2π )2 /12. and increasing target parameter estimation quality. On one
It should be noted here that if the delay estimator does not hand reducing the estimation rate by reducing process noise
achieve the CRLB, the estimation rate is lowered. frees up more of the spectrum to be used by communications
systems, aiding spectrum sharing, whereas increasing the esti-
mation rate by improving target parameter estimation quality
B. Altering the Estimation Rate increases the radar system performance. Accordingly, attempts
As we have stated before, the estimation rate is simply a should be made to maximize the estimation rate from an SNR
measure of the amount of information about the target that perspective, while jointly considering estimation error perfor-
can be gained through the radar channel through illumination. mance. That is, estimation error should never be increased
Thus, an increase in the estimation rate implies an increased to increase the estimation rate, but steps to maximize the
amount of information about the target is gained by the radar mutual information for a fixed process noise should always
system through the channel. As we see, increasing the esti- been taken.
mation rate can lead to better target parameter estimation
performance whereas reducing the estimation rate can result
in reduced spectral access. From Equation (3), we see that the IV. J OINT R ADAR -C OMMUNICATIONS ACCESS
estimation rate can be altered by: Now that we have seen both the communications user
1) Changing Process Noise: Process noise represents and the radar user in an information theory context, we
the amount of information of the target that is unknown. jointly consider the systems here. We start an exposition of
From Equation (3), we see that by increasing the process the multiple-access communications performance bound for
noise, we increase the estimation rate. Increasing the process two communications users as motivation to develop inner
noise essentially means that the target behaves in an unex- bounds on the performance of a joint radar-communications
pected manner when compared to how the target was modeled system. We then present a number of inner bounds on per-
by the radar system. Thus, the amount of information that can formance for a joint radar-communications system and the
be gained about the target through radar illumination increases signaling schemes that are used to achieve these inner bounds.
and this is reflected via an increase in the estimation rate. The bounds presented in this section have been previously
However, if the target was modeled accurately, then the derived in much greater detail in [2] and [4]. Furthermore, to
information content gained through radar illumination is low simplify the discussion, we consider only a single radar tar-
because much of the true uncertainty about the target was get with delay τ and gain-propagation-cross-section product
bought down by accurately modeling the target. This is ben- a. Additionally, we employ a blending ratio, α, to regu-
eficial since, as seen in [8] and [9], by reducing the process late radar-communications sharing in the temporal or spectral
noise, the radar system can illuminate less frequently. Thus, by dimension.
using a more accurate model of the target and reducing pro-
cess noise, the radar system needs less spectral access which
is beneficial for cooperative radar-communications spectrum A. Multiple-Access Communications Performance Bound
sharing. We consider a scenario in which the channel propagation
2) Changing Estimation Performance: From Equation (3), gain for the first communications system is given by a1 and
we see that by improving the estimation performance or channel propagation gain for the second communications sys-
decreasing the mean-squared estimation error σest2 , we increase tem is given by a2 . The power of the first communications
the estimation rate. By improving the estimation performance, transmitter is denoted by P1 and the power of the second com-
the radar system is able to extract more information about the munications transmitter is given by P2 . The scenario under
target, thus increasing the estimation rate. Increasing the consideration is shown in Figure 3.
6 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

Fig. 5. Isolated sub-band bandwidth split. The noise floor is flat across
the total bandwidth. The radar user and communications user are then given
some complementary fraction of the overall bandwidth B, parameterized by
the blending ratio α.

Fig. 3. Physical multiple access communications system scenario with 2

users. It is assumed both users are occupying the same bandwidth, and their section can be extended for more than two communications
space-time converge at the communications base station. As a result, their systems. For N different communications systems, the resul-
communication rates must be considered jointly. tant achievable rate region is an N-dimensional polytope [56].

B. Isolated Sub-Band Inner Bound

In this section, we derive an inner bound by considering a
scenario in which we partition the total bandwidth into two
sub-bands, one for radar only and the other for communi-
cations only, which is the standard, isolated solution. Each
system functions without any interference in their respective
sub-band [2], [4]. The bandwidth is split between the two
sub-bands according to some blending ratio α such that,
B = Brad + Bcom , Bcom = α B , Brad = (1 − α) B , (8)
as shown in Figure 5. The corresponding communications rate
(for the communications only sub-band) is given by
Fig. 4. Pentagon containing two-user communications multiple-access  
achievable rate region. Lines 1,2 and 3,4 are the rates achieved considering
b2 Pcom
each communications user in an isolated bandwidth. The bisecting diagonal Rcom ≤ Bcom log2 1 + , (9)
is the joint achievable rate. As a result, the convex hull of the three lines kB Ttemp Bcom
constructs the achievable region of two-user communications within a given
shared bandwidth [4, Fig. 2]. where b is the combined gain and communications propagation
loss product defined in Equation (1). The corresponding radar
estimation rate is given by
Their corresponding rates are denoted R1 and R2 . Assuming   
that the noise variance is given by σnoise
2 , fundamental limits 1
Rest ≤ log2 1 + 8 π 2 στ,proc
2
γ 2 B2rad ISNR . (10)
on the communications rate are shown in Figure 4. Vertices 2 Tpri
are found by jointly solving the two bounds to get [2], [4],
   C. Successive Interference Cancellation Inner Bound
a1 2 P1
{R1 , R2 } = log2 1 + , As stated in Section III, we have some knowledge of the
1 + a2 2 P2 radar target parameter (in this case, range or time-delay) up to
 
