entropy

Article
Design of Nonlinear Autoregressive Exogenous Model Based
Intelligence Computing for Efficient State Estimation of
Underwater Passive Target
Wasiq Ali 1,2 , Wasim Ullah Khan3, * , Muhammad Asif Zahoor Raja 4 , Yigang He 3, * and Yaan Li 1

1 School of Marine Science and Technology, Northwestern Polytechnical University, Xi’an 710072, China;
[email protected] (W.A.); [email protected] (Y.L.)
2 Department of Electrical and Computer Engineering, COMSATS University Islamabad, Attock Campus,
Attock 43600, Pakistan
3 School of Electrical Engineering and Automation, Wuhan University, Wuhan 430072, China
4 Future Technology Research Center, National Yunlin University of Science and Technology, 123 University
Road, Section 3, Douliou, Yunlin 64002, Taiwan; [email protected]
* Correspondence: [email protected] (W.U.K.); [email protected] (Y.H.)

Abstract: In this study, an intelligent computing paradigm built on a nonlinear autoregressive

exogenous (NARX) feedback neural network model with the strength of deep learning is presented
for accurate state estimation of an underwater passive target. In underwater scenarios, real-time
motion parameters of passive objects are usually extracted with nonlinear filtering techniques. In
filtering algorithms, nonlinear passive measurements are associated with linear kinetics of the target,
governing by state space methodology. To improve tracking accuracy, effective feature estimation





 and minimizing position error of dynamic passive objects, the strength of NARX based supervised
learning is exploited. Dynamic artificial neural networks, which contain tapped delay lines, are
Citation: Ali, W.; Khan, W.U.;
suitable for predicting the future state of the underwater passive object. Neural networks-based
Raja, M.A.Z.; He, Y.; Li, Y. Design of
intelligence computing is effectively applied for estimating the real-time actual state of a passive
Nonlinear Autoregressive Exogenous
Model Based Intelligence Computing
moving object, which follows a semi-curved path. Performance analysis of NARX based neural
for Efficient State Estimation of networks is evaluated for six different scenarios of standard deviation of white Gaussian measurement
Underwater Passive Target. Entropy noise by following bearings only tracking phenomena. Root mean square error between estimated
2021, 23, 550. https://doi.org/ and real position of the passive target in rectangular coordinates is computed for evaluating the
10.3390/e23050550 worth of the proposed NARX feedback neural network scheme. The Monte Carlo simulations are
conducted and the results certify the capability of the intelligence computing over conventional
Academic Editor: Boris Ryabko nonlinear filtering algorithms such as spherical radial cubature Kalman filter and unscented Kalman
filter for given state estimation model.
Received: 1 April 2021
Accepted: 27 April 2021
Keywords: nonlinear autoregressive with exogenous input (NARX); state estimation; artificial neural
Published: 29 April 2021
network; measurement noise; nonlinear filtering; intelligent computing

Publisher’s Note: MDPI stays neutral

with regard to jurisdictional claims in
published maps and institutional affil-
1. Introduction
iations.
The basic aim in state estimation of the underwater passive target is to accurately
approximate the motion variables and target’s actual trajectory by obtaining valuable
data from noisy measurements of hydrophones [1,2]. Real-time state estimation has many
Copyright: © 2021 by the authors.
practical applications in both civil and military fields like target detection [3], aircraft
Licensee MDPI, Basel, Switzerland.
surveillance [4], navigation [5], precise guidance [6] and object localization [7]. To achieve
This article is an open access article
accurate state estimation in these real-life problems, nonlinear filtering algorithms are
distributed under the terms and widely applied by the research community in the last two decades because of their effi-
conditions of the Creative Commons ciency and robustness [8]. Consequently, the accuracy of state estimation phenomena is
Attribution (CC BY) license (https:// largely dependent on the convergence of a specific filtering algorithm [9]. Usually, Bayesian
creativecommons.org/licenses/by/ filtering techniques offer better results in state prediction problems, where real-time param-
4.0/). eters like velocity, position and possibly trajectory of the dynamic target are extracted [10].

Entropy 2021, 23, 550. https://doi.org/10.3390/e23050550 https://www.mdpi.com/journal/entropy

Entropy 2021, 23, 550 2 of 26

In nonlinear systems, Bayesian filtering algorithms are applied in a recursive manner to

achieve an efficient estimation of state parameters [11]. Bearings-only passive state approx-
imation problems in complex ocean environment are specifically solved with nonlinear
filtering techniques and this scheme is usually referred to target motion analysis [12].
In the literature, the famous Kalman filter (KF), following Bayesian methodology, is
widely applied to solve state estimation problems for linear systems [13]. In this filter both
passive bearings and estimated dynamics of a moving vehicle are used to compute the
next probability density function (PDF) of object state. In KF, optimum state estimation of
the dynamic target is calculated using the root mean square error (RMSE) criterion [14].
Practically, in a complex underwater atmosphere, the performance of KF declines due to
nonlinear and undetermined characteristics of the observation model [15]. Therefore, this
phenomenon reduces the applications of KF in practical nonlinear systems. To solve the
issue of non linearity in the measurement model, researchers proposed a Taylor series
based first partial optimal nonlinear variant of KF known as extended Kalman filter
(EKF) [16]. Still, time-varying and undefined noise covariance in real-time state estimation
problems diverge the performance of EKF. To obtain precise position estimation of a passive
object, many adaptive EKF schemes have been established by the research community
that have the ability to update weights of noise covariance throughout the prediction
process [17]. Although better covariance approximation offers an adaptive EKF algorithm,
still many robust nonlinear systems produce poor posterior PDF of state vector by Taylor
series approximation.
A serial Monte Carlo simulations-based technique, known as particle filter (PF),
was introduced to handle the issue of non linearity by applying the bulk of particles for
evaluating posterior density distribution [18]. PF shows a satisfactory performance in
nonlinear location estimation problems but at the cost of complexity and high numerical
computations. As an alternative to linearization, the unscented transform method is
used in unscented Kalman filter (UKF) to solve state error covariance by propagating
sigma points [19]. UKF algorithm has the advantage of less computational cost than
PF and high filtering precision than EKF in nonlinear systems [20]. Along with these
filtering techniques, researchers also proposed many nonlinear filters like cubature Kalman
filter [21], Gaussian filters [22], Gauss-Hermite quadrature filter [23], the sparse grid
quadrature filter [24], shifted Rayleigh filter [25] and ensemble Kalman filter [26]. Some
scholars integrate information entropy theory with parameter extraction techniques to
investigate characteristics of underwater object. Also in multi target tracking the maximum
entropy theory has been combined with JPDA method [27]. Hence, information entropy
theory has also important research value in underwater passive target tracking.
In the last decade, researchers have proposed many soft computing methodologies
by following artificial intelligence studies. In these intelligence computing schemes, the
strength of artificial neural networks (ANN) and optimization with local and global search
algorithms are exploited [28]. The spectrum of ANN modeling is spreading extensively
in various real-life studies such as electromagnetism [29], nonlinear optics [30], nanotech-
nology [31], bioinformatics [32], mathematical equations [33], meteorology [34], fluid
dynamics [35], thermodynamics [36], rotating electrical machine [37], electric motors [38],
atomic physics [39], plasma physics [40] and astrophysics [41]. ANNs show outstanding
and significant performance in real-life applications and emerged as one of the dominating
models [42]. In the comparison with other methods, ANNs have provided a better con-
vergence rate as described in several relevant studies [39–41]. ANNs have the potential to
produce excellent results in non-linear data modeling.
In many practical engineering systems, whose characteristics are time-varying and
which rely on their present state, non linearity is a big challenge to handle. In these systems,
the NARX neural network paradigm can be convenient [43,44]. Recently, NARX has found
wide use in a variety of real life applications such as quantitative investment forecast-
ing [45], air quality forecasting [46], predicting marine engine performance [47], intelligent
proportional integral derivative systems [48] and prediction of solar radiation [49]. The
Entropy 2021, 23, 550 3 of 26

