Detection Performance of Distributed MIMO Radar

with Asynchronous Propagation

and Timing/Phase Errors
Fangzhou Wang, Cengcang Zeng, and Hongbin Li Mark A. Govoni
ECE Department, Stevens Institute of Technology Army Research Laboratory
Hoboken, NJ 07030, USA Aberdeen Proving Ground, MD 21005, USA

Abstract—Distributed multi-input multi-output (MIMO) radar which allows one to exploit spatial or geometric diversity for
has received significant interest in recent years. However, most target detection and recognition.
prior works assume orthogonal transmit waveforms, which can Most existing studies on MIMO radar assume orthogo-
be perfectly separated at the receiver using a set of waveform-
specific matched filters. With widely separated antennas, the nal waveforms, in which case the receiver can separate the
waveforms often arrive at the receiver at different times, waveforms using a set of filters matched to each individual
losing their orthogonality. The problem becomes worse when waveform to avoid mutual interference. However, such ideal
timing/phase errors are present. In this paper, we examine waveform separation is impossible across all Doppler and time
target detection in distributed MIMO radar with synchronization delay resolutions [10]–[12]. In addition, phase synchronization
errors. A general signal model is developed that takes into
account asynchronous propagation and possible timing/phase provides a coherent processing gain beneficial to direction
synchronization errors in distributed MIMO radar. Two sets finding, moving target indication, etc. [13]. Time/phase syn-
of linear frequency modulation (LFM) based waveforms with chronization is a non-trivial task for MIMO radar systems.
different characteristics, along with coherent and non-coherent Although in principle the transmit/receive antennas can be
detectors, are employed to examine the impact of asynchronous synchronized to a common time/frequency source, timing and
propagation and synchronization errors on target detection.
Simulation results are provided to compare the target detection phase errors may occur due to the imperfect knowledge of the
performance of these detectors under different scenarios. Our antenna locations and local oscillator/electronics characteris-
results show that while the coherent detector is in general tics of the antennas. A few prior studies considered the effect
sensitive to synchronization errors, the non-coherent detector of phase errors on MIMO radar working under some specific
may benefit from the cross interference among the transmit conditions [14]–[17].
waveforms, which makes it less sensitive to synchronization
errors, in particular for highly correlated waveforms. In this paper, we develop a general signal model that
Index Terms—distributed MIMO radar, asynchronous propa- considers synchronization errors for a distributed MIMO radar.
gation, sync errors, coherent and non-coherent detection To study the impacts of synchronization errors on the target
detection performance of the MIMO radar, two sets of linear
I. I NTRODUCTION frequency modulation (LFM) based waveforms with different
Multi-input multi-output (MIMO) radar, which is equipped characteristics suitable for MIMO radar are discussed and
with multiple transmit and receive antennas along with mul- the cross ambiguity function of these waveforms are studied.
tiple probing waveforms, has received significant interest in Two MIMO detectors with both non-coherent and coherent
recent years (e.g., [1]–[9]). Compared with the conventional processing are employed for testing under cases with and
phased-array radar using one transmit aperture with a single without synchronization errors. The detection performance of
probing waveform, MIMO radars offer a number of unique these detectors is numerically evaluated and compared under
benefits including higher spatial resolution, more degrees of several scenarios.
freedom (DOFs), enhanced target detection and identifiability,
as well as better spatial coverage. MIMO radar has two stan- II. S IGNAL M ODEL WITH A SYNCHRONOUS P ROPAGATION
dard array configurations, one involving co-located antennas AND S YNCHRONIZATION E RRORS
[1]–[5] and the other using distributed antennas [6]–[9], with Consider a distributed MIMO radar geometry with M
distinct performance characteristics. In this paper, we focus widely separated transmit antennas (Tx’s) and N widely
on distributed MIMO radar with widely separated antennas, separated receive antennas (Rx’s). The transmit and receive
antennas are assumed to be on stationary platforms and
Research was sponsored by the Army Research Office and was accom-
plished under Cooperative Agreement Number W911NF-19-2-0234. The their locations are assumed known. Pulsed transmission is
views and conclusions contained in this document are those of the authors and employed by the M transmit antennas to probe a common
should not be interpreted as representing the official policies, either expressed area of interest using M waveforms. Each transmitter sends
or implied, of the Army Research Office or the U.S. Government. The U.S.
Government is authorized to reproduce and distribute reprints for Government a succession of K periodic pulses, i.e., K repetitions of the
purposes notwithstanding any copyright notation herein. same waveform, over a coherent processing interval (CPI).