a2 2 P2 some range fluctuation (also called process noise). Using this
log2 1 + . (7) information, we can generate a predicted radar return and sub-
σnoise
2
tract it from the joint radar-communications received signal.
The other vertex can be found by switching the subscripts After suppressing the radar return, the receiver then decodes
1 and 2 in Equation (7). The region that satisfies these the- and removes the communications signal from the observed
oretical bounds is depicted in Figure 4. The vertices 1 and waveform to obtain a radar return signal free of communi-
3 shown in Figure 4 can be achieved by utilizing an optimal cations interference. This method of interference cancellation
multiuser detection technique called Successive Interference is called SIC. SIC is the same optimal multiuser detection
Cancellation (SIC) [4], [56]. technique mentioned in Section IV-A, except it is now refor-
The achievable rate region is obtained by taking the convex mulated for a communications and radar user instead of two
hull [57] of the vertices 1-4. Because a radar signal return communications users. The block diagram of the joint radar-
is not derived from a countable dictionary, the fundamen- communications system considered in this scenario is shown
tal assumption of a communications signal is violated, and in Figure 6.
the bounds presented here can not be achieved by a joint If Rest ≈ 0 (for example, because of a low power return or
radar-communications system [2]. The result presented in this well modeled target), it is as if the radar interference is not
CHIRIYATH et al.: RADAR-COMMUNICATIONS CONVERGENCE: COEXISTENCE, COOPERATION, AND CO-DESIGN 7

Fig. 6. Joint radar-communications system block diagram for SIC scenario.

The radar and communications signals have two effective channels, but arrive
converged at the joint receiver. The radar signal is predicted and removed,
allowing a reduced rate communications user to operate. Assuming near
perfect decoding of the communications user, the ideal signal can be recon- Fig. 7. Constant information radar time-sharing scheme. For the kth iteration
structed and subtracted from the original waveform, allowing for unimpeded of cycle time T, part of the cycle is allocated to radar operation, where a
radar access. reduced rate communications user is operating using SIC. This allocation
depends on whether the radar information measure based on observations
from previous cycle times, Ik | k−1 (x; y), is above or below a certain threshold,
Iconst . For the remainder of the cycle, the communications user is free to
present and the communications system can operate at a data operate without any radar emissions. Note that the radar access time can be
rate determined by the isolated communications bound, fixed at the duration for a single spectral access, and then the cycle time can
  be varied.
b2 Pcom
Rcom ≤ B log2 1 + . (11)
kB Ttemp B slower at the SIC node. Note that the radar access time or
If the estimation rate is non-trivial, then the residual con- CPI can be fixed at the duration for a single spectral access,
tributes to the communications system’s noise floor. We can and then the cycle time can be varied. α is increased when
mitigate this by reducing Rcom for a given transmit power. the target is well-modeled, or the SNR is low. In the former
After subtraction of the predicted radar return, the receiver case, there is little information gained through measurement,
can decode the communications signal. With knowledge of the since the target is well predicted [8], [9]. For the SNR, mea-
communications system, forward error correction and spectral surement noise dominates the entropy, and very little “good”
shaping can be reapplied, and the radar system can remove information is obtained through access of the spectrum. The
the ideal communications signal from the observed wave- rates are then given by
 
form, leaving just the radar return. Thus, radar parameter b2 Pcom
estimation can be done without corruption from any out- Rcom ≤ (1 − α)B log2 1 +
side interference. This implies that from the communications σint+n
2
 
receiver’s perspective, it observes interference plus noise as b2 Pcom
described by σint+n
2 = Prad a2 B2rms στ,proc
2 + kB Ttemp B [2], + αB log2 1 + , (13)
kB Ttemp B
[4], where a is the radar gain-propagation-cross-section prod-
1   
uct as defined in the beginning of Section IV, and the Rest ≤ (1 − α) log2 1 + 8 π 2 στ,proc
2
B2rms ISNR .
corresponding communications rate is given by 2 Tpri
  (14)
b2 Pcom
Rcom ≤ B log2 1 + . (12)
σint+n
2
D. Communications Water-Filling Bound
In this regime, the corresponding estimation rate bound Rest In this section, we consider a scenario in which the total
is given by Equation (5). The SIC inner bound is given bandwidth is split into two sub-bands, one sub-band for com-
by connecting the points given by Equations (5), (11), and munications only and the other sub-band for both radar and
(12). This is equivalent to time sharing between full band communications. It is not necessary that the sub-bands be of
SIC operation (normal radar, reduced communications), and equal bandwidth. We use water-filling to distribute the total
communications only (no radar). In [8] and [9], the con- communications power between the two sub-bands [2], [4].
stant information radar (CIR) algorithm was developed which Water-filling optimizes the power and rate allocation between
proposed to modulate radar spectral access based on the esti- multiple channels [56], [58]. The mixed use channel oper-
mation rate measure, or more specifically, the radar estimation ates at the SIC rate vertex defined by Equations (5) and (12).
information, I(x; y). The goal was to delay target revisit until The block diagram of the joint radar-communications system
an arbitrary number of bits of information about the target, considered in this scenario is shown in Figure 8.
Iconst , would become available. This fixed the information rate As in the isolated sub-band case, we use the blending ratio
locally around radar access, enforcing a spectral efficiency for α to split the overall bandwidth B into a communications only
allowing radar access. During periods when the target was sub-band, and a mixed sub-band operating at the SIC node:
well-modeled, or the SNR was low, insufficient information
could be obtained, and so the communications user is allowed B = Bcom + Bmix , Bcom = α B , Bmix = (1 − α) B. (15)
to broadcast freely for a longer period of time. Figure 7 shows We then optimize the power utilization, β, between sub-bands,
this scheme in action. Our blending ratio α now modulates
radar time access or radar coherent processing interval (CPI), Pcom = PCO + PMU , PCO = β Pcom , PMU = (1 − β) Pcom .
in which the communications user is allowed to communicate (16)
8 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

