Professional Documents
Culture Documents
Multitarget Parameter Estimation of Monopulse Radar Based On RJ-MCMC Algorithm
Multitarget Parameter Estimation of Monopulse Radar Based On RJ-MCMC Algorithm
Research Article
Multitarget Parameter Estimation of Monopulse Radar Based on
RJ-MCMC Algorithm
Jian Gong ,1,2,3 Yiduo Guo,2 Hui Yuan,2 and Qun Wan1
1
University of Electronic Science and Technology of China, Chengdu, Sichuan, 611731, China
2
Air Force Engineering University, Xi’an, Shaanxi, 710051, China
3
TongFang Electronic Science and Technology Co. Ltd., Jiujiang Jiangxi, 332007, China
Received 29 June 2018; Revised 13 November 2018; Accepted 21 November 2018; Published 6 January 2019
Copyright © 2019 Jian Gong et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
A multiple parameter estimation method based on RJ-MCMC for multiple nondiscernible targets is proposed in this paper.
Different from the traditional estimation methods, the proposed method can simultaneously complete the joint estimation of the
target number and the target location parameters. More importantly, the method proposed in this chapter is applicable to many
situations with different power and nondistinguishable target. The simulation results show that the method proposed in this chapter
requires less observation time to obtain similar and even better estimation performance than the ML-MDL method, which is of
great significance for real-time processing.
1. Introduction same distance and angle resolution units are still possible
to cause interference to the radar angle measurement and
In the process of fighter penetration, as shown in Figure 1, tracking system. Serious deviation will occur when the angle
a two-tier echelon and a team’s wedge formation are usually of the target is estimated by using the traditional single pulse
used [1]. processing.
It can be seen that two or more than two targets are likely From the spatial domain, the composite jamming
to exist in the same distance-angle resolution unit of the radar described above can deceive monopulse radar from a mul-
when the target is launched with a double “Echelon” or a tisource coherent/incoherent angle. The solution to this
team ’s “wedge” formation. With the influence of side lobe problem from the beam is the angle superresolution method
jamming and false target jamming [2–5], radar’s display on based on the improved monopulse processing method, and it
the two-dimensional display of distance angle is shown in mainly includes complex monopulse processing method [6–
Figure 2. The red pentagram represents the target that radar 10], parameter estimation method based on joint bin process-
is detecting or tracking, and the dot represents other targets ing (JBP) [11, 12], and maximum likelihood (ML) estimation.
or false targets. The complex monopulse processing method assumes that the
From Figure 2, we can see that, during the penetration target just falls on the sampling point of matched filter and
process, many false targets may be observed in the range only uses one matched filter to sample information, so it
angle dimension of radar. The black dots represent the false can only solve the problem of angle estimation of up to two
targets in the sidelobe region, which can be suppressed by indistinguishable targets. The ML estimation algorithm based
the side lobe countermeasures. The blue dots represent the on JBP model is computationally intensive and easy to fall
false targets in the different range resolution units in the into local extremum. When the number of samples is small
radar main beam resolution unit and can be identified or and the SNR is low, the minimum description length (MDL)
suppressed by the method of distance dimension false target criterion is used to estimate the number of targets in the
interference. However, the false targets (green dots) in the beam and it is very easy to fail. The standard MCMC method
2 Mathematical Problems in Engineering
3 4
50~100m
50~100m
3
50~100m
100m 100m 1
50~100m
±50m ±50m
2
(a) Two-shipper echelon (b) “Wedge” team vertical view (c) “Wedge” team flat view
Azimuth/Pitch
cannot “jump” between subspaces of different dimensions, proposed method in Section 5. Finally, we conclude the paper
so it does not have the ability to estimate the number of in Section 6.
targets. A direct solution is similar to the method proposed
by [11, 12]. In order to get the target parameters estimation,
2. Joint Bin Processing Model
the order MCMC sampling algorithm is run independently
and the number of targets is fixed at each time. Then the Before giving the model, we first make the following assump-
number of targets is judged by MDL criterion. However, there tions:
is no substantial improvement in the amount of computation (I) Assuming that the target model is a point target model,
and estimation accuracy. Therefore, in this paper, we use RJ- the extended target is not considered.