major benefit of this soft computing technique is the potential to handle dynamic inputs
expressed by time-series data sets. Time series prediction by NARX is a variable approach,
in which information about the procedure that initializes the time series is not important.
NARX can be applied for the prediction of the next instant for any input signal. This
property makes it suitable for state estimation, where the output of a target is referred
to as a noise-free interpretation of the input signal. Hence, bearings-only passive state
estimation of an underwater moving target is a purely nonlinear problem in which the
dynamics of a target are continuously time-varying.
In our proposed study, we model NARX neural intelligence to approximate the current
state of the underwater dynamic object in different ocean environments. Our key objective
in this study is to design an appropriate framework to investigate the time series statistics
of passive noisy measurements from the passive target. This proposed work establishes
a novel state estimation application of neural networks to analyze different underwater
scenarios for semi-curved trajectory of the target. The simulation results are compared with
our recent reported study [21], in which spherical radial cubature Kalman filter (SRCKF)
and UKF are used for the same state estimation model.
The numerical values of white Gaussian measurement noise are varied to assume
cluttered and perfect underwater environment. Therefore, in the proposed study, NARX
neural computing is investigated for position RMSE among actual and predicted location
of an underwater kinetic vehicle. A comprehensive flow chart of the proposed work is
illustrated in Figure 1. The highlights of the given scheme are summarized as follows:
• A performance based intelligent computing is presented for the accurate location
estimation of a passive underwater target by manipulating the capability of NARX
based neural network.
• Wiener Velocity Motion (WVM) model is exploited for designing the dynamics of
target in semi-curved trajectory with the parameter of standard deviation of measure-
ment noise.
• State estimates and position error for passive target prove the worth of proposed
intelligent computing over conventional nonlinear variants of KF based on SRCKF
and UKF.
• The performance is further endorsed through minimum RMSE criterion in terms of
accuracy and better convergence rate than conventional nonlinear KFs.
The rest of the study is organized as follows—Section 2 initiate the passive state
estimation system modeling based on state-space phenomena. Then, in Section 3 the
methodology and architecture of NARX neural intelligence is presented with the procedure
of training, testing and validation. Section 4 covers the experimentation of neural network
and its detailed simulation results. The conclusion and a proposal for future study are
revealed in Section 5.
Entropy 2021, 23, 550 4 of 26

Figure 1. Flow chart of the state estimation scheme.

2. Passive State Estimation System Model

In this portion of the study, a passive bearings-only state estimation system model is
designed by following the state-space methodology. The kinetics of an underwater dynamic
object are designed in a two-dimensional Cartesian coordinates system for precise real-time
position estimation. There are eight acoustic sensors for collecting and combining passive
bearings emitted from target. The passive bearings collected from acoustic sensors depend
upon the horizontal angle and orientation of each array component. In the given system
model, movement of the target is assumed in a continuous real semi-curved trajectory,
which we aim to approximate with NARX neural network. This passive state estimation
architecture is shown in Figure 2.
Entropy 2021, 23, 550 5 of 26

Figure 2. Passive state estimation architecture.

2.1. Dynamic Model

In the dynamic model, target state vector consist on position ( xn , yn ) and velocity
(dxn , dyn ) at time step n in the two-dimensional rectangular coordinate system [21]. These
motion variables are illustrated in state vector of the target Stn as:
T
Stn = xnt ytn dxnt dytn

, (1)

where [.]T is representing transpose of matrix in the above equation. Also, at the observer
platform the state vector is represented as:
T
Sbn = xnb ybn dxnb dybn

. (2)

A comparative state vector of target and observer is defined with the following relation-
ship as:
T
Sn = Stn − Sbn = xn yn dxn dyn

. (3)
In this study, kinetics of underwater dynamic passive target are associated with discrete-
time WVM model. Thus, the dynamic model based on the state vector is defined as:

S n = F n −1 S n −1 + υ n −1 , (4)

meanwhile state conversion matrix in the above equation is represented with Fn−1 , which
is used to analyze the response of the given model. Independent white Gaussian process
noise of zero mean is denoted with υn−1 in the above model. Sampling interval di for
process noise is given as:
di = [in−1 − in ], (5)
whereas the state conversion matrix Fn−1 is described as:
 
1 0 di 0
 0 1 0 di 
F n −1 =
 0
. (6)
0 1 0 
0 0 0 1

The dynamic model given in equation (4) should be in discrete-time for accurate state
estimation with NARX neural model. The discrete dynamic model equation will be applied
for evaluating the system for time steps n, while this time step is multiple of di. The
discrete-time dynamic model with appropriate values of its parameters is expressed as:
Entropy 2021, 23, 550 6 of 26

  
1 0 di 0 x n −1
 0 1 0 di   y n −1 
Sn =    + υ n −1 . (7)
 0 0 1 0   dxn−1 
0 0 0 1 dyn−1
| {z }| {z }
F n −1 S n −1

Process noise υn−1 in the above discrete-time dynamic model can be calculated with its
covariance Cdn−1 as:
υn−1 ≈ N(0, Cdn−1 ), (8)
meanwhile above equation can be written as:
h i
Cdn−1 = E υn−1 υnT−1 . (9)

Likewise, dynamic model, the covariance of process noise should also be given in discrete
form with spectral density v as:
1 3 1 2
3 di 2 di
 
0 0
1 3 1 2 
Cdn−1 = 
0 3 di 0 2 di  v.

1 2 (10)
2 di 0 di 0 

1 2
0 2 di 0 di

2.2. Measurement Model

Moreover, state-space methodology is also associated with the measurement model [19].
At time step n the measurement model is designed as:

M n = h ( S n , Γ n ). (11)

In the above model, measurement function h(.) depends upon current noisy passive mea-
surements from acoustic sensors at time step n. Independent white Gaussian measurement
noise in the above measurement model is denoted with Γn . Passive bearings θn impinging
at array element z and current position of the dynamic object ( xn , yn ) at time step n are
combined in measurement function as:
yn − αzy
 
h(Sn ) = θn = . (12)
xn − αzx

Based upon measurement function and measurement noise the inclusive measurement
model Mzn at time step n for antenna element z is expressed as:

yn − αzy
 
−1
Mzn = tan +Γzn . (13)
xn − αzx
| {z }
θn

In measurement model αzy , αzx are representing the localization function of sensors z in
two-dimensional rectangular coordinates, while independent measurement noise Γzn have
zero mean and covariance Cmn−1 . The covariance of measurement noise can be written as:

Γzn ≈ N(0, Cm
n −1 ), (14)

whereas
Cm 2
n−1 = diag( ΨM ). (15)
In the above covariance equation, Ψ is representing the standard deviation of measurement
noise which have larger impact in given passive state estimation scheme. In the proposed
study, the standard deviation of measurement noise Ψ is oscillated for designing ideal
Entropy 2021, 23, 550 7 of 26

and cluttered noisy scenarios in ocean environment. In simulations, standard deviation of

measurement noise is chosen as a criterion for analyzing the accuracy of neural scheme and
nonlinear KFs. The starting location of concern target is assumed at S0 = [−2 − 0.5 1 0]T .
The prior distribution for the initial state is S0 ≈ N (0, P0 ) and P0 is assumed as:

0.1 0 ··· 0 0
 
 0 0.1 ··· 0 0 
.. ..
 
P0 = 
 .. 
. (16)
 . . . 0 0 
 0 0 0 10 0 
0 0 0 0 10.

3. Artificial Neural Network (ANN) Modeling

In this section of the study, ANN modeling is briefly explained with a systematic
description. ANNs, simply known as neural networks, are actually intelligent computing
frameworks developed by the biological neural systems that integrate animal brains. A
typical ANN is constructed from a set of bound nodes called artificial neurons. Each
bound node is used to propagate signals between neurons. In our proposed study, NARX
based ANN is designed for the accurate state prediction of an underwater target. Detailed
modeling of NARX with its training method is explained in the below subsections.