The mixed use communications rate inner bound, Rcom,MU , is

given by
 
b2 (1 − β) Pcom
Rcom,MU ≤ Bmix log2 1 + , (21)
σint+n
2

where σint+n
2 is given by Equation (18). The corresponding
radar estimation rate inner bound is then given by
1   
Rest ≤ log2 1 + 8 π 2 στ,proc
2
γ 2 B2mix ISNR . (22)
2 Tpri
Fig. 8. Joint radar-communications system block diagram for commu-
nications only and mixed use sub-bands. One band is operating only for E. Not All Bits Are Equal
communications, and is spectrally isolated from the radar operation. The other As stated on numerous occasions through out this paper, the
sub-band is operating using SIC, where the communications and radar RF
energy converge at the receiver. The optimal power split is determined using communications and estimation rates represent the amount of
water-filling. information, in bits, gained through the respective channels
through message transmission or radar illumination. However,
the bits that are used to represent the amount of information
gained for each system can be prioritized differently and the
information rate metrics do not clearly highlight this.
For the multiple-access communications system described
previously, an increase in performance by 1 bit for the first
communications system may not be as critical as an increase
in performance by 1 bit for the second communications system
and vice versa. For example, in Figure 4, if the first commu-
nications system with rate R1 represents a user receiving an
emergency broadcast message and the second communications
system with rate R2 represents a local Wi-Fi network connec-
Fig. 9. Water-filling bandwidth allocation. As in the isolated sub-band case, tion, an increase in R1 by 1 bit is more critical than a similar
we use our blending ratio α to parameterize the allocation of bandwidth
between the communications only user, and the joint radar-communications improvement in R2 .
band operating using SIC. After the radar is predicted and subtracted, the A similar case exists for joint radar-communications sys-
mixed band has a radar residual contributing to the communications noise tems as well. As we see in the next section, we can use the
floor. We then have two channels with differing noise degradations, and the
normal water-filling solution follows. estimation rate to generate achievable rate regions for the joint
radar-communications system, such as in Figure 4. The bits
used by the radar system can have more value or priority than
the bits used by the communications system and vice versa.
The water-filling power and bandwidth allocation, and the SIC
Another consequence of the bits not being equal is that
algorithm are all shown in Figure 9. There are two effective
there is an intrinsic disparity in power required to increase the
channels
estimation rate by 1 bit when compared to the communications
b2 b2 rate. A bit of higher value (or priority) may require more power
μcom = , μmix = , (17) for a 1 bit performance increase when compared to the other
kB Ttemp Bcom σint+n
2
system.
σint+n
2
= a2 Prad γ 2 B2mix στ,proc
2
+ kB Ttemp Bmix . (18) For example, in Figure 10, consider a joint radar-
communications system in which the communications system
The first for the communications only channel, and the second is used to stream a video and the radar system is monitoring
for the mixed use channel. We apply the water-filling result air-traffic. An increase in the communications rate by 1 bit
derived in [2] and see that the optimal power distribution (β) is not as critical as a similar improvement in the estimation
between the two sub-channels is given by: rate. As highlighted by the examples provided in this discus-
  sion, the importance of bits are application specific. As shown
1 α−1 α in Section V-A, a system engineer can assign priorities to bits
β =α+ + ;
Pcom μcom μmix for different systems and use the complete profile of achievable
α 1 rate regions, such as the ones provided by Figures 4 and 10, to
when Pcom ≥ − . (19)
(1 − α) μmix μcom determine the operating point for a joint radar-communications
system which is the set of appropriate rates for each system.
The resulting communications rate bound in the The importance of this concept is further emphasized when
communications-only sub-band, Rcom,CO , is given by looking at the weighted spectral efficiency of various inner
  bounds on performance against the blending ratio, α in order
β Pcom b2 to find the optimal operating point for both systems, as is
Rcom,CO ≤ Bcom log2 1+ . (20)
kB Ttemp Bcom done in Section V-B. The weighted spectral efficiency, which
CHIRIYATH et al.: RADAR-COMMUNICATIONS CONVERGENCE: COEXISTENCE, COOPERATION, AND CO-DESIGN 9