MCMC method to jointly estimate target number and target
(II) When the noise is not considered, the output of
position parameters. RJ-MCMC method can “jump” from
matched filter is triangular waveform; that is, the radar
different dimension space when the dimension of parameter
transmits the signal with rectangular envelope.
space changes; that is, it can directly sample according to the
joint distribution of A and finally get the estimation results of (III) The sampling interval after matched filtering is pulse
target number and target position parameters. width T, which means that a target can only appear in its
The remainder of this paper is organized as follows. In adjacent matched filtering sampling points.
Section 2, the joint bin processing (JBP) model is established. Figure 3 shows the “leakage” model of two targets at two
Then, the prior and posterior distributions of Bayesian adjacent sampling points. Δ𝑇 represents the time offset of
parameter estimation are derived in Section 3, followed by the real target position relative to the sampling point 1. 𝑥 =
a multitarget parameter estimation algorithm for monopulse 𝑎 cos(𝜙) (or 𝑥 = 𝑎 sin(𝜙)) is the peak amplitude of the target
radar Based on RJ-MCMC is proposed in Section 4. Some echo in the same phase (or quadrature) channel after matched
simulations are conducted to verify the performance of the filtering. If 𝛼 = Δ𝑇/𝑇 is defined, the amplitude of the target
Mathematical Problems in Engineering 3
Target 1 matched filtering peak amplitude x1 Target 2 matched filtering peak amplitude x2
Δ T1 sampling point 2
sampling point 1
Δ T2