3.1. Nonlinear Autoregressive with Exogenous Input (NARX) Model

Here, the methodology of NARX neural intelligence is mathematically designed for
state estimation of the underwater moving object. The importance of relationship among
designed time series data and supplementary external data in real-life applications cannot
be ignored. A common phenomenon in state estimation problems is that the performance
of the estimation algorithm heavily depends upon noisy bearings. Therefore, the previous
knowledge of noisy bearings can be handy for the modeling of time series to obtain better
state estimation performance. A multilayer structure of NARX neural network is depicted
in Figure 3.

Figure 3. Multilayer structure of NARX neural network.

In the methodology of NARX, neural learning is applied for predicting future values
in time series by efficiently incorporating previous data. External input and latter outputs
of time series data are responsible for estimating the next value in nonlinear model of the
NARX neural network. The multilayer structure of the NARX is designed with hidden layer,
delay layer, input layer and output layer. Here we want to accurately predict state vector
series S(n) of passive object for j previous values of real data series S with additional input
of measurement series M(n) having input delay of k. Based on these notations behavior of
NARX hidden nodes for predicting time series is modeled as:
 
M(n − 1), M(n − 2), ..., M(n − k)
S(n) = R + e ( n ). (17)
S(n − 1), S(n − 2), ..., S(n − j)
Entropy 2021, 23, 550 8 of 26

In the above nonlinear model of the NARX neural scheme [34], measurements M(n) is a Q
dimension input vector while S(n) is representing output vector having R dimension. k
and j are input and output delay orders correspondingly, whereas e(n) is showing neural
network error.
y ( n ) = gv1 ( M ( n ) u v1 + a v1 ), (18)
S ( n ) = gv2 ( y ( n ) u v2 + a v2 ). (19)
In equation (18) hidden layer vector is denoted with y(n) having dimensions of v. Thresh-
olds of hidden layer and input layer are represented with av1 and av2 accordingly. Weighted
connections between hidden, delay and output layers are uv1 and uv2 respectively. gv1 is
denoting the transfer function of hidden nodes while gv2 is representing the activation
function of output nodes.

3.2. NARX Neural Network Architecture

NARX is a recursive discrete-time neural network consist on time series [44] and can
be designed as:
 
S(w), S(w − 1), ..., S(w − d, +1)
S ( w + 1) = R . (20)
M(w − n + 1), ..., M(w − d, − n + 1)

Assuming process dead time delay term n in the above model is approaching to zero and
following this updated NARX, the neural model can be written as:
 
S(w), S(w − 1), ..., S(w − d, +1)
S ( w + 1) = R . (21)
M(w), M(w + 1), ..., M(w − d, 1)

The vector form of the above model is expressed as:

S(w + 1) = R(S(w), M(w)), (22)

where input and output regresses are represented with vectors M(w) and S(w) correspond-
ingly in the above vector form. Here we applied the strength of Levenberg– Marquardt
(LM) algorithm to train neurons of NARX model. Architecture of the NARX neural network
is shown in Figure 4.

Figure 4. Architecture of NARX neural network.

3.3. Levenberg–Marquardt (LM) Training Method

The LM algorithm is widely applied in the literature as the fastest learning mechanism
for the NARX neural network. This algorithm is formulated to compute derivatives of
second-order without calculating the Hessian matrix so its training speed is increased. Two
Entropy 2021, 23, 550 9 of 26

researchers Kenneth Levenberg and Donald Marquardt introduced LM training technique

by integrating Gauss-Newton algorithm (GNM) with steepest descent method (SDM).
The working principle of LM training scheme is joint training procedure of neurons by
initially quadratic estimation is conducted with SDM and after that weights are updated
and improved with GNM. Minimum of function L(M) is initially computed in the training
algorithm from nonlinear functions as:
c
[lb (M)]2
L( M ) = ∑ 2 . (23)
b =1

Like the above model when sum of squares are combined in performance function, then
Hessian matrix H and gradient χ can be calculated in the following form as:

H = ZTn Zn , (24)

χ = ZTn en . (25)
In the Hessian matrix and gradient equation, Z represents the Jacobian matrix. First-order
derivatives of neural network error are combined in this Jacobian matrix in correspon-
dence with biases and weights, while in all training data points en is representing neural
network error vector. LM training technique involves the Jacobian matrix for computations
purpose by incorporating the mean of the sum of the squared errors are combined in the
performance function. The search space of LM algorithm is represented by the following
equation as:
(ZTn Zn + β n I) pn = −ZTn ln . (26)
In the above search space equation, I is denoting the identity matrix while positive scalars
are shown with β n . Thus, in LM training method the rule of updating weights un is
expressed as:
Zn e n
u n +1 = u n − T . (27)
(Zn Zn + β n I)

3.4. Performance Evaluation Criterion

RMSE among true position and predicted position from state estimation techniques
at each time instant n is taken as an evaluation criterion for analyzing the performance.
The minimum value of RMSE is showing better accuracy of each algorithm. Position error
equation in the form of RMSE is computed with the following relationship as:
r
1 N
∑n=1 |SnTrue − SnEst |
2
RMSE(n) = . (28)
N
The actual state vector of the moving target in the above fitness function is symbolized
with SnTrue , whereas the predicted state of object from the proposed algorithm NARX neural
network and conventional Kalman filtering techniques is denoted by SnEst . The maximum
number of samples or data symbols is 500, which are shown with N in above RMSE
function. In the semi curved path at each point n, the position error is computed in this
study for performance analysis.

4. Simulation Results and Discussion

Here in the simulation results and discussion section, the state estimates, position
error, error histogram and regression analysis of proposed NARX neural scheme are
comprehensively explained. Arithmetical values of measurement noise are accelerated
from 0.1 to 2.5 radian for creating six different scenarios in the ocean environment. The least
value of measurement noise Ψ = 0.1 radian is showing an ideal underwater atmosphere,
while the maximum value of measurement noise Ψ = 2.5 radian is representing the complex
noisy ocean environment. Both measurements from the sensor network and true state
Entropy 2021, 23, 550 10 of 26

of the dynamic object are used as input of NARX neural model to accurately find the
estimated state of the dynamic target. The structure of NARX neural model is simulated in
the neural network toolbox of MATLAB. This structure consists of one input, one hidden
and one output layer with feedback path as revealed in Figure 5. True state of passive target
is considered as one input value in NARX model to enhance prediction output. In hidden
layer, the activation function of neurons is based on sigmoid function. In simulations
of NARX structure, the hidden layer is based upon 100 hidden neurons for better state
estimation results.
LM algorithm is used to train NARX neural network in epoch form by following the
Back-propagation algorithm through time phenomena. The network is trained using the in
epoch wise mode. In the designing procedure, passive bearings collected at the acoustic
sensors and true position data are used for estimation and evaluation of results. In overall
time series, 70% data is used for training, validation is using 15% data and the remaining
15% data is used for testing purposes of the results. In this study, the LM algorithm is used
to train the neural network. In the below figure MATLAB toolbox structure of NARX is
given in which we give passive measurements M(n) at the input of x(t), while the target
state from true trajectory S(n) is given at y(t) input.

Figure 5. Toolbox structure of NARX Model for state estimation.

State estimation framework consists of various parameters that need to be properly

tuned for obtaining better performance from state prediction algorithms. The parameters
and their relevent setting is selected on a similar manner as in reported studies [3,19,21,22].
Appropriate values of these parameters are listed in Table 1.

Table 1. State estimation parameters and their suitable values.