TABLE I
PARAMETERS FOR E XAMPLE P ERFORMANCE B OUND

is just a linear combination of the communications and esti-

mation rate, can be modified to account for disparate priorities Fig. 10. Multiple access bounds for joint radar-communications access. The
between a communications and estimation bit. Assigning dif- red lines, given by Equations (1) and (3), are created by considering each
user independently in the entire bandwidth, without interference. The isolated
ferent priorities to radar and communications bits alters the sub-band, given by Equations (9) and (10), is represented by the yellow line,
optimal or most spectrally efficient operating point for joint where the blending ratio α is swept from 0 to 1, which allocates the overall
radar-communications differently. Thus, by assigning incor- bandwidth B proportionally to the radar or communications user. The constant
information time share line is represented by the solid blue linear interpolation
rect priorities, both systems could be operating in a spectrally between the SIC node, given by Equations (5) and (12), and the radar-free
inefficient manner thereby wasting precious spectral resources. communications point, given by Equation (11). Finally, the optimal water-
filling solution, given by Equations (20) to (22), is represented by the solid
green line. The proportion of B allocated to communications only and the
V. E XAMPLES mixed-use SIC band is swept with α.

Here we demonstrate the joint radar-communications mul-

tiple access information bounds for the basic multiple-access obtained by sweeping the blending ratio, α from 0 to 1.
scenario for a given parameter set. We go over the key parame-
Changing the blending ratio alters the operating point of a joint
ters, and present the multiple-access bounds similar to previous radar-communications system along the performance bound
works [2], [3], [7]. In order to make a fair comparison, we then curves shown in Figure 10.
plot the weighted spectral efficiencies for the inner bounds
The concept that not all bits are equal discussed
using the cooperative/co-designed techniques, and the equiv- in Section IV-E can also be used to find the appropriate oper-
alent isolated solution. It should be noted that the spectral ating point for a joint radar-communications system from a
isolation bound that is plotted in Figures 10, 12, and 13 is complete profile of achievable rate regions such as the one
a simple, unachievable outer bound that corresponds to each shown in Figure 10. By assigning suitable weights to radar
system utilizing the full bandwidth with out the presence of and communications bits, plotting this information against the
each other, given by Equations (9) and (10) with α = 1 and complete profile of achievable rate regions indicates the appro-
0 respectively. priate operating point for the given joint system. This process
is further highlighted in Figure 11, where we show two cases,
A. Comparison of Joint Radar-Communications one in which a radar bit is worth 10000 communications bits
Performance Bounds and another where a radar bit is worth 4000 communications
In Figure 10, we display an example of the inner bounds bits. The two lines indicate on each inner bound what the
on performance. The parameters used in the example are dis- appropriate operating points are for each radar bit weight case.
played in Table I. In general, the inner bound is produced
by the convex hull of all contributing inner bounds. There B. Weighted Spectral Efficiency of Joint
are some important subtleties with this figure. For example, Radar-Communications Performance Bounds
the CIR time sharing scheme shows a linear interpolation The plots shown in Figures 10 and 11, while useful, does
between the full bandwidth SIC node, and the radar free opera- not give us a notion of how spectrally efficient we are.
tion (communications only). While it shows a linear decrease In Figures 12 and 13, we attempt to do a more fair comparison
in estimation rate, for any given radar spectrum-space-time by looking at the weighted spectral efficiency of each bound.
access, the radar is operating over the full bandwidth, unim- Here, the weighted spectral efficiency of each performance
peded by the communications user. Contrast that with the bound is given by
isolated sub-band (ISB), where traveling along the curve
wR Rest + wC Rcom
toward the communications only axis implies a reduction Eweighted = , (23)
in radar bandwidth, which impacts specific radar parameter (wR + wC ) Btot
estimation [59]. The same applies for water-filling. Finally, where wR is the radar bit weight, wC is the communications
it should be stated that the performance bound curves are bit weight, and Btot is the sum of all the bandwidth consumed.
10 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

Fig. 13. Weighted spectral efficiency (measured here in bits per sec-
Fig. 11. Multiple access bounds for joint radar-communications access ond per kilohertz) plots for joint radar-communications access, weighted for
describing operating point selection. The dashed and dotted black lines rep- importance. In this example, we weighted the radar bits 3000x what the
resent two cases where a radar bit is valued against a communication bit. communications bits are worth. This may be true for certain military radar
The slopes of the dashed lines indicate how much a radar bit is worth when applications, and the weighting may be scenario dependent. With proper
compared against a communication bit. In the case of the dotted line, a radar weighting, the maximum point of spectral efficiency has more meaning and
bit is worth 10000 communications bits and in the case of the dashed line, utility.
a radar bit is worth 4000 communications bits. The solid lines depict the
performance bounds shown in Figure 10. The intersection of a dashed line
against a performance bound indicates the appropriate operating point for a
given radar bit weight.
communications system to achieve the same communications
rate as a co-designed system. For the CIR time sharing scheme,
we solve for the effective isolated bandwidth by solving
 