978-1-7281-6813-5/20/$31.00 ©2020 IEEE 13

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.
Specifically, at the m-th transmitter, the transmitted burst of of the propagation delay τni . Moreover, let Λni  fni − fni
pulses are given by s̃m (t) = bm um (t)ej(2πfc t+ψm ) , where denote the Doppler mismatch between the actual Doppler shift
K−1
um (t) = k=0 pm (t − kTr ) is the baseband transmitted and the compensated Doppler shift of the i-th matched filter
signal, pm (t) is the complex envelope of a single pulse at the n-th receiver. Then, the samples can be written as
for antenna m, bm is the transmit amplitude at the m-th 

transmitter, fc is the carrier frequency, and ψm is the initial xni (k) = xni (t)
t=τni +Δni +kTr
phase. The pulse waveform pm (t) is of the same duration
jψi Λni −j2πfc τni j2πkfni j2π fniT−Λni Δni
Tp for all transmitters and has unit energy. Therefore, |bm |2 = bi hni e χii (Δni ,
)e e e r

denotes the energy transmitted in a single pulse. Tr

 Λni
Suppose there is a moving target at a distance RT,m to the + bm hnm ejψm χmi (τni + Δni − τnm , )
m-th transmitter and a distance RR,n to the n-th receiver. The Tr
m=i
baseband signal sn (t) observed at the n-th receiver consists fni −Λni
× e−j2πfc τnm ej2πkfnm ej2π Tr (τni +Δni −τnm )
(4)
of echoes from the target illuminated by M waveforms
M i = 1, 2, . . . , M ; n = 1, 2, . . . , N ; k = 0, 1, . . . , K − 1.

sn (t) = bm hnm um (t − τnm )
m=1
Note that the output sample xni (t) consists of one auto
jψm −j2πfc τnm −j2πFnm τnm j2πFnm t term between the i-th transmit waveform and the i-th matched
×e e e e . (1)
filter, and M − 1 cross terms between the other M − 1
where hnm is a complex parameter that lumps together the transmit waveforms and the i-th matched filter. The cross
target amplitude and the channel coefficient associated with terms vanish when waveforms pm (t), m = 1, 2, . . . , M , are
the (m, n)-th transmit-receive (Tx-Rx) pair, τnm = (RT,m + orthogonal to each other, a standard assumption used in the
RR,n )/c is the (m, n)-th Tx-Rx delay, and Fnm is the bistatic MIMO literature. In practice, maintaining strict orthogonality
target Doppler frequency as observed by the n-th receiver in across time and frequency is a challenging problem. With non-
response to the radar waveform transmitted from the m-th orthogonal waveforms or waveforms that are orthogonal only
transmitter. with zero delay/Doppler, cross terms are present as residual
At the n-th receiver, a set of M matched filters, each interferences, which may become non-negligible and need to
matched to one of M transmit waveforms, are used to extract be accounted for with increasing synchronization errors.
the baseband signal corresponding to the (m, n)-th Tx-Rx Next, we stack the K discrete samples of (4) into a vector
pair. Specifically, each pulse of sn (t) is matched separately xni = [xni (0), xni (1), · · · , xni (K − 1)]T to formulate a noisy

with gi (t) = p∗i (−t)ej2πFni t , i = 1, 2, . . . , M , where Fni discrete observation signal yni as
denotes the Doppler compensation applied by the matched
filter. In general, Fni is the estimated Doppler frequency. yni = xni + wni , (5)
The output at the i-th matched filter of the n-th receiver is
xni (t) = sn (μ)gi (t−μ)dμ. Let us define the cross ambiguity where wni is the noise component for the (n, i)-th Tx-
function (CAF) as Rx pair, which is assumed to have a complex Gaussian
 2
distribution with zero mean and covariance matrix σni I, i.e.,
χmi (t, f ) = pm (μ)p∗i (μ − t)ej2πf μ dμ. (2) wni ∼ CN (0, σni 2
I).
In this paper, the problem of interest is to study the
Following the above discussion, the output of the i-th matched
target detection performance of distributed MIMO radar with
filter at the n-th receiver xni (t) can be rewritten as
asynchronous propagation and synchronization errors.
M
 
xni (t) = bm hnm ejψm e−j2πfc τnm e−j2πFnm τnm ej2πFni t
III. MIMO R ADAR WAVEFORMS AND D ETECTORS
m=1