Variables Setting
Initial position and velocity of object S0 = [−2 − 0.5 1 0]T
Location function of observer elements (αzy , αzx )
Number of sensors z=8
Spacing of sensors 0.5
Standard deviation of measurement noise Ψ = 0.1 → 2.5 radian
Spectral density of process noise v = 0.01
Step size di = 0.01
Number of delays k=2
Data points 500
Number of hidden neurons 100
Number of target time step 1000
Entropy 2021, 23, 550 11 of 26

4.1. State Prediction of Passive Target by Varying Standard Deviation of Measurement Noise
Here, MATLAB simulation results of SRCKF, UKF and ANN are conducted for es-
timating the real-time state of the passive moving target in rectangular coordinates. The
performance of ANN for efficient state estimation is analyzed over six different scenarios
of standard deviation of measured noise. These scenarios represent the ideal and noisy
underwater environment in which the estimation capability of ANN is exploited. For each
scenario, performance analysis of filtering and NARX neural network can be seen from
Figure 6 to Figure 23 in the form of state estimates, position error, error histogram and
regression of actual and estimated trajectory. Here six different variations of measurement
noise are represented with their mathematical equations and simulations results as:

4.1.1. Scenario 1: Measurement Noise = 0.1 Radian

In this scenario, standard deviation of measurement noise is taken as Ψ = 0.1 radian
which is showing a smooth ocean atmosphere. Whereas the covariance of measured noise
is calculated from this value of standard deviation of measurement noise as:

Cm 2
n−1 = diag( ΨM ). (29)

Covariance in this scenario is developing independent white Gaussian measurement noise

for z sensors at time step n as:
Γzn ≈ N(0, Cm
n −1 ). (30)
Above computed independent white Gaussian measurement noise is adding in measure-
ment model as:
yn − αzy
 
Mzn = tan−1 +Γzn . (31)
xn − αzx
| {z }
θn

The overall measurement model M for sensor z at time step n is described above in which
passive measurements from eight sensors are combined with independent white Gaussian
measurement noise. These passive measurements of sensors computed in the above model
are applied to NARX neural network as input time series Mzn . Target’s time series which is
defining desired output is taken from state vector of the dynamic model as:
T
Sn = Stn − Sbn =

xn yn dxn dyn . (32)

In the above, state vector xn and yn are defining the position of object in x-y axis corre-
spondingly at time step n. These real positions of the dynamic passive object are designed
for complete semi-curved trajectory and applied in NARX neural network modeling as
target time series for approximating desired current states of the object. Simulations results
in the form of state estimates, location error, error histogram and regression of actual and
estimated trajectory for this scenario are shown below.
In scenario 1, an ideal underwater environment is taken by assuming the lowest value
of the standard deviation of measured noise. The major observations in this scenario are
explained as:
• In Figure 6 state estimates of ANN are compared with other two conventional filtering
techniques and it is clear that ANN is exactly following the true trajectory of the
dynamic object with respect to other techniques.
Entropy 2021, 23, 550 12 of 26

State Estimates with SRCKF, UKF and ANN

0.8

Real Semi Curved Trajectory

0.6
SRCKF Estimates
UKF Estimates
0.4 ANN Estimates

0.2

Y axis
0

-0.2

-0.4

-0.6

-0.8
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
X axis

Figure 6. State estimates of passive target.

• In Figure 7 RMSE between true and estimated position of dynamic object is repre-
sented which is clearly showing that ANN is performing far better from conventional
filtering algorithms of SRCKF and UKF.

Standard Deviation of Measurement Noise = 0.1 rad

0.35

SRCKF
0.3 UKF
ANN

0.25

0.2
RMSE

0.15

0.1

0.05

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 7. Position error between true and estimated trajectory.

• In Figure 8a, the error histogram is shown between target values S(n), S(n − 1), ...,
S(n − j) and the estimated value SnEst of target’s position after training feedback ANN.
As these error values indicate how predicted values differ from the target value and it
can be negative. The total error of neural network is distributed in 20 smaller vertical
bars in the graph of Figure 8a, which are called bins. In a specific bin, the number of
data points from the total data set are shown on the Y-axis. IN the middle of the plot, a
bin corresponding to the error of −0.00379 and the height of that bin for training data
set lies near to 700 while validation and test data set lie between 600 and 700. This
means that many samples from different data sets have an error lies in that following
range. The zero error line corresponds to the zero-error value on the error axis (i.e.,
X-axis). In this case, the zero error point falls under the bin with center −0.00379.
• In Figure 8b, the regression results of the ANN scheme are presented for the training,
validation and testing process. The overall data set is divided into these three with a
ratio of 75%, 15% and 15% correspondingly. Neural network analysis of regression is
actually a combination of statistical procedures for predicting correspondence among
single output variable SnEst and target variable Sn . In the regression results, it can be
seen that the actual target and predicted output are overlapping each other. An almost
linear behavior among target and output values is observed, which is clear evidence
for the effectiveness of the proposed NARX model.
Entropy 2021, 23, 550 13 of 26

Error Histogram with 20 Bins Training: R=1 Validation: R=0.99977

Output ~= 1*Target + -0.00074

Output ~= 1*Target + 1.7e-13
2 2
Data Data
Training 1.5 Fit 1.5 Fit
700 Validation
Y=T Y=T
1 1
Test 0.5 0.5
600 Zero Error 0 0

-0.5 -0.5

500 -1 -1
Instances

-1.5 -1.5

400 -1 0 1 2 -1 0 1 2
Target Target

300
Test: R=0.99963 All: R=0.99986

Output ~= 1*Target + -0.00057

Output ~= 1*Target + -0.00026

2 2
Data Data
200 1.5 Fit 1.5 Fit
Y=T Y=T
1 1
100 0.5 0.5

0 0

0 -0.5 -0.5
0.0177

0.1037
-0.08977
-0.07902
-0.06828
-0.05753
-0.04678
-0.03603
-0.02529
-0.01454
-0.00379
0.006954
-0.1005

0.02845
0.03919
0.04994
0.06069
0.07144
0.08218
0.09293
-1 -1

-1.5 -1.5

-1 0 1 2 -1 0 1 2
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 8. Error histogram and regression analysis of ANN in scenario 1.

4.1.2. Scenario 2: Measurement Noise = 0.5 Radian

In scenario 2, standard deviation of measurement noise is taken as Ψ = 0.5 radian,
which produces a small amount of noisy bearings at acoustic sensors. While the covariance
of measurement noise from this value of standard deviation of measurement noise is
given as:
Cm 2
n−1 = diag( ΨM ). (33)
Covariance in this above equation is based on independent white Gaussian measurement
noise for z sensors at time step n as:

Γzn ≈ N(0, Cm
n −1 ). (34)

Above computed independent white Gaussian measurement noise is adding in measure-

ment model as:
yn − αzy
 
z −1
Mn = tan +Γzn . (35)
xn − αzx
| {z }
θn

Total passive measurements M combining from acoustic sensor z at time step n are de-
scribed in the above measurement model with a measurement noise of 0.5 radian. These
passive measurements computed in the above model are applied to NARX neural network
as input time series Mzn . Target time series is taken from the state vector of the dynamic
model as: T
Sn = Stn − Sbn = xn yn dxn dyn

. (36)
These real positions of the dynamic passive object are designed for complete semi-curved
trajectory and applied in NARX neural network modeling as target time series for estimat-
ing desired real-time states of the object. Simulations results in the form of state estimates,
position error, error histogram and regression results of actual and estimated trajectory for
this scenario are shown in Figures 9–11.
The explanation of all simulation results is given as:
• In Figure 9, state estimates of ANN are compared with other two conventional filtering
techniques and it is clear that ANN is also showing better accuracy rate for estimating
the true trajectory of the dynamic object in this scenario.
Entropy 2021, 23, 550 14 of 26

• In Figure 10, RMSE between true and estimated position of the dynamic object is
represented which is expressing competency of ANN over conventional filtering
algorithms of SRCKF and UKF.
• In Figure 11a, error histogram is shown between target values S(n), S(n − 1), ..., S(n − j)
and estimated value SnEst of the target’s position after training feedback ANN. At the
mid of the plot, a bin corresponding to the error of −0.00846 and the height of that
bin for training data set lies near to 700 samples while validation and test data set lie
between 700 and 800 samples. It means that many samples from different data sets
have an error lies in that following range. In this scenario, zero error point falls under
the bin with center −0.00846.
• In Figure 11b, the regression results of ANN scheme are presented for training, vali-
dation and testing process. In regression results efficiency of ANN is seen by linear
trend and the adjacent response of actual target and predicted output.

State Estimates with SRCKF, UKF and ANN

0.8

Real Semi Curved Trajectory

0.6 SRCKF Estimates
UKF Estimates
0.4 ANN Estimates

0.2
Y axis

-0.2

-0.4

-0.6

-0.8
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
X axis

Figure 9. State estimates of passive target.