Beff = R−1
com αRcom,free + (1 − α)Rcom,sic , (25)

where Rcom,free is the communications rate when there are

no other users given by Equation (11), and Rcom,sic is the
reduced communications rate operating at the SIC node, given
by Equation (12). That is, we solve for B in Equation (11),
given that the left hand side is equal to the total communi-
cations rate for a given point along the CIR time share line.
This is the sum of the duty-cycled communications only rate
given by Equation (11), and the complementary duty-cycled
SIC communications rate operating after radar prediction sub-
traction given by Equation (12). We utilize a similar technique
for the water-filling scenario.
Fig. 12. Weighted spectral efficiency plots for joint radar-communications To emphasize the importance of the concept that not all bits
access. Note the accompanying colored dashed lines are the equivalent, iso-
lated weighted spectral efficiencies. For example, the dashed blue line is the are equal that was discussed in great detail in Section IV-E,
weighted spectral efficiency obtained if the two systems were operating at we assign both the radar and communications bits a weight
the same rate given by the CIR time share scheme in the solid blue line, but of 1 and calculate the weighted spectral efficiency, as shown
isolated in frequency. This means the communications user that is operating
after subtraction of the radar would be in its own equivalent band. in Figure 12. Here, with solid and dashed lines represent-
ing the co-designed system and the equivalent isolated system
(systems operating at same rates but isolated in frequency)
In Figures 12 and 13, we look at the weighted spectral effi- respectively, we see that cooperation outperforms isolation for
ciency of the performance bounds discussed in Section IV this case and that the water-filling bound is most spectrally
for co-designed systems as well as their respective equivalent, efficient. However, on recognizing that the blending ratio α is
isolated systems (spectrally isolated systems operating at the the x-axis for this plot, it then becomes clear that the peak
same rates) for a given spectral allocation, B. For these isolated of this plot is not the optimal operating point, given that
systems, the total consumed bandwidth, Btot is given by α = 1 implies no radar use. Since both radar and communi-
cations bits are assigned equal priorities of 1 in this scenario,
Btot = B + Beff , (24)
it is evident that not all bits are equal in this optimization
where B is the bandwidth consumed by an isolated radar sys- process. We can underscore this by setting wR = 3000 and
tem and Beff is the effective bandwidth required by a isolated wC = 1 in Equation (23), as shown in Figure 13. With proper
CHIRIYATH et al.: RADAR-COMMUNICATIONS CONVERGENCE: COEXISTENCE, COOPERATION, AND CO-DESIGN 11