echo at the first and second sampling points is (1 − 𝛼)𝑥 and The observation model given by Formula (1) is expressed
𝛼𝑥, respectively. as vector matrix form:
Suppose that there are k targets between the two sampling
points. When the radar launches the first pulse, we can get the 𝑍 (:, 𝑚)
target’s observed values at two sampling points, as shown in T
Formula (1): = [𝑠𝑖 (𝑚) 𝑠𝑞 (𝑚) 𝑑𝑎𝑖 (𝑚) 𝑑𝑎𝑞 (𝑚) 𝑑𝑒𝑖 (𝑚) 𝑑𝑒𝑞 (𝑚)] (2)
= 𝐷𝑋 (:, 𝑚) + 𝑊 (:, 𝑚)
𝑘
𝑠𝑖1 (𝑚) = ∑ (1 − 𝛼𝑗 ) 𝑥𝑗 (𝑚) + 𝑛𝑠𝑖1 (𝑚) where
𝑗
𝑘
1 − 𝛼1 1 − 𝛼2 ⋅⋅⋅ 1 − 𝛼𝑘
[ ]
𝑠𝑖2 (𝑚) = ∑𝛼𝑗 𝑥𝑗 (𝑚) + 𝑛𝑠𝑖2 (𝑚) [ 𝛼1 𝛼2 ⋅⋅⋅ 𝛼𝑘 ]
[ ]
𝑗 [ ]
[(1 − 𝛼1 ) 𝜂𝑎1 (1 − 𝛼2 ) 𝜂𝑎2 ⋅⋅⋅ (1 − 𝛼𝑘 ) 𝜂𝑎𝑘 ]
𝐷=[
[ 𝛼𝜂
],
𝛼𝑘 𝜂𝑎𝑘 ]
𝑘
[ 1 𝑎1 𝛼2 𝜂𝑎2 ⋅⋅⋅ ]
𝑑𝑎𝑖1 (𝑚) = ∑ (1 − 𝛼𝑗 ) 𝜂𝑎𝑗 𝑥𝑗 (𝑚) + 𝑛𝑑𝑎𝑖1 (𝑚) [ ]
[ (1 − 𝛼 ) 𝜂 (1 − 𝛼 ) 𝜂 ⋅⋅⋅ (1 − 𝛼𝑘 ) 𝜂𝑒𝑘 ]
𝑗 [ 1 𝑒1 2 𝑒2 ]
(1)
𝑘 [ 𝛼 𝜂
1 𝑒1 𝛼 𝜂
2 𝑒2 ⋅⋅⋅ 𝛼𝑘 𝜂𝑒𝑘 ]
𝑑𝑎𝑖2 (𝑚) = ∑𝛼𝑗 𝜂𝑎𝑗 𝑥𝑗 (𝑚) + 𝑛𝑎𝑖2 (𝑚) (3)
𝑗 𝑛𝑠𝑖1 (1) 𝑛𝑠𝑖1 (2) ⋅ ⋅ ⋅ 𝑛𝑠𝑖1 (𝑀)
[ ]
[ 𝑛𝑠𝑖2 (1) 𝑛𝑠𝑖2 (2) ⋅ ⋅ ⋅ 𝑛𝑠𝑖2 (𝑀) ]
𝑘 [ ]
𝑑𝑒𝑖1 (𝑚) = ∑ (1 − 𝛼𝑗 ) 𝜂𝑒𝑗 𝑥𝑗 (𝑚) + 𝑛𝑒𝑖1 (𝑚) [ ]
[𝑛𝑑𝑎𝑖1 (1) 𝑛𝑑𝑎𝑖1 (2) ⋅ ⋅ ⋅ 𝑛𝑑𝑎𝑖1 (𝑀)]
𝑊=[
[𝑛
].
𝑛𝑑𝑎𝑖2 (2) ⋅ ⋅ ⋅ 𝑛𝑑𝑎𝑖2 (𝑀)]
𝑗
[ 𝑑𝑎𝑖2 (1) ]
[ ]
𝑘 [ 𝑛 (1) 𝑛𝑑𝑒𝑖1 (2) ⋅ ⋅ ⋅ 𝑛𝑑𝑒𝑖1 (𝑀) ]
𝑑𝑒𝑖2 (𝑚) = ∑𝛼𝑗 𝜂𝑒𝑗 𝑥𝑗 (𝑚) + 𝑛𝑒𝑖2 (𝑚) [ 𝑑𝑒𝑖1 ]
𝑗 [ 𝑛𝑑𝑒𝑖2 (1) 𝑛𝑑𝑒𝑖2 (2) ⋅ ⋅ ⋅ 𝑛𝑑𝑒𝑖2 (𝑀) ]
3. Bayesian Parameter Estimation Method where 𝑝(𝑧 | 𝑘, 𝜃𝑘 ) is a likelihood function, which can be
obtained by Formula (4), that is,
3.1. Prior Distribution Hypothesis. Because the actual location
of the target is unknown, it can be assumed that it is evenly 2
−6𝑀/2 𝑧 − 𝑅 (𝑠𝑘 ) 𝑎𝑘
distributed between two consecutive sampling points [13]: 𝑝 (𝑧 | 𝑘, 𝜃𝑘 ) = (2𝜋𝜎𝑘2 ) exp [− ] (12)
2𝜎𝑘2
𝛼𝑗 ∼ 𝑈 (0, 1) ,
The joint posterior distribution of (𝑘, 𝜃𝑘 ) can be obtained by