K−1
 In this section, we first introduce two sets of linear frequen-
× pm (μ − kTr − τnm )p∗i (μ − t)ej2π(Fnm −Fni )μ dμ cy modulation (LFM) based waveforms that can be employed
k=0
for MIMO radar, and examine the cross ambiguity function
M
fni fni (CAF) of these waveforms. We then discuss two MIMO detec-
= bm hnm ejψm e−j2πfc τnm ej2π Tr t e−j2π Tr τnm
tors with non-coherent and, respectively, coherent processing,
m=1
K−1 which are used for testing in this study.
 fnm − fni j2πk(fnm −fni )
× χmi (t − τnm − kTr , )e , (3)
Tr A. LFM Based Waveforms
k=0

where fnm  Fnm Tr and fni  Fni Tr denote the normalized The LFM waveforms, often called chirp waveforms, are
Doppler frequencies. frequently used in radar. In this subsection, we introduce two
The continuous-time signal xni (t) is sampled at time in- sets of chirp waveforms and their ambiguity functions. Each
stants t = τni +Δni +kTr , k = 0, 1, · · · , K−1, where Δni de- set contains multiple waveforms that are suitable for MIMO
notes the sampling time error due to the inaccurate knowledge radar.

14

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.
 PXOWLEDQGFKLUS VLQJOHEDQGFKLUS PXOWLEDQGFKLUS]HUR'RSSOHU&$)
   

  

  

   

  

  

QRUPDOL]HGIUHTXHQF\I

QRUPDOL]HGIUHTXHQF\I

  

 

DPSOLWXGH



 

 
      
     
QRUPDOL]HGGHOD\

QRUPDOL]HGWLPHW
QRUPDOL]HGWLPHW

(a)
Fig. 1. The instantaneous frequency of the chirp waveforms. Left: multi-band VLQJOHEDQGFKLUS]HUR'RSSOHU&$)
chirp (M = 2 and η = 3); right: single-band chirp (L = 1 and κ = 3).  

1) Multi-Band Chirp Waveforms: Consider a set of multi-  

DPSOLWXGH
band chirp waveforms given by
1 2
pm (t) = √ ejπβ(t /τ +ηmt) (6)  
τ XSFKLUS XSFKLUS
0 ≤ t ≤ τ, m = 1, 2, . . . , M, GRZQFKLUS GRZQFKLUS
XSFKLUS GRZQFKLUS
GRZQFKLUS XSFKLUS
where β is the waveform bandwidth of each chirp waveform, τ  
    