Standard Deviation of Measurement Noise = 0.5 rad

0.8

SRCKF
0.7 UKF
ANN
0.6

0.5
RMSE

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 10. Position error between true and estimated trajectory.

Entropy 2021, 23, 550 15 of 26

Error Histogram with 20 Bins Training: R=0.99999 Validation: R=0.99459

800

Output ~= 0.98*Target + 0.0033

Output ~= 1*Target + -0.00049
2 Data 2 Data
Training Fit Fit
700 Validation 1
Y=T
1
Y=T

Test
600 Zero Error 0 0

-1 -1
500
Instances

-2 -2
400 -2 -1 0 1 2 -2 -1 0 1 2
Target Target

300
Test: R=0.99351 All: R=0.99763

Output ~= 0.99*Target + 0.00096

Output ~= 0.98*Target + 0.003
2 Data 2 Data
200 Fit Fit
Y=T Y=T
1 1
100
0 0

0
-0.404

0.1234
0.1673
0.2113
0.2553
0.2992
0.3432
0.3871
0.4311
-0.09637
-0.05241
-0.00846
-0.3601
-0.3161
-0.2722
-0.2282
-0.1843
-0.1403

0.03549
0.07944

-1 -1

-2 -2
-2 -1 0 1 2 -2 -1 0 1 2
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 11. Error histogram and regression of ANN in scenario 2.

4.1.3. Scenario 3: Measurement Noise = 1 Radian

In this scenario, the standard deviation of measurement noise is taken as Ψ = 1 radian
which depicts that there is sufficient increment in noise. While the covariance of the
measurement noise for this value of standard deviation is computed as:

Cm 2
n−1 = diag( ΨM ). (37)

Covariance in this scenario depends upon independent white Gaussian measurement noise
for z sensors at time step n as:
Γzn ≈ N(0, Cm
n −1 ). (38)
Above computed independent white Gaussian measurement noise is adding in measure-
ment model as:
yn − αzy
 
z −1
Mn = tan +Γzn . (39)
xn − αzx
| {z }
θn

The overall measurement model M for sensor z at time step n is described above in which
passive measurements from eight acoustic sensors are combining with independent white
Gaussian measurement noise assumed in this scenario. These passive measurements of
sensors computed in the above model are applied to NARX neural network as input time
series Mzn . Target time series is taken from state vector of the dynamic model as:
T
Sn = Stn − Sbn =

xn yn dxn dyn . (40)

Real positions of the dynamic passive object are modeled for complete semi-curved trajec-
tory and applied in NARX neural network modeling as target time series for estimating
desired real-time states of the object. Simulation results for this scenario in the form of
state estimates, least position error, error histogram and regression analysis of actual and
estimated trajectory are shown below.
Simulation results in this scenario are explained as:
• In Figure 12, the state estimates of all techniques are shown in which it is wort noting
that again ANN is estimating real state object with efficiency as compared to SRCKF
and UKF.
Entropy 2021, 23, 550 16 of 26

State Estimates with SRCKF, UKF and ANN

1.2

0.8

0.6

0.4

Y axis 0.2

-0.2
Real Semi Curved Trajectory
-0.4 SRCKF Estimates
UKF Estimates
-0.6 ANN Estimates

-0.8
-3 -2 -1 0 1 2 3
X axis

Figure 12. State estimates of passive target.

• In Figure 13, RMSE between true and estimated position of the dynamic object is rep-
resented which is also depicting the competency of ANN as compared to conventional
filtering algorithms.
Standard Deviation of Measurement Noise = 1 rad
1.8

SRCKF
1.6
UKF
ANN
1.4

1.2

1
RMSE

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 13. Position error between true and estimated trajectory.

• In Figure 14a, error histogram is shown between target values S(n), S(n − 1), ..., S(n − j)
and the estimated value SnEst of target’s position after training feedback ANN. At the
mid of the plot, a bin corresponding to the error of 0.05589 and the height of that bin
for training data set lie near to 700 data points while validation and test data set lies
between 600 and 700 data points. It means that many samples from different data sets
have an error that lies in that given range. In this scenario, the zero error point falls
under the bin with center 0.05589.
• In Figure 14b, regression results of ANN scheme are presented for training, validation
and testing process. In regression results, there is small divergence between actual
target and predicted output because of sufficient amount of noise.
Entropy 2021, 23, 550 17 of 26

Error Histogram with 20 Bins Training: R=1 Validation: R=0.96679

Output ~= 0.97*Target + -0.017

700

Output ~= 1*Target + -3e-14

Data Data
Training 2 Fit 2 Fit
Y=T Y=T
Validation
600 1 1
Test
Zero Error 0 0

500 -1 -1
Instances

-2 -2
400
-2 -1 0 1 2 -2 -1 0 1 2
Target Target
300
Test: R=0.96561 All: R=0.98146

Output ~= 0.97*Target + -0.037

Output ~= 0.98*Target + -0.016

200 Data Data
2 Fit 2 Fit
Y=T Y=T

1 1
100
0 0

0 1.03
-1.161

-0.431

0.1776
0.2993
0.4211
0.5428
0.6645
0.7862
-0.06583
-1.04

0.908

1.151
-1
-0.9179
-0.7962
-0.6745
-0.5527

-0.3093
-0.1876

-1
0.05589

-2 -2

-2 -1 0 1 2 -2 -1 0 1 2
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 14. Error histogram and regression of ANN in scenario 3.

4.1.4. Scenario 4: Measurement Noise = 1.5 Radian

In this scenario, the standard deviation of measurement noise is taken as Ψ = 1.5
radian, which shows a relatively noisy underwater environment. While the covariance of
measurement noise computed in this scenario is defined with the following mathematical
expression as:
Cm 2
n−1 = diag( ΨM ). (41)
Covariance is depending on independent white Gaussian measurement noise for z sensors
at time step n as:
Γzn ≈ N(0, Cm
n −1 ). (42)
Above computed independent white Gaussian measurement noise is adding in measure-
ment model as:
yn − αzy
 
Mzn = tan−1 +Γzn . (43)
xn − αzx
| {z }
θn

Measurement model M for sensor z at time step n is designed above in which passive mea-
surements from acoustic array elements are combining with independent white Gaussian
measurement noise. These passive bearings are applied to NARX neural network as input
time series Mzn . Target time series is taken from state vector of the dynamic model as:
T
Sn = Stn − Sbn =

xn yn dxn dyn . (44)

xn and yn in above state vector are defining the position of object in x-y axis correspondingly
at time step n. These real positions of the dynamic passive object are designed for complete
semi-curved trajectory and applied in NARX neural network modeling as target time series
for estimating desired real-time state of the object. Simulations results in the form of state
estimates, position error, error histogram and regression analysis of actual and estimated
trajectory for this scenario are shown in Figures 15–17.
Entropy 2021, 23, 550 18 of 26

State Estimates with SRCKF, UKF and ANN

1.5

0.5

Y axis

Real Semi Curved Trajectory

-0.5 SRCKF Estimates
UKF Estimates
ANN Estimates

-1
-4 -3 -2 -1 0 1 2 3 4
X axis

Figure 15. State estimates of passive target.

The major observations in this scenario are discussed as:

• In Figure 15, state estimates of ANN, SRCKF and UKF are presented in which it is
clearly seen that ANN is converging more than other two algorithms.
• In Figure 16, RMSE between true and estimated position of the dynamic object is
represented which is also showing that ANN has less position error from SRCKF
and UKF.
Standard Deviation of Measurement Noise = 1.5 rad
2.5

SRCKF
UKF
ANN
2

1.5
RMSE

0.5

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 16. Position error between true and estimated trajectory.