𝜂𝑎𝑗 ∼ 𝑈 (−1, 1) , chain decomposition method, that is,
(5)
𝜂𝑒𝑗 ∼ 𝑈 (−1, 1) 𝑝 (𝑘, 𝜃𝑘 ) = 𝑝 (𝑘, 𝑠𝑘 , 𝑎𝑘 , 𝜎𝑘2 ) = 𝑝 (𝑘, 𝑠𝑘 , 𝑎𝑘 | 𝜎𝑘2 ) 𝑝 (𝜎𝑘2 ) (13)
𝑗 = 1, 2, ⋅ ⋅ ⋅ 𝑘
According to the prior distribution hypothesis, we can get the
Because the independence between targets, the parameter joint distribution of (𝑘, s𝑘 , a𝑘 ) as follows:
space of 𝑠𝑘 can be expressed as
Λ𝑘 𝑒−Λ 1
𝑝 (𝑘, 𝑠𝑘 , 𝑎𝑘 | 𝜎𝑘2 ) ∝ ⋅ 𝑘
Φ𝑘 ≜ (0, 1)𝑘 × (−1, 1)𝑘 × (−1, 1)𝑘 (6) 𝑘! 4
(14)
Under the condition that the number of targets is 𝑘, the exp (−𝑎𝑇𝑘 Σ−1 𝑎𝑘 /2𝜎𝑘2 )
⋅
location parameters of all 𝑘 targets are 2𝜋𝜎2 Σ1/2
𝑘
𝑘
1 1 1 1 Using the above assumptions and derivation, we can get the
𝑝 (𝑠𝑘 | 𝑘) = ∏ ( ⋅ ⋅ )= 𝑘 (7)
𝑗=1 1 − 0 1 − (−1) 1 − (−1) 4 following:
The prior distribution of target number 𝑘 obeys censored 𝑝 (𝑘, 𝜃𝑘 | 𝑧) ∝ 𝑝 (𝑧 | 𝑘, 𝜃𝑘 ) 𝑝 (𝑘, 𝑠𝑘 , 𝑎𝑘 | 𝜎𝑘2 ) 𝑝 (𝜎𝑘2 )
Poisson distribution:
−6𝑀/2 (𝛾0 + 𝑧𝑇 𝑃𝑘 𝑧) −𝜐0 /2−1
Λ −Λ 𝑘 ∝ (2𝜋𝜎𝑘2 ) exp [− ] (𝜎𝑘2 )
𝑝 (𝑘) ∝ 𝑒 , 𝑘 ∈ [0, 𝑘max ] (8) 2𝜎𝑘2
𝑘!
(15)
Λ𝑘 1 −1/2
Document [14] points out that the number of targets that ⋅ ⋅ ⋅ 2𝜋𝜎𝑘2 Σ
can be distinguished within the beam cannot exceed 5 at 𝑘! 4𝑘
most, that is, 𝑘max = 5. Λ can be understood as the number (𝑎𝑘 − 𝑛𝑘 ) 𝑁−1
𝑇
𝑘 (𝑎𝑘 − 𝑛𝑘 )
of desired targets, and its influence can be eliminated by ⋅ exp [− ]
calculating Bias factor 𝑝(𝑧 | 𝑘1 )/𝑝(𝑧 | 𝑘2 ). 2𝜎𝑘2
Under the condition of (𝑘, 𝑠𝑘 ) is known, the target
amplitude 𝑎𝑘 obeys the Gauss distribution whose mean value where 𝑁−1 𝑘 = Σ
−1
+ 𝑅𝑇 (𝑠𝑘 )𝑅(𝑠𝑘 ), 𝑛𝑘 = 𝑁𝑘 𝑅𝑇 (𝑠𝑘 )𝑧, and 𝑃𝑘 =
𝑇
is zero and the covariance is 𝜎𝑘2 Σ. Σ = diag(Σ 𝑘 , ⋅ ⋅ ⋅ , Σ𝑘 ),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 𝐼 − 𝑅(𝑠𝑘 )𝑁𝑘 𝑅 (𝑠𝑘 ).