is the pulse duration, and η is a bandwidth gap parameter that
QRUPDOL]HGGHOD\

is selected to keep the frequency bands of different waveforms
non-overlapping. As one example, Fig. 1 (left plot) depicts the (b)
instantaneous frequency for 2-band chirps with M = 2 and Fig. 2. Zero-Doppler cut of the auto and cross ambiguity functions. (a)
η = 3. Multi-band chirp (η = 3); (b) single-band chirp (κ = 3).
Multi-band chirps are orthogonal waveforms under certain
conditions, e.g., when the time-bandwidth product βτ is an
2) Single-Band Chirp Waveforms: For comparison purpose,
integer and the waveforms are synchronous, which can be
we also consider a set of single-band chirp waveforms with
determined from the ambiguity function. The CAF of the
overlapping instantaneous frequency:
multi-band chirps can be obtained from (2). Specifically, for ⎧ 1 jπβg (t)
t ≥ 0, ⎪
⎨ √τ e
1
, 0 ≤ t < −1L τ,
 τ
1 jπβ( μ2 +ηmμ) p (t) = (9)
χ̂mi (t, fd ) = e τ
⎪
⎩ √1 jπβg2 (t) −1
t τ e , τ ≤ t ≤ τ,
 (μ−t)2 τ L
× e−jπβ τ +ηi(μ−t) j2πfd μ
e dμ  2
 τ where  = 1, 2, . . . , L,, g1 (t) = (−1)γ tτ + L− +1
L 2t +2γt+
j2πρmi τ
1 j2πρmi μ e − ej2πρmi t (t− −1 τ )2
=φ e dμ = φ κt, and g2 (t) = (−1)γ L
τ + 2γt + κt. In addition, κ
t τ j2πρmi τ
is a constant that controls the center frequency of the chirp
jπρmi (τ −t) −jπρmi (τ −t)
e −e waveform and γ = 0 or 1. Specifically, when γ = 0 the pulse
= φejπρmi (t+τ ) , (7)
j2πρmi τ is an up chirp whose instantaneous frequency has a positive
t2 slope on the time-frequency plane; if γ = 1, it is a down chirp
where ρmi = (m−i)ηβτ +2βt
+ fd and φ = e−jπβ( τ +ηit) . A
2τ with a negative slope. Eq. (9) defines a total of 2L waveforms
similar derivation can be carried out for the case t < 0. The
in the band, i.e., L up-chirp waveforms and L down-chirp
ambiguity function is the magnitude of χ̂mi (t, fd ):
⎧  waveforms. As indicated by (9), the instantaneous frequency
sin[πρmi (τ +t)]  of all the single-band chirp waveforms sweeps between κβ
⎪
⎪   , −τ ≤ t < 0, 2 Hz
⎨ πρ τ
and (β + κβ
mi
) Hz. When L = 1, (9) reduces to a simple up
χmi (t, fd ) = |χ̂mi (t, fd )| =  
2
⎪
⎪ chirp with γ = 0:
⎩ sin[πρmi (τ −t)]  , 0 ≤ t ≤ τ.
πρmi τ 1 2
(8) pu (t) = √ ej(πβt /τ +κπβt) , 0 ≤ t < τ, (10)
τ
Note that when m = i, χmi (t, fd ) represents the auto-
ambiguity function (AAF) of the chirp waveform while m = i and a simple down chirp with γ = 1:
gives the CAF of the two chirp waveforms. The zero-Doppler 1 2
pd (t) = √ ej(−πβt /τ +2πβt+κπβt) , 0 ≤ t ≤ τ. (11)
cut of the AAF and CAF is shown in Fig. 2(a). τ

15

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.
The instantaneous frequency of the simple up- and down-chirp IV. N UMERICAL S IMULATIONS
waveforms is shown in Fig. 1 (right plot) with κ = 3. In this section, numerical simulations are provided to
The CAF of those two waveforms is given by demonstrate the performance of the coherent and non-coherent
⎧  τ +t j2π(f + μ−t−τ β)μ 
⎪ g(t) dμ , −τ ≤ t < 0 detectors by using the multi-band and single-band chirp
⎪
⎨ 0
e d τ
waveforms. We consider a stationary target and the effect
χud (t, fd ) =   of Doppler is neglected for simplicity, i.e., both the Doppler
⎪
⎩g(t)  τ ej2π(fd + μ−t−τ
⎪
τ β)μ
dμ