• In Figure 17a, error histogram is shown between target values S(n), S(n − 1), ..., S(n − j)
and estimated value SnEst of target’s position after the training feedback ANN model.
At the mid of the plot, a bin corresponding to the error of 0.058 and the height of that
bin for training data set lies near to 700 samples while validation and test data set
lie between 600 and 700 samples. It means that many samples from different data
sets have an error lies in that following range. Zero error line corresponding to the
zero-error value on the error axis. In this scenario, zero error point falls under the bin
with center 0.058.
• In Figure 17b, the regression analysis of ANN modeling is presented for the training,
validation and testing process. In regression results some distance is appearing
between actual target and predicted output because of noisy ocean condition.
Entropy 2021, 23, 550 19 of 26

Error Histogram with 20 Bins Training: R=1 Validation: R=0.95532

3

Output ~= 1*Target + 4.3e-14

3

Output ~= 1*Target + 0.024

Data Data
700 Training Fit Fit
2 Y=T 2 Y=T
Validation
600 Test 1 1
Zero Error
0 0
500
-1 -1
Instances

-2 -2
400
-2 -1 0 1 2 3 -2 -1 0 1 2 3
Target Target
300
Test: R=0.95529 All: R=0.97635

Output ~= 0.98*Target + 0.0016

Output ~= 0.99*Target + 0.0043

3 3
200 Data Data
Fit Fit
2 Y=T 2 Y=T

100 1 1

0 0
0
-1.439
-1.273
-1.106

-0.441

0.2243
0.3907

0.7233
0.8897
-0.94

0.058

0.557

1.056
1.222
1.389
1.555
1.721
-0.7737
-0.6073

-0.2747
-0.1083

-1 -1

-2 -2

-2 -1 0 1 2 3 -2 -1 0 1 2 3
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 17. Error histogram and regression of ANN in scenario 4.

4.1.5. Scenario 5: Measurement Noise = 2 Radian

In this scenario, standard deviation of measurement noise is taken as Ψ = 2 radian.
While the covariance of measurement noise from this value of standard deviation of
measurement noise is computed as:

Cm 2
n−1 = diag( ΨM ). (45)

Covariance in this scenario develops from independent white Gaussian measurement noise
for z sensors at time step n as:
Γzn ≈ N(0, Cm
n −1 ). (46)
The above computed independent white Gaussian measurement noise is adding up in
measurement model as:
yn − αzy
 
z −1
Mn = tan +Γzn . (47)
xn − αzx
| {z }
θn

Passive bearings collected at sensors are computed in the above model and applied to
NARX neural network as input time series Mzn . Target time series which is defining desire
output is taken from state vector of the dynamic model as:
T
Sn = Stn − Sbn =

xn yn dxn dyn . (48)

These real-time positions of the dynamic passive object are designed for each point of
complete semi-curved trajectory and applied in NARX neural network modeling as target
time series for estimating desired real-time states of the object. Simulations results in the
form of state estimates, position error, error histogram and regression analysis of actual
and estimated trajectory for this scenario are shown below.
Concluding remarks on simulation results of this scenario are given as:
• State estimates of SRCKF, UKF and ANN are presented in Figure 18 and again ANN per-
forms far better at estimating the semi-curved trajectory than the other two algorithms.
Entropy 2021, 23, 550 20 of 26

State Estimates with SRCKF, UKF and ANN

2

1.5

0.5

Y axis
0

-0.5
Real Semi Curved Trajectory
SRCKF Estimates
-1 UKF Estimates
ANN Estimates

-1.5
-3 -2 -1 0 1 2 3
X axis

Figure 18. State estimates of passive target.

• In Figure 19, RMSE between true and estimated position of dynamic object is rep-
resented which is showing that ANN has less position error as compared SRCKF
and UKF.
Standard Deviation of Measurement Noise = 2 rad
2

1.8 SRCKF
UKF
1.6 ANN

1.4

1.2
RMSE

0.8

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 19. Position error between true and estimated trajectory.

• In Figure 20a, the error histogram is shown between target values S(n), S(n − 1),
..., S(n − j) and estimated value SnEst of target’s position after training feedback ANN.
At the mid of the plot, a bin corresponding to the error of −0.04745 and the height of
that bin for training data set lie near to 600 samples while validation and test data set
lie between 500 and 600 samples. It means that many samples from different data sets
have an error lies in that following range. In this case, the zero error point falls under
the bin with center −0.04745.
• In Figure 20b, the regression results of ANN modeling are presented for the training,
validation and testing process. In the regression results, some divergence between the
actual target and the predicted output appears because of complex noisy bearings.
Entropy 2021, 23, 550 21 of 26

Error Histogram with 20 Bins Training: R=1 Validation: R=0.96932

Output ~= 1*Target + -5.1e-14

Output ~= 1*Target + -0.018

Data Data
Training 2 Fit 2 Fit
600 Y=T Y=T
Validation
1 1
Test
500 Zero Error 0 0

-1 -1
Instances

400
-2 -2

-2 -1 0 1 2 -2 -1 0 1 2
300 Target Target

Test: R=0.97182 All: R=0.98466

200

Output ~= 1*Target + 0.00052

Output ~= 1*Target + -0.0026

Data Data
2 Fit 2 Fit
Y=T Y=T

100 1 1

0 0

0
-1.009

-0.528

0.1448
0.2409

0.4331
0.5293
0.6254
0.7215
0.8176
-0.04745

0.337
-0.9125
-0.8164
-0.7203
-0.6242

-0.4319
-0.3358
-0.2397
-0.1436

0.04867

-1 -1

-2 -2

-2 -1 0 1 2 -2 -1 0 1 2
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 20. Error histogram and regression of ANN in scenario 5.

4.1.6. Scenario 6: Measurement Noise = 2.5 Radian

In the last scenario of this study, the standard deviation of measurement noise is
taken as Ψ = 2.5 radian, which shows an extremely noisy and cluttered ocean environment.
The covariance of measurement noise calculated from this value of standard deviation of
measurement noise is written as:

Cm 2
n−1 = diag( ΨM ). (49)

Covariance is designed from independent white Gaussian measurement noise for z sensors
at time step n as:
Γzn ≈ N(0, Cm
n −1 ). (50)
Above computed independent white Gaussian measurement noise is adding in measure-
ment model as:
yn − αzy
 
z −1
Mn = tan +Γzn . (51)
xn − αzx
| {z }
θn

Measurement model M for sensor z at time step n is presented above in which passive
measurements from eight acoustic sensors are combined with independent white Gaussian
measurement noise. These passive measurements of sensors computed in the above model
are applied to the NARX neural network as input time series Mzn . The target time series,
which defines the required output is taken from the state vector of the dynamic model as:
T
Sn = Stn − Sbn =

xn yn dxn dyn . (52)

These real positions of the dynamic passive object shown in the above equation are designed
for complete semi-curved trajectory and applied in NARX neural network modeling as
a target time series for estimating the desired real-time states of object. In this scenario,
the simulation results in the form of state estimates, position error, error histogram and
regression results of actual and estimated trajectory are shown below.
In scenario 6, a noisy underwater environment is taken by assuming the highest value
of standard deviation of measured noise. Here simulation results are discussed as:
Entropy 2021, 23, 550 22 of 26

• In Figure 21, state estimates of ANN are analyzed with other two conventional filtering
techniques and it is worth to note that ANN is also showing its command even in the
cluttered ocean environment.
State Estimates with SRCKF, UKF and ANN
3

2.5

1.5

1
Y axis

0.5

-0.5 Real Semi Curved Trajectory

SRCKF Estimates
-1 UKF Estimates
ANN Estimates

-1.5
-3 -2 -1 0 1 2 3
X axis

Figure 21. State estimates of passive target.

• In Figure 22, RMSE between true and estimated position of dynamic object is repre-
sented which showing that ANN is estimating position of passive dynamic object
with less position error.
Standard Deviation of Measurement Noise = 2.5 rad
4

SRCKF
3.5 UKF
ANN
3

2.5
RMSE

1.5

0.5

0
0 50 100 150 200 250 300 350 400 450 500
Data Points

Figure 22. Position error between true and estimated trajectory.