𝑀 According to the prior distribution hypothesis of 𝑎𝑘 and
Σ−1 −2 2
𝑘 = 𝛿 𝐼𝑘×𝑘 . 𝛿 denotes the expected signal-to-noise ratio 𝜎𝑘2 , we can eliminate the 𝑎𝑘 and 𝜎𝑘2 in Formula (15) and finally
and obeys the conjugate Gama prior distribution, that is, get the joint posterior distribution of (𝑘, 𝑠𝑘 ) as
𝛼0 𝛽0 𝑝 (𝑘, s𝑘 | z)
𝛿2 ∼ 𝐼𝑔 ( , ) (9)
2 2
𝑀/2 𝑘 (16)
𝛾 + z P𝑘 z𝑇 −(6𝑀+𝜐0 )/2 [Λ/ (1 + 𝛿2 ) ]
The noise power 𝜎𝑘2 can also be assumed to obey the conjugate ∝( 0 )
Gama prior distribution: 2 𝑘! ⋅ 4𝑘
𝜐0 𝛾0
𝜎𝑘2 ∼ 𝐼𝑔 ( , ) (10) Because the location parameters of each target are indepen-
2 2 dent of each other, if 𝑘, 𝑠𝑘𝑖 , and 𝑧 are known, then the location
parameter vector 𝑠𝑘 (:, 𝑖) of the 𝑖 target follows the following
And 𝛼0 , 𝛽0 , 𝜐0 , and 𝛾0 stand for superparameters. distribution:
−(6𝑀+𝜐0 )/2
3.2. Posteriori Distribution Derivation. According to Bayes 𝛾0 + 𝑧𝑇 𝑃𝑘 𝑧
principle, the joint posterior distribution of all parameters is 𝑝 (𝑠𝑘 (:, 𝑖) | 𝑘, 𝑠𝑘𝑖 , 𝑧) ∝ ( ) (17)
2
𝑝 (𝑘, 𝜃𝑘 , 𝑧)
𝑝 (𝑘, 𝜃𝑘 | 𝑧) = ∝ 𝑝 (𝑧 | 𝑘, 𝜃𝑘 ) 𝑝 (𝑘, 𝜃𝑘 ) (11) where 𝑠𝑘𝑖 ≜ 𝑠𝑘 \ 𝑠𝑘 (:, 𝑖) represents all the parameters except
𝑝 (𝑧) column i in matrix 𝑠𝑘 , 1 ≤ 𝑖 ≤ 𝑘.
Mathematical Problems in Engineering 5
Step 1. Initialization: (𝑘(0) , 𝜃(0) If 𝑢 ≤ 𝐴 birth , the state of Markov chain is changed to
𝑘 )∈Θ
{𝑘 + 1, 𝑠𝑘+1 };
Step 2. The ith iteration process is as follows:
Otherwise, the original state {𝑘, 𝑠𝑘 } is kept.
Sample 𝑢 ∼ 𝑈(0, 1); 4.2. The Destruction Process. Suppose the current parameter
space is {𝑘 + 1, 𝑠𝑘+1 }, 𝑘 > 0. Then we remove one target
If 𝑢 ≤ 𝑏𝑘(𝑖) , execute the birth process; randomly and get the new parameter space {𝑘, 𝑠𝑘 }. Similar
to the birth process, we can get the reception rate and the
reception probabilities of the destruction process are
Otherwise, if 𝑢 ≤ (𝑏𝑘(𝑖) + 𝑑𝑘(𝑖) ), then execute the
destruction process;
𝑝 (𝑘 − 1, 𝑠𝑘−1 | 𝑧) 𝑞 (𝑘, 𝑠𝑘 | 𝑘 − 1, 𝑠𝑘−1 )
𝑟death = ⋅
Otherwise, execute the renewal process. 𝑝 (𝑘, 𝑠𝑘 | 𝑧) 𝑞 (𝑘 − 1, 𝑠𝑘−1 | 𝑘, 𝑠𝑘 )
(6𝑀+]0 )/2
Step 3. 𝑖 ← 𝑖 + 1, and return to Step 2. 𝛾 + 𝑧𝑇 𝑃 𝑧 𝑀/2 (22)
=( 0 𝑇 𝑘 ) ⋅ (1 + 𝛿2 ) ⋅𝑘
𝛾0 + 𝑧 𝑃𝑘−1 𝑧
4.1. The Birth Process. Suppose the current parameter space 𝐴 death = min {1, 𝑟death }
is (𝑘, 𝑠𝑘 ) and 𝑘 < 𝑘max . After randomly generating a new
target, the parameter space becomes (𝑘 + 1, 𝑠𝑘+1 ). According
The specific steps of the destruction process can described as
to the prior distribution of target parameters, we can get the
below.
recommended distribution of birth process and extinction
process, that is, Step 1. Remove one target randomly from the existing 𝑘 + 1
goals.