, 0 ≤ t ≤ τ, frequency fni and errors Λni are set to zero. For comparison
t
(12) purpose, we consider the following three scenarios.
2
1 j( πβt +2πβt+κπβt)
where g(t) = τ e τ . The AAF of the up-chirp • Clairvoyant: The first scenario is an ideal case that the
waveform can be expressed as in (8) with m = i. The zero- output sample of the matched filter contains only the
Doppler cut ambiguity function of the simple up- and down- auto term in (4) with sampling error Δni = 0. Since the
chirp signal is shown in Fig. 2(b). cross terms in (4) are absent, the clairvoyant waveforms
are assumed to be orthogonal and the orthogonality is
B. Detectors maintained for different delays and Dopplers.
Given the radar observations (5) in the presence of a target, • Asynchronous with no sync errors: This reflects a more
the detection problem is to select between the following two practical case where the waveforms arrive at a receiver
hypotheses: with different propagation delays and Δni = 0 (no timing
errors). In this case, the matched filter output contains
H0 : yni = wni ,
both auto and cross terms in (4) with Δni = 0 and Λni =
H1 : yni = xni + wni , (13) 0. In other words, the matched filter is synchronized with
i = 1, 2, · · · , M, n = 1, 2, · · · , N. the desired waveform but not with the other waveforms.
As such, the target echo is sampled at the peak location
Next, we discuss two detectors for the above hypothesis testing of the auto term but also contains the cross terms due to
problem. mutual interference among different waveforms.
1) Non-Coherent Detector: The nature of non-coherent • Asynchronous with sync errors: This represents the
integration is the combining of the energy of the received most practical case when synchronization error Δni = 0
signals from different Tx-Rx pairs. This leads to an energy and cannot be neglected. Note that the presence of timing
detector for target detection [6]: errors also introduces phase errors because they are
N 
 M coupled [see, e.g., (16)].
H1
H
TNCD  yni yni ≷ γNCD , (14) The simulation scenarios involve two transmitters (M =
n=1 i=1 H0
2) and one receiver (N = 1) unless otherwise stated. In the
where γNCD is a threshold set for a tolerated level of false simulation, the transmitted amplitude is bm = 1, the carrier
alarm. frequency is fc = 3 GHz, and the waveform bandwidth is β =
2) Coherent Detector: Coherent processing in spatial and 1 MHz. The target is moving with relative location determined
Doppler domains promises a higher resolution and perfor- by τ11 = 0.5τ and τ12 = 0.1τ , where τ = 10−5 s denotes
mance gains in target detection and estimation than non- the pulse duration. The normalized Doppler frequencies are
coherent. This requires additional phase synchronization f11 = 0.4 and f12 = 0.38. The number of pulse within a
across all sensors, i.e., finding accurate estimates of the coherent processing interval (CPI) is K = 12. For the multi-
initial phase ψm , pair-wise Tx-Rx delay τmn , and pair-wise band chirp waveform, η = 3. For the single-band waveforms
Doppler frequency fmn , for all n and m. Let us define given by (10) and (11), we have κ = 3. The probability of false
θnik = ψi −2πfc τni +2πkfni . Under perfect condition, θnik is alarm is 10−4 . The signal-to-noise ratio (SNR) is defined as
known and can be used to perform phase compensation for the
|bm |2 E{|hnm |2 }
coherent detector [18]. On the other hand, with timing error SNR = SNRnm = 2
, (17)
Δni and Doppler estimation error Λni , the phase compensation σnm
is based on the estimates of the delay and Doppler, i.e., where the noise variance is chosen as σ 2 = 1 and hnm ∼
τni + Δni and fni + Λni . Therefore, the coherent detector CN (0, σh2 ) is randomly generated from trial to trial for each
can be implemented as SNR. σh2 is computed according to a given SNR value.
 2 In Fig. 3, the probability of detection is presented as a
 N  M K−1
  H1
 −jθ̂nik 
TCD   e yni (k) ≷ γCD (15) function of SNR. It is observed that the performance of
  H0 the multi-band chirp signals for both detectors under the
n=1 i=1 k=0

where asynchronous case with Δni = 0 is very close to that of

the clairvoyant result while there is a performance gap for
θ̂nik = θnik + 2πkΛni − 2πfc Δni , (16)
the single-band chirp signals. This is because, as shown in
yni (k) denotes the k-th element of yni , and γCD is set to Fig. 2, the CAF of the multi-band chirp is relatively small,
ensure the required probability of false alarm. whereas the CAF for the single-band chirp is considerably

16

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.
 QRQFRKHUHQWGHWHFWRU QRQFRKHUHQWGHWHFWRU
 

 

SUREDELOLW\RIGHWHFWLRQ

SUREDELOLW\RIGHWHFWLRQ
 

 


 

 
 
 
 

      
 

  

 
 

  
   
 

  


 
 

  
   
 

  

 
         
615G% 615G%

(a) (b)
FRKHUHQWGHWHFWRU FRKHUHQWGHWHFWRU
 

 

SUREDELOLW\RIGHWHFWLRQ
SUREDELOLW\RIGHWHFWLRQ

 

 


 

 
 
 
 

      
 

  

 
 

  
   
 

  


 
 

  
   
 

  

 
         
615G% 615G%

(c) (d)
Fig. 3. Probability of detection versus SNR for different detectors and waveforms with various timing errors. (a) Non-coherent detector with multi-band chirp
signals; (b) non-coherent detector with single-band chirp signals; (c) coherent detector with multi-band chirp signals; (d) coherent detector with single-band
chirp signals.