• In Figure 23a, the error histogram is shown between target time seriesS(n), S(n − 1),
..., S(n − j) and estimated value SnEst of target’s position after training the neural
network. At the center of the graph, a bin incorporating the error of 0.05052 and the
height of that bin for training data set lies near to 600 samples while the validation
and test data set lie between 500 and 600 samples. It means that many samples from
different data sets have an error that lies in that following range. In this case, the zero
error point falls under the bin with center 0.05052.
• In Figure 23b, the regression output of the ANN modeling is presented for training,
validation and testing purposes. In the regression results, many divergence points
between the actual target and the predicted output appear because this scenario has
extremely noisy passive bearings.
Entropy 2021, 23, 550 23 of 26

Error Histogram with 20 Bins Training: R=1 Validation: R=0.95836

Output ~= 1*Target + -5.3e-15

Output ~= 1*Target + 0.006

Data Data
Training 2 Fit 2 Fit
600 Y=T Y=T
Validation
1 1
Test
500 Zero Error 0 0

-1 -1
Instances

400 -2 -2

-2 -1 0 1 2 -2 -1 0 1 2
300 Target Target

Test: R=0.96941 All: R=0.9825

200

Output ~= 1*Target + 0.0055

Output ~= 1*Target + 0.013
Data Data
2 Fit 2 Fit
Y=T Y=T

100 1 1

0 0

0
0.38
-1.158
-1.048
-0.938

-0.279

0.1604
0.2702

0.4899
0.5997
0.7095
0.8193
0.9292
-0.05931

-1
-0.8281
-0.7183
-0.6085
-0.4986
-0.3888

-0.1691

0.05052

-1

-2 -2

-2 -1 0 1 2 -2 -1 0 1 2
Errors = Targets - Outputs Target Target

(a) Error histogram between target values and (b) Regression results of neural network
predicted values

Figure 23. Error histogram and regression of ANN in scenario 6.

Simulation results from all scenarios show that, for higher values of standard deviation
of measured noise Ψ, state estimation techniques experience difficulties in approaching
the real position of the underwater dynamic object. However, in the comparison between
all estimation techniques, neural network intelligence paradigm NARX shows a better
performance, which is clearly evident for its effectiveness in nonlinear state estimation
problems in an underwater atmosphere.
Along with the simulation results, we also compute RMSE in meters between the
actual and predicted position of the target. These position error responses also endorse
the above results that the accuracy of ANN is more than double from Kalman filters and it
shows the effectiveness of neural network applications in the state estimation of passive
target problems. These position errors computed from SRCKF, UKF and ANN are given in
Figure 24.

Figure 24. Average RMSEs for estimating the position with SRCKF, UKF and ANN by varying
measurement noise
Entropy 2021, 23, 550 24 of 26

5. Conclusions
Intelligent computing through the NARX based neural network is investigated effec-
tively for underwater bearings-only passive state estimation application. The instantaneous
position of a passive dynamic target is predicted in two-dimension x-y coordinates at each
time instant. Initially, bearings only and state space-based state estimation framework
of dynamic and measurement model is mathematically designed. Later, the intelligent
computing paradigm based on NARX is designed for the particular problem of the state
estimation of the passive object. The NARX based supervised neural network is analyzed
for 500 data samples. The competency of neural computing is assessed for semi-curved
target movement in the form of minimum root mean square position error. Appropriate
numerical values of the white Gaussian measured noise are applied to examine the perfor-
mance of the proposed methodology. Simulation results clearly depict that the accuracy
of the neural network is superior from conventional nonlinear filtering algorithms like
SRCKF and UKF. In the noisy underwater scenario, the exponential decay is noticed in
the results of all algorithms. Thus, obtaining accurate state estimation in a complex ocean
environment is still a challenging task and has wide room for development.
In the future, fractional evolutionary and swarming techniques [50–53] can be investi-
gated for obtaining better state prediction of highly maneuvering object for non-Gaussian
distribution of measurement noise, which is still an exciting research dimension in the state
estimation of underwater single or multi targets. This work can also be extended for real
implementation of the proposed state estimation scheme.

Author Contributions: Conceptualization, W.A., Y.L., and M.A.Z.R.; Methodology, W.A. and M.A.Z.R.;
Software, W.A. and Y.L.; Investigation, W.A. and W.U.K.; Resources, Y.H.; Writing original draft
preparation, W.A. and W.U.K.; Writing review and editing, Y.L., M.A.Z.R., W.U.K., Y.H.; Visualiza-
tion, W.A. and W.U.K.; Supervision, Y.L. and M.A.Z.R.; Project administration, Y.L. and M.A.Z.R.;
Funding acquisition, W.U.K. and Y.H. All authors have read and agreed to the published version of
the manuscript.
Funding: This work was supported by the National Natural Science Foundation of China un-
der Grant No. 51977153, 51977161, 51577046, the State Key Program of National Natural Science
Foundation of China under Grant No. 51637004, the national key research and development plan
“important scientific instruments and equipment development” Grant No.2016YFF0102200, Equip-
ment research project in advance Grant No.41402040301, Wuhan Science and Technology Plan Project
Grant No.20201G01.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ge, B.; Zhang, H.; Jiang, L.; Li, Z.; Butt, M.M. Adaptive unscented Kalman filter for target tracking with unknown time-varying
noise covariance. Sensors 2019, 19, 1371. [CrossRef]
2. Li, Y.; Chen, X.; Yu, J.; Yang, X. NA fusion frequency feature extraction method for underwater acoustic signal based on variational
mode decomposition, duffing chaotic oscillator and a kind of permutation entropy. Electronics 2019, 8, 61. [CrossRef]
3. Ali, W.; Li, Y.; Raja, M.A.Z.; Ahmed, N. Generalized pseudo Bayesian algorithms for tracking of multiple model underwater
maneuvering target. Appl. Acoust. 2020, 166, 107345. [CrossRef]
4. Patra, N.; Sadhu, S.; Ghoshal, T.K. Adaptive state estimation for tracking of civilian aircraft. IET Sci. Meas. Technol. 2018, 12,
777–784. [CrossRef]
5. Fan, X.; Shang, S.; Song, D.; Sun, W.; Wen, Y. Weak target detection technology of passive radar on the navigation satellite signal.
J. Eng. 2019, 2019, 6671–6674. [CrossRef]
6. Hong, J.; Laflamme, S.; Dodson, J.; Joyce, B. Introduction to state estimation of high-rate system dynamics. Sensors 2018, 18, 217.
[CrossRef]
7. Wirnshofer, F.; Schmitt, P.S.; Meister, P.; Wichert, G.V.; Burgard, W. State estimation in contact-rich manipulation. In Proceedings
of the 2019 International Conference on Robotics and Automation (ICRA), Montreal, QC, Canada, 20–24 May 2019; pp. 3790–3796.
8. Luo, J.; Han, Y.; Fan, L. Underwater Acoustic Target Tracking: A Review. Sensors 2018, 18, 112. [CrossRef]
9. Ahmed, H.; Ullah, I.; Khan, U.; Qureshi, M.B.; Manzoor, S.; Muhammad, N.; Shahid Khan, M.U.; Nawaz, R. Adaptive Filtering on
GPS-Aided MEMS-IMU for Optimal Estimation of Ground Vehicle Trajectory. Sensors 2019, 19, 5357. [CrossRef]
10. Li, X.; Zhao, C.; Yu, J.; Wei, W. Underwater Bearing-Only and Bearing-Doppler Target Tracking Based on Square Root Unscented
Kalman Filter. Entropy 2019, 21, 740. [CrossRef]
Entropy 2021, 23, 550 25 of 26