weighting, the maximum in this plot becomes more meaning- [8] B. Paul and D. W. Bliss, “Constant information radar for dynamic shared
ful when considering spectral allocation. It should be noted spectrum access,” in Proc. 49th Asilomar Conf. Signals Syst. Comput.,
Pacific Grove, CA, USA, Nov. 2015, pp. 1374–1378.
that since the radar bits are weighted 3000x what the com- [9] B. Paul and D. W. Bliss, “The constant information radar,”
munications bits are worth, this implies that more power may Entropy, vol. 18, no. 9, p. 338, 2016. [Online]. Available:
be required to increase the estimation rate as compared to the http://www.mdpi.com/1099-4300/18/9/338
[10] B. Paul, A. R. Chiriyath, and D. W. Bliss, “Joint communications and
communications data rate. radar performance bounds under continuous waveform optimization: The
waveform awakens,” in Proc. IEEE Radar Conf., Philadelphia, PA, USA,
VI. C ONCLUSION May 2016, pp. 865–870.
[11] B. Paul and D. W. Bliss, “Estimation information bounds using
In this paper, we describe the problem of radar communi- the I-MMSE formula and Gaussian mixture models,” in Proc. 50th
cations coexistence and describe the challenges in achieving Annu. Conf. Inf. Sci. Syst. (CISS), Princeton, NJ, USA, Mar. 2016,
pp. 292–297.
radar communications RF convergence. We perform a sur-
[12] D. Guo, S. Shamai, and S. Verdú, The Interplay Between Information
vey of previously proposed solutions and by considering the and Estimation Measures (Foundations and Trends in Signal
RF convergence problem as a joint information problem, we Processing). Boston, MA, USA: Now, 2012.
present a novel approach to constructing future solutions. We [13] B. Paul, A. R. Chiriyath, and D. W. Bliss, “Survey of RF
communications and sensing convergence research,” IEEE Access,
develop multiple solutions to cooperative radar and communi- vol. 5, pp. 1–19, 2017.
cations coexistence for a simple multi-access scenario, which [14] C. Sturm and W. Wiesbeck, “Waveform design and signal processing
can be applied to more complicated scenarios. We discuss aspects for fusion of wireless communications and radar sensing,” Proc.
IEEE, vol. 99, no. 7, pp. 1236–1259, Jul. 2011.
our choice of information as a joint performance metric and [15] A. Turlapaty, Y. Jin, and Y. Xu, “Range and velocity estimation of radar
present a novel parameterization of the radar in terms of target targets by weighted OFDM modulation,” in Proc. IEEE Radar Conf.,
information gained through radar illumination, the estimation Cincinnati, OH, USA, May 2014, pp. 1358–1362.
[16] T. Guo and R. Qiu, “OFDM waveform design compromising spec-
rate. We see how the estimation rate is a metric analogous tral nulling, side-lobe suppression and range resolution,” in Proc. IEEE
to the communications data rate. We also note how the infor- Radar Conf., Cincinnati, OH, USA, May 2014, pp. 1424–1429.
mation measured by the estimation and communications rate [17] G. Lellouch, A. Mishra, and M. Inggs, “Impact of the Doppler modu-
lation on the range and Doppler processing in OFDM radar,” in Proc.
in bits for each system can have different values or priority, IEEE Radar Conf., Cincinnati, OH, USA, May 2014, pp. 803–808.
depending on the importance of each system. Using the esti- [18] S. C. Thompson and J. P. Stralka, “Constant envelope OFDM for
mation rate and the communications data rate, we then develop power-efficient radar and data communications,” in Proc. Int. Waveform
Diversity Design Conf., Kissimmee, FL, USA, Feb. 2009, pp. 291–295.
several cooperative signaling schemes that are used to develop
[19] J. R. Krier et al., “Performance bounds for an OFDM-based joint
inner bounds on joint performance. Finally, we compare the radar and communications system,” in Proc. IEEE Mil. Commun. Conf.,
weighted spectral efficiency of various bounds on performance Tampa, FL, USA, Oct. 2015, pp. 511–516.
as well traditional solutions to the RF convergence problem, [20] G. Hakobyan and B. Yang, “A novel OFDM-MIMO radar with
non-equidistant subcarrier interleaving and compressed sensing,” in
such as complete spectral isolation, and observe that coopera- Proc. 17th Int. Radar Symp. (IRS), Kraków, Poland, May 2016, pp. 1–5.
tive and co-design techniques provide high spectral efficiency [21] Y. L. Sit, B. Nuss, S. Basak, M. Orzol, and T. Zwick, “Demonstration of
and outperform traditional solutions. interference cancellation in a multiple-user access OFDM MIMO radar-
communication network using USRPs,” in Proc. IEEE MTT-S Int. Conf.
Microw. Intell. Mobility (ICMIM), San Diego, CA, USA, May 2016,
ACKNOWLEDGMENT pp. 1–4.
[22] X. Shaojian, C. Bing, and Z. Ping, “Radar-communication integration
The views expressed are those of the authors and do not based on DSSS techniques,” in Proc. 8th Int. Conf. Signal Process.,
reflect the official policy or position of the Department of vol. 4. Beijing, China, Nov. 2006, pp. 1–4.
Defense or the U.S. Government. [23] Y. Xie, R. Tao, and T. Wang, “Method of waveform design for radar and
communication integrated system based on CSS,” in Proc. 1st Int. Conf.
Instrum. Meas. Comput. Commun. Control, Beijing, China, Oct. 2011,
R EFERENCES pp. 737–739.
[1] H. Griffiths et al., “Radar spectrum engineering and management: [24] M. Roberton and E. R. Brown, “Integrated radar and communications
Technical and regulatory issues,” Proc. IEEE, vol. 103, no. 1, based on chirped spread-spectrum techniques,” in Proc. IEEE MTT-S
pp. 85–102, Jan. 2015. Int. Microw. Symp. Dig., vol. 1. Philadelphia, PA, USA, Jun. 2003,
[2] D. W. Bliss, “Cooperative radar and communications signaling: The pp. 611–614.
estimation and information theory odd couple,” in Proc. IEEE Radar [25] A. Khawar, A. Abdel-Hadi, and T. C. Clancy, “MIMO radar waveform
Conf., Cincinnati, OH, USA, May 2014, pp. 50–55. design for coexistence with cellular systems,” in Proc. IEEE Int. Symp.
[3] B. Paul and D. W. Bliss, “Extending joint radar-communications bounds Dyn. Spectr. Access Netw. (DYSPAN), McLean, VA, USA, Apr. 2014,
for FMCW radar with Doppler estimation,” in Proc. IEEE Radar Conf., pp. 20–26.
Arlington County, VA, USA, May 2015, pp. 89–94. [26] Y. L. Sit and T. Zwick, “MIMO OFDM radar with communication
[4] A. R. Chiriyath, B. Paul, G. M. Jacyna, and D. W. Bliss, “Inner bounds and interference cancellation features,” in Proc. IEEE Radar Conf.,
on performance of radar and communications co-existence,” IEEE Trans. Cincinnati, OH, USA, May 2014, pp. 265–268.
Signal Process., vol. 64, no. 2, pp. 464–474, Jan. 2016. [27] B. Li, A. P. Petropulu, and W. Trappe, “Optimum co-design for spectrum
[5] A. R. Chiriyath and D. W. Bliss, “Effect of clutter on joint radar- sharing between matrix completion based MIMO radars and a MIMO
communications system performance inner bounds,” in Proc. 49th communication system,” IEEE Trans. Signal Process., vol. 64, no. 17,
Asilomar Conf. Signals Syst. Comput., Pacific Grove, CA, USA, pp. 4562–4575, Sep. 2016.
Nov. 2015, pp. 1379–1383. [28] A. Khawar, A. Abdel-Hadi, and T. C. Clancy, “Spectrum sharing
[6] A. R. Chiriyath, B. Paul, and D. W. Bliss, “Joint radar-communications between S-band radar and LTE cellular system: A spatial approach,” in
information bounds with clutter: The phase noise menace,” in Proc. Proc. IEEE Int. Sympo. Dyn. Spectr. Access Netw. (DYSPAN), McLean,
IEEE Radar Conf., Philadelphia, PA, USA, May 2016, pp. 690–695. VA, USA, Apr. 2014, pp. 7–14.
[7] A. R. Chiriyath and D. W. Bliss, “Joint radar-communications per- [29] S. Sodagari, A. Khawar, T. C. Clancy, and R. McGwier, “A projection
formance bounds: Data versus estimation information rates,” in Proc. based approach for radar and telecommunication systems coexistence,”
IEEE Mil. Commun. Conf. (MILCOM), Tampa, FL, USA, Oct. 2015, in Proc. IEEE Glob. Commun. Conf. (GLOBECOM), Anaheim, CA,
pp. 1491–1496. USA, Dec. 2012, pp. 5010–5014.
12 IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 1, MARCH 2017