1 1
𝑞 (𝑘 + 1, 𝑠𝑘+1 | 𝑘, 𝑠𝑘 ) = 𝑏𝑘 ⋅ ⋅ Step 2.
1 − 0 1 − (−1)
(19) Calculate the 𝐴 death by Formula (22);
1 𝑏
⋅ = 𝑘
1 − (−1) 4 Take sample 𝑢 ∼ 𝑈(0, 1).
𝑑𝑘+1 Step 3.
𝑞 (𝑘, 𝑠𝑘 | 𝑘 + 1, 𝑠𝑘+1 ) = (20)
𝑘+1
If 𝑢 ≤ 𝐴 death , the state of Markov chain is changed to
{𝑘, 𝑠𝑘 };
Combining formulas (19) and (20), we can get the reception
rate and the reception probability are Otherwise, the original state {𝑘 + 1, 𝑠𝑘+1 } is kept.
6 Mathematical Problems in Engineering
For 𝑗 = 1, ⋅ ⋅ ⋅ , 𝑘
Take sample 𝑢1 ∼ 𝑈(0, 1);
If 𝑢1 < 𝜆 (0 < 𝜆 < 1), take the proposal distribution 𝑞1 (𝑠∗𝑘 (:, 𝑗) | 𝑠𝑘 (:, 𝑗)). And the target distribution of 𝑝(𝑠𝑘 (:, 𝑗) | 𝑧, 𝑠𝑘𝑗 ) as
invariant distribution of 𝜋(𝑠𝑘 (:, 𝑗)), MH performed a step sampling;
Else, take the proposal distribution 𝑞2 (𝑠∗𝑘 (:, 𝑗) | 𝑠𝑘 (:, 𝑗)), And the target distribution of 𝑝(𝑠𝑘 (:, 𝑗) | 𝑧, 𝑠𝑘𝑗 ) as invariant distribution
of 𝜋(𝑠𝑘 (:, 𝑗)), MH performed a step sampling;
End if
End for
RMSE of
3 0.55 0.1 -0.4
0.04
4 0.75 0.5 0.8
5 0.95 0.9 -0.8 0.03
0.02
4.3. The Renewal Process. The renewal process is actually a 0.01
standard MCMC sampling process. In this article, we use
mixed MH sampling to generate 𝑠𝑘 , in order to achieve 0
optimal search and traversal of target location parameter 10 12 14 16 18 20 22 24 26 28 30
space, as shown in Table 1. SNR (dB)
The receiving probability of the update process can be 5 Targets ML-MDL M=10 5 Targets ML-MDL M=50
obtained as 4 Targets ML-MDL M=10 4 Targets ML-MDL M=50
3 Targets ML-MDL M=10 3 Targets ML-MDL M=50
2 Targets ML-MDL M=10 2 Targets ML-MDL M=50
𝐴 update = min {1, 𝑟update } 1 Target ML-MDL M=50
1 Target ML-MDL M=10
(6𝑀+]0 )/2
5 Targets RJ-MCMC M=10
{ 𝛾 + 𝑧𝑇 𝑃𝑘 𝑧 𝑞𝑖 (𝑠𝑘,𝑗 | 𝑠∗𝑘,𝑗 ) } (23) 4 Targets RJ-MCMC M=10
= min {1, ( 0 ) }
𝛾0 + 𝑧𝑇 𝑃∗𝑘 𝑧 𝑞𝑖 (𝑠∗𝑘,𝑗 | 𝑠𝑘,𝑗 ) }
3 Targets RJ-MCMC M=10
{ 2 Targets RJ-MCMC M=10
1 Target RJ-MCMC M=10
To determine the convergence of the direct determination Figure 4: RMSE of 𝛼 estimation under the same power, different
method, calculate a mean ergodic every 50 iterations, if the target numbers, and different SNR conditions.
absolute value of three consecutive mean calculation results
the difference of the control coefficient error is less than the
set, to determine the estimate convergence and to stop the
iteration. azimuth, and pitch angle estimation accuracy, under different
SNR conditions and different real target number conditions.