higher and the performance loss is caused by the mutual a smaller averaged performance loss. On the other hand, the
interference among the transmitted waveforms, which becomes multi-band chirp cannot benefit from the cross term, which is
non-negligible. For the asynchronous case with timing errors, too small compared with the auto term at Δni = 0.1τ .
we consider two different synchronous errors Δni = 0.05τ To offer further insight into the effects of the synchronous
and Δni = 0.1τ , respectively. It is seen from all four plots in error, we test the performance of both the coherent and non-
Fig. 3 that in general, as the synchronization error increases, coherent detectors for the asynchronous case by using multi-
the detection performance of the non-coherent and coherent band chirp waveform with no timing error and, respectively,
detectors degrades. This is expected since a larger sampling a small timing error Δni = 10−4 τ . Fig. 4 shows that the non-
error implies the sampling location is further away from the coherent detector in both scenarios has the same performance
peak of the AAF (see Fig. 2), which results in a higher loss of due to negligible amplitude drop of the AAF value when the
the energy of the desired auto term and the associated SNR. timing error is small [cf. Fig. 2(a)]. In contrast, the coherent
For the non-coherent detector, a closer examination of detector experiences a notable degradation even with such a
Figs. 3(a) and 3(b) reveals that the performance loss of the small timing error. This is because a small timing error can still
single-band chirp at Δni = 0.1τ versus Δni = 0.05τ is significantly affect the phase of the AAF, which can introduce
considerably smaller than that of the multi-band chirp. This additional phase mismatch for the coherent detector.
is caused by the fact that at Δni = 0.1τ , the cross term Finally, we examine the effects of the number of transmit-
of the single-band chirp contains significant energy (Fig. 2(b) ters. Fig. 5 depicts the probability of detection for the coherent
shows the cross term and auto term have similar energy at and non-coherent detectors using the multi-band chirp signal
Δni = 0.1τ ). The auto and cross terms may add up construc- with various numbers of transmitters for the asynchronous sce-
tively when the difference of their phase angles is between nario with no timing error. It can be seen that the performance
−π/2 to π/2, or destructively when otherwise. Since the of both detectors improves when the number of transmitters
detection results are averaged over many channel realizations increases due to the additional power gain. Specifically, there
(with varying hnm ), there is about 50 percent chance that the is a 3 dB and 6 dB performance improvement with M = 2
auto and cross terms would add up constructively, in which and M = 4 Tx’s, respectively, over the case with M = 1 Tx,
case the SNR would benefit from the cross term, leading to for the coherent detector. On the other hand, the benefit of

17

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.
 PXOWLEDQGFKLUS
 is shown that the non-coherent detector may benefit from cross

 





 





 interference among the transmit waveforms, which makes
 
 




it less sensitive to synchronization errors than the coherent
SUREDELOLW\RIGHWHFWLRQ

 






detector, in particular for highly correlated waveforms.

R EFERENCES
 [1] J. Li and P. Stoica, “MIMO radar with colocated antennas,” IEEE Signal
Processing Magazine, vol. 24, no. 5, pp. 106–114, Sep. 2007.
 [2] H. Li and B. Himed, “Transmit subaperturing for MIMO radars with co-
located antennas,” IEEE Journal of Selected Topics in Signal Processing,
vol. 4, no. 1, pp. 55–65, Feb 2010.
 [3] G. Cui, H. Li, and M. Rangaswamy, “MIMO radar waveform design
      
with constant modulus and similarity constraints,” IEEE Transactions
615G%
on Signal Processing, vol. 62, no. 2, pp. 343–353, Jan 2014.
[4] M. A. Govoni, R. Elwell, T. Dogaru, and D. Liao, “A practical look
Fig. 4. Probability of detection versus SNR for coherent and non-coherent
at target detection using MIMO radar,” in SPIE Defense, Security, and
detectors with no or a small timing error (Δni = 10−4 τ ).
Sensing Conference, vol. 9461, May 2015, pp. 228–238.
[5] P. Chen, L. Zheng, X. Wang, H. Li, and L. Wu, “Moving target detection
FRKHUHQWGHWHFWRU using colocated MIMO radar on multiple distributed moving platforms,”