[30] H. Shajaiah, A. Abdelhadi, and C. Clancy, “Spectrum sharing approach [53] M. R. Bell, “Information theory and radar waveform design,” IEEE
between radar and communication systems and its impact on radar’s Trans. Inf. Theory, vol. 39, no. 5, pp. 1578–1597, Sep. 1993.
detectable target parameters,” in Proc. IEEE 81st Veh. Technol. [54] J. R. Guerci, R. M. Guerci, A. Lackpour, and D. Moskowitz, “Joint
Conf. (VTC Spring), Glasgow, U.K., May 2015, pp. 1–6. design and operation of shared spectrum access for radar and commu-
[31] G. N. Saddik, R. S. Singh, and E. R. Brown, “Ultra-wideband nications,” in Proc. IEEE Radar Conf., Arlington County, VA, USA,
multifunctional communications/radar system,” IEEE Trans. Microw. May 2015, pp. 761–766.
Theory Techn., vol. 55, no. 7, pp. 1431–1437, Jul. 2007. [55] R. A. Romero and N. A. Goodman, “Cognitive radar network:
[32] R. Saruthirathanaworakun, J. M. Peha, and L. M. Correia, “Opportunistic Cooperative adaptive beamsteering for integrated search-and-track appli-
sharing between rotating radar and cellular,” IEEE J. Sel. Areas cation,” IEEE Trans. Aerosp. Electron. Syst., vol. 49, no. 2, pp. 915–931,
Commun., vol. 30, no. 10, pp. 1900–1910, Nov. 2012. Apr. 2013.
[33] F. Paisana, J. P. Miranda, N. Marchetti, and L. A. DaSilva, “Database- [56] T. M. Cover and J. A. Thomas, Elements of Information Theory, 2nd ed.
aided sensing for radar bands,” in Proc. IEEE Int. Symp. Dyn. Spectr. Hoboken, NJ, USA: Wiley, 2006.
Access Netw. (DYSPAN), McLean, VA, USA, Apr. 2014, pp. 1–6. [57] M. de Berg, O. Cheong, M. van Kreveld, and M. Overmars,
[34] A. Lackpour, M. Luddy, and J. Winters, “Overview of interference mit- Computational Geometry: Algorithms and Applications, 3rd ed.
igation techniques between WiMAX networks and ground based radar,” Heidelberg, Germany: Springer-Verlag, 2008.
in Proc. 20th Annu. Wireless Opt. Commun. Conf., Newark, NJ, USA, [58] D. W. Bliss and S. Govindasamy, Adaptive Wireless Communications:
Apr. 2011, pp. 1–5. MIMO Channels and Networks. New York, NY, USA: Cambridge Univ.
[35] H. Shajaiah, A. Khawar, A. Abdel-Hadi, and T. C. Clancy, “Resource Press, 2013.
allocation with carrier aggregation in LTE advanced cellular system shar- [59] H. L. Van Trees, Detection, Estimation, and Modulation Theory: Radar-
ing spectrum with S-band radar,” in Proc. IEEE Int. Symp. Dyn. Spectr. Sonar Signal Processing and Gaussian Signals in Noise. Malabar, FL,
Access Netw. (DYSPAN), McLean, VA, USA, Apr. 2014, pp. 34–37. USA: Krieger, Jul. 1992.
[36] J. R. Guerci and R. M. Guerci, “RAST: Radar as a subscriber technol-
ogy for wireless spectrum cohabitation,” in Proc. IEEE Radar Conf., Alex R. Chiriyath received the B.S.E.E. (Cum
Cincinnati, OH, USA, May 2014, pp. 1130–1134. Laude) degree in electrical engineering from the
[37] Y. Nijsure, Y. Chen, C. Yuen, and Y. H. Chew, “Location-aware spectrum University of Michigan at Ann Arbor in 2012,
and power allocation in joint cognitive communication-radar networks,” and the M.S. degree in electrical engineering from
in Proc. 6th Int. ICST Conf. Cogn. Radio Orient. Wireless Netw. Arizona State University in 2014, where he is
Commun. (CROWNCOM), Yokohama, Japan, Jun. 2011, pp. 171–175. currently pursuing the Ph.D. degree in electrical
[38] A. Aubry et al., “Cognitive radar waveform design for spectral coex- engineering.
istence in signal-dependent interference,” in Proc. IEEE Radar Conf.,
Cincinnati, OH, USA, May 2014, pp. 474–478.
[39] K.-W. Huang, M. Bică, U. Mitra, and V. Koivunen, “Radar waveform
design in spectrum sharing environment: Coexistence and cognition,”
in Proc. IEEE Radar Conf., Arlington County, VA, USA, May 2015,
pp. 1698–1703.
[40] J. Jakabosky, B. Ravenscroft, S. D. Blunt, and A. Martone, “Gapped Bryan Paul received the B.S. (Highest Hons.)
spectrum shaping for tandem-hopped radar/communications & cognitive degree in electrical engineering from the University
sensing,” in Proc. IEEE Radar Conf. (RadarConf), Philadelphia, PA, of Illinois at Urbana–Champaign in 2010, and the
USA, May 2016, pp. 1–6. M.S. degree in electrical engineering from Arizona
[41] M. Bica, K.-W. Huang, V. Koivunen, and U. Mitra, “Mutual information State University in 2014, where he is currently pur-
based radar waveform design for joint radar and cellular commu- suing the Ph.D. degree with the School of Electrical,