5. Computer Simulation Results From Figures 4–6, we can see that, in any case, the
estimated RMSE will increase as the number of real targets
Assume there is a maximum of 5 goals between two con- increases. For ML-MDL method, increasing the number of
secutive sampling assumptions. Location parameters and observation pulses can improve the accuracy of parameter
their settings in the document [11] are the same, as shown estimation to a certain extent. In addition, we can see that the
in Table 2. Given the hyperparameters and other related estimation performance of the RJ-MCMC method is between
parameters as [12], 𝛼0 = 2, 𝛽0 = 10, ]0 = 𝛾0 = 0.1, 𝜆 = 0.2, and the other two cases under the condition of 10 subpulses.
Σ𝑞2 = 𝐼/(30𝑀). Monte Carlo simulation 200 times is to get That is to say, compared with the ML-MDL method, the RJ-
the following results. We will call the method in the document MCMC method has better estimation performance under the
[11, 14] abbreviated as ML-MDL. same sub pulse number condition.
Experiment 1 (performance analysis for estimating target Experiment 2 (RMSE estimation of two targets under different
location parameters). When the target has the same power, SNR conditions). Suppose there are two different power
Figures 4–6 show the wavelet door relative distance of target 1, targets and the SNR of target 1 changes from 15dB to 30dB
Mathematical Problems in Engineering 7
0.25 0.045
0.04
0.2
0.035
0.15 0.03
RMSE of ;
RMSE of
0.025
0.1
0.02
0.015
0.05
0.01
0 0.005
10 12 14 16 18 20 22 24 26 28 30 15 20 25 30
SNR (dB) SNR (dB)
5 Targets ML-MDL M=10 5 Targets ML-MDL M=50
4 Targets ML-MDL M=10 4 Targets ML-MDL M=50 Target 1
3 Targets ML-MDL M=10 3 Targets ML-MDL M=50 Target 2
2 Targets ML-MDL M=10 2 Targets ML-MDL M=50
Figure 7: RMSE of 𝛼 estimation of target 1 and target 2 under the
1 Target ML-MDL M=10 1 Target ML-MDL M=50
different power, and different SNR conditions.
5 Targets RJ-MCMC M=10
4 Targets RJ-MCMC M=10
3 Targets RJ-MCMC M=10
2 Targets RJ-MCMC M=10
1 Target RJ-MCMC M=10 0.08
0.075
Figure 5: RMSE of 𝜂𝑎 estimation under the same power, different
target numbers, and different SNR conditions. 0.07
0.065
0.06
RMSE of ;
0.35 0.055
0.05
0.3
0.045
0.25 0.04
0.035
RMSE of ?
0.2
0.03
15 20 25 30
0.15 SNR (dB)
Target 1
0.1
Target 2
0.06
No potential conflicts of interest were reported by the authors.
0.05
Authors’ Contributions
0.04
Jian Gong and Hui Yuan designed the algorithm scheme.
0.03 Jian Gong performed the experiments and analysed the
experiment results. Qun Wan, Hui Yuan, and Yiduo Guo
0.02 contributed to the manuscript drafting and critical revision.
15 20 25 30
All authors read and approved the final manuscript.
SNR (dB)
Target 1
Target 2
Acknowledgments
Figure 9: RMSE of 𝜂𝑒 estimation of target 1 and target 2 under the This work was supported by the National Natural Sci-
different power, and different SNR conditions. ence Foundation of China under Grant nos. 61601502 and
61501501.
International
Journal of
Mathematics and
Mathematical
Sciences
Journal of
Hindawi
Optimization
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018
International Journal of
Engineering International Journal of
Mathematics
Hindawi
Analysis
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018