IEEE Transactions on Signal Processing, vol. 65, no. 17, pp. 4670–4683,
Sep. 2017.
 [6] A. M. Haimovich, R. S. Blum, and L. J. Cimini, “MIMO radar with
SUREDELOLW\RIGHWHFWLRQ

widely separated antennas,” IEEE Signal Processing Magazine, vol. 25,


no. 1, pp. 116–129, 2008.
[7] P. Wang, H. Li, and B. Himed, “Moving target detection using dis-
tributed MIMO radar in clutter with nonhomogeneous power,” IEEE
 Transactions on Signal Processing, vol. 59, no. 10, pp. 4809–4820, Oct
2011.


 [8] P. Wang, H. Li, and B. Himed, “A parametric moving target detector





 for distributed MIMO radar in nonhomogeneous environment,” IEEE


 Transactions on Signal Processing, vol. 61, no. 9, pp. 2282–2294, May
 2013.
      
[9] H. Li, Z. Wang, J. Liu, and B. Himed, “Moving target detection in
615G% distributed MIMO radar on moving platforms,” IEEE Journal of Selected
(a) Topics in Signal Processing, vol. 9, no. 8, pp. 1524–1535, Dec 2015.
QRQFRKHUHQWGHWHFWRU
[10] Y. I. Abramovich and G. J. Frazer, “Bounds on the volume and height
 distributions for the MIMO radar ambiguity function,” IEEE Signal
Processing Letters, vol. 15, pp. 505–508, 2008.
[11] M. Akcakaya and A. Nehorai, “MIMO radar sensitivity analysis for

SUREDELOLW\RIGHWHFWLRQ

target detection,” IEEE Transactions on Signal Processing, vol. 59, no. 7,

pp. 3241–3250, July 2011.
 [12] P. Wang, H. Li, and B. Himed, “Moving target detection for distributed
MIMO radar with imperfect waveform separation,” in 2013 IEEE Radar
Conference (RadarCon13), April 2013, pp. 1–5.

[13] N. H. Lehmann, A. M. Haimovich, R. S. Blum, and L. Cimini, “High


 resolution capabilities of MIMO radar,” in 2006 Fortieth Asilomar


 Conference on Signals, Systems and Computers, Oct 2006, pp. 25–30.


 [14] Q. He and R. S. Blum, “Cramer-Rao bound for MIMO radar target



localization with phase errors,” IEEE Signal Processing Letters, vol. 17,

       no. 1, pp. 83–86, Jan 2010.
615G% [15] H. Godrich and A. M. Haimovich, “Localization performance of co-
herent MIMO radar systems subject to phase synchronization errors,”
(b) in 2010 4th International Symposium on Communications, Control and
Fig. 5. Probability of detection versus SNR for different detectors with Signal Processing (ISCCSP), March 2010, pp. 1–5.
various numbers of transmitters. (a) Coherent detector; (b) non-coherent [16] M. Akcakaya and A. Nehorai, “MIMO radar detection and adaptive
detector. design under a phase synchronization mismatch,” IEEE Transactions on
Signal Processing, vol. 58, no. 10, pp. 4994–5005, Oct 2010.
[17] Y. Yang and R. S. Blum, “Phase synchronization for coherent MIMO
radar: Algorithms and their analysis,” IEEE Transactions on Signal
multiple transmitters to the non-coherent detector is smaller. Processing, vol. 59, no. 11, pp. 5538–5557, Nov 2011.
[18] J. A. Nanzer, R. L. Schmid, T. M. Comberiate, and J. E. Hodkin, “Open-
loop coherent distributed arrays,” IEEE Transactions on Microwave
V. C ONCLUSION Theory and Techniques, vol. 65, no. 5, pp. 1662–1672, May 2017.
In this paper, a performance analysis for target detection in
distributed MIMO radar subject to asynchronous propagation
and timing/phase synchronization errors is provided. Two sets
of chirp waveforms, together with a coherent and non-coherent
detector, are discussed and employed to evaluate the impact
of imperfect waveform separation caused by asynchronous
transmission, which is inherent in distributed MIMO radar. It

18

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:56:18 UTC from IEEE Xplore. Restrictions apply.