nication systems,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Computer and Energy Engineering.
Process. (ICASSP), Pudong, China, Mar. 2016, pp. 3671–3675. From 2002 to 2008, he was enlisted in the Illinois
[42] J. M. Chapin and W. H. Lehr, “Cognitive radios for dynamic spectrum Air National Guard completing multiple tours of
access—The path to market success for dynamic spectrum access tech- duty in support of Operation Enduring Freedom
nology,” IEEE Commun. Mag., vol. 45, no. 5, pp. 96–103, May 2007. and Operation Iraqi Freedom. From 2010 to 2012,
[43] W. Zhang, S. Vedantam, and U. Mitra, “Joint transmission and state he was with Validus Technologies, Peoria, IL, USA, working in the area
estimation: A constrained channel coding approach,” IEEE Trans. Inf. of embedded software and digital signal processing. He is currently with
Theory, vol. 57, no. 10, pp. 7084–7095, Oct. 2011. General Dynamics Mission Systems, Scottsdale, AZ, USA, working in the
[44] S. Howard, W. Moran, A. R. Calderbank, H. Schmitt, and C. Savage, area of digital signal processing for satellite payload systems. His current
“Channel parameters estimation for cognitive radar systems,” in Proc. research interests include signal processing, information theory, radar, and
IEEE Int. Conf. Acoust. Speech Signal Process., vol. 5. Philadelphia, communications.
PA, USA, Mar. 2005, pp. 897–900. He was a recipient of numerous military decorations, including the
[45] A. Pezeshki, A. R. Calderbank, W. Moran, and S. D. Howard, “Doppler Humanitarian Service Medal. He is a co-inventor of one awarded U.S. patent.
resilient Golay complementary waveforms,” IEEE Trans. Inf. Theory,
vol. 54, no. 9, pp. 4254–4266, Sep. 2008. Daniel W. Bliss (F’15) received the B.S.E.E.
[46] B. Li and A. Petropulu, “MIMO radar and communication spec- degree in electrical engineering from Arizona State
trum sharing with clutter mitigation,” in Proc. IEEE Radar University in 1989, and the M.S. degree in physics
Conf. (RadarConf), Philadelphia, PA, USA, May 2016, pp. 1–6. and the Ph.D. degree from the University of
[47] A. Hassanien, M. G. Amin, Y. D. Zhang, F. Ahmad, and B. Himed, California at San Diego in 1995 and 1997, respec-
“Non-coherent PSK-based dual-function radar-communication systems,” tively. He was employed by General Dynamics from
in Proc. IEEE Radar Conf. (RadarConf), Philadelphia, PA, USA, 1989 to 1993, where he designed rocket avionics
May 2016, pp. 1–6. and performed magnetic field calculations and opti-
[48] A. Hassanien, M. G. Amin, Y. D. Zhang, and F. Ahmad, “Dual-function mization for high-energy particle-accelerator super-
radar-communications: Information embedding using sidelobe control conducting magnets. His doctoral work (1993–1997)
and waveform diversity,” IEEE Trans. Signal Process., vol. 64, no. 8, was in the area of high-energy particle physics. He
pp. 2168–2181, Apr. 2016. was a senior member of the technical staff with MIT Lincoln Laboratory
[49] E. BouDaher, A. Hassanien, E. Aboutanios, and M. G. Amin, “Towards from 1997 to 2012. He is currently an Associate Professor with the School
a dual-function MIMO radar-communication system,” in Proc. IEEE of Electrical, Computer and Energy Engineering, Arizona State University.
Radar Conf. (RadarConf), Philadelphia, PA, USA, May 2016, pp. 1–6. His current research topics include multiple-input multiple-output (MIMO)
[50] D. Gaglione et al., “Fractional Fourier based waveform for a joint wireless communications, MIMO radar, cognitive radios, radio network
radar-communication system,” in Proc. IEEE Radar Conf. (RadarConf), performance, geolocation, and statistical signal processing for anticipatory
Philadelphia, PA, USA, May 2016, pp. 1–6. physiological analytics. He has been the Principal Investigator on numerous
[51] C. Shannon, “A mathematical theory of communication,” Bell Syst. Tech. programs with applications to radio, radar, and medical monitoring. He has
J., vol. 27, no. 3, pp. 379–423 and pp. 623–656, Jul./Oct. 1948. made significant contributions to robust multiple-antenna communications,
[52] P. M. Woodward, “Information theory and the design of radar receivers,” including theory, patents, and the development of advanced prototypes. He
Proc. IRE, vol. 39, no. 12, pp. 1521–1524, Dec. 1951. is responsible for some of the foundational MIMO radar literature.