11. Li, Y.; Cheng, Y.; Li, X.; Wang, H.; Hua, X.; Qin, Y. Bayesian Nonlinear Filtering via Information Geometric Optimization. Entropy
2017, 19, 655. [CrossRef]
12. Nguyen, N.H.; Doğançay, K. Multistatic pseudolinear target motion analysis using hybrid measurements. Signal Process. 2017,
130, 22–36. [CrossRef]
13. Pei, Y.; Biswas, S.; Fussell, D.S.; Pingali, K. An elementary introduction to kalman filtering. Commun. ACM 2019, 62, 122–133.
[CrossRef]
14. He, Y.; Jiang, J.; Huang, H.; Zhuo, S.; Wu, Y. Kalman Filtering Algorithm for Systems with Stochastic Nonlinearity Functions,
Finite-Step Correlated Noises, and Missing Measurements. Discret. Dyn. Nat. Soc. 2018, 2018, 1516028 . [CrossRef]
15. Zhang, M.; Li, K.; Hu, B.; Meng, C. Comparison of Kalman Filters for Inertial Integrated Navigation. Sensors 2019, 19, 1426.
[CrossRef]
16. Julier, S.J.; Uhlmann, J.K. New extension of the Kalman filter to nonlinear systems. Signal Process. Sens. Fusion Target Recognit. VI
1997, 3068, 182–193.
17. Leung, H.; Zhu, Z.; Ding, Z. An aperiodic phenomenon of the extended Kalman filter in filtering noisy chaotic signals. Mech. Syst.
Signal Process. 2020, 143, 106837. [CrossRef]
18. Lee, Y.; Majda, A.J. State estimation and prediction using clustered particle filters. Proc. Natl. Acad. Sci. USA 2016, 113,
14609–14614. [CrossRef]
19. Ali, W.; Li, Y.; Tanoli, S.A.K.; Raja, M.A.Z.; Javaid, K.; Ahmed, N. SConvergence Analysis of Unscented Transform for Underwater
Passive Target Tracking in Noisy Environment. In Proceedings of the 2019 IEEE International Conference on Signal Processing,
Communications and Computing (ICSPCC), Dalian, China, 20–22 September 2019; pp. 1–6.
20. Wang, Y.; Qiu, Z.; Qu, X. An improved unscented Kalman filter for discrete nonlinear systems with random parameters. Discret.
Dyn. Nat. Soc. 2017, 2017, 7905690 . [CrossRef]
21. Ali, W.; Li, Y.; Chen, Z.; Raja, M.A.Z.; Ahmed, N.; Chen, X. Application of Spherical-Radial Cubature Bayesian Filtering and
Smoothing in Bearings Only Passive Target Tracking. Entropy 2019, 21, 1088. [CrossRef]
22. Su, J.; Li, Y.; Ali, W. Underwater angle-only tracking with propagation delay and time-offset between observers. Signal Process.
2020, 176, 107581. [CrossRef]
23. Qin, W.; Wang, X.; Cui, N. Maximum correntropy sparse Gauss–Hermite quadrature filter and its application in tracking ballistic
missile. IET Radar Sonar Navig. 2017, 11, 1388–1396. [CrossRef]
24. Jia, B.; Xin, M.; Cheng, Y. Sparse-grid quadrature nonlinear filtering. Automatica 2012, 48, 327–341. [CrossRef]
25. Hou, J.; Yang, Y.; Gao, T. Variational Bayesian based adaptive shifted Rayleigh filter for bearings-only tracking in clutters. Sensors
2019, 19, 1512. [CrossRef]
26. Zhang, Z.; Fu, K.; Sun, X.; Ren, W. Multiple target tracking based on multiple hypotheses tracking and modified ensemble
Kalman filter in multi-sensor fusion. Sensors 2019, 19, 3118. [CrossRef] [PubMed]
27. Chen, X.; Li, Y.; Yu, J.; Li, Y. Developing the fuzzy c-means clustering algorithm based on maximum entropy for multitarget
tracking in a cluttered environment. J. Appl. Remote Sens. 2018, 12, 016019. [CrossRef]
28. Ahmad, I.; Zahid, H.; Ahmad, F.; Raja, M.A.Z.; Baleanu, D. Design of computational intelligent procedure for thermal analysis of
porous fin model. Chin. J. Phys. 2019, 59, 641–655. [CrossRef]
29. Khan, J.A.; Raja, M.A.Z.; Rashidi, M.M.; Syam, M.I.; Wazwaz, A.M. Nature-inspired computing approach for solving non-linear
singular Emden–Fowler problem arising in electromagnetic theory. Connect. Sci. 2015, 27, 377–396. [CrossRef]
30. Ahmad, I.; Ahmad, S.; Awais, M.; Ahmad, S.U.I.; Raja, M.A.Z. Neuro-evolutionary computing paradigm for Painlevé equation-II
in nonlinear optics. Eur. Phys. J. Plus 2018, 133, 184. [CrossRef]
31. Mehmood, A.; Zameer, A.; Ling, S.H.; Raja, M.A.Z. Design of neuro-computing paradigms for nonlinear nanofluidic systems of
MHD Jeffery–Hamel flow. J. Taiwan Inst. Chem. Eng. 2018, 91, 57–85. [CrossRef]
32. Raja, M.A.Z.; Umar, M.; Sabir, Z.; Khan, J.A.; Baleanu, D. A new stochastic computing paradigm for the dynamics of nonlinear
singular heat conduction model of the human head. Eur. Phys. J. Plus 2018, 133, 364. [CrossRef]
33. Sabir, Z.; Wahab, H.A.; Umar, M.; Sakar, M.G.; Raja, M.A.Z. Novel design of Morlet wavelet neural network for solving second
order Lane–Emden equation. Math. Comput. Simul. 2020, 172, 1–14. [CrossRef]
34. Bukhari, A.H.; Sulaiman, M.; Islam, S.; Shoaib, M.; Kumam, P.; Raja, M.A.Z. Neuro-fuzzy modeling and prediction of summer
precipitation with application to different meteorological stations. Alex. Eng. J. 2020, 59, 101–116. [CrossRef]
35. Raja, M.A.Z.; Mehmood, A.; ur Rehman, A.; Khan, A.; Zameer, A. Bio-inspired computational heuristics for Sisko fluid flow and
heat transfer models. Appl. Soft Comput. 2018, 71, 622–648. [CrossRef]
36. Ahmad, I.; Raja, M.A.Z.; Bilal, M.; Ashraf, F. Neural network methods to solve the Lane–Emden type equations arising in
thermodynamic studies of the spherical gas cloud model. Neural Comput. Appl. 2017, 28, 929–944. [CrossRef]
37. Raja, M.A.Z.; Niazi, S.A.; Butt, S.A. An intelligent computing technique to analyze the vibrational dynamics of rotating electrical
machine. Neurocomputing 2017, 219, 280–299. [CrossRef]
38. Ahmad, I.; Ahmad, F.; Raja, M.A.Z.; Ilyas, H.; Anwar, N.; Azad, Z. Intelligent computing to solve fifth-order boundary value
problem arising in induction motor models. Neural Comput. Appl. 2018, 29, 449–466. [CrossRef]
39. Sabir, Z.; Manzar, M.A.; Raja, M.A.Z.; Sheraz, M.; Wazwaz, A.M. Neuro-heuristics for nonlinear singular Thomas-Fermi systems.
Appl. Soft Comput. 2018, 65, 152–169. [CrossRef]
Entropy 2021, 23, 550 26 of 26

40. Raja, M.A.Z.; Shah, F.H.; Tariq, M.; Ahmad, I. Design of artificial neural network models optimized with sequential quadratic
programming to study the dynamics of nonlinear Troesch’s problem arising in plasma physics. Neural Comput. Appl. 2018, 29,
83–109. [CrossRef]
41. Ahmad, I.; Raja, M.A.Z.; Bilal, M.; Ashraf, F. Bio-inspired computational heuristics to study Lane–Emden systems arising in
astrophysics model. SpringerPlus 2016, 5, 1866. [CrossRef] [PubMed]
42. Esteves, J.T.; de Souza Rolim, G.; Ferraudo, A.S. Rainfall prediction methodology with binary multilayer perceptron neural
networks. Clim. Dyn. 2019, 52, 2319–2331. [CrossRef]
43. Hatata, A.Y.; Eladawy, M. Prediction of the true harmonic current contribution of nonlinear loads using NARX neural network.
Alex. Eng. J. 2018, 57, 1509–1518. [CrossRef]
44. Solanki, V.; Joshi, M. Energy Efficient NARX Model for Target Tracking in Wireless Sensor Network. In Proceedings of the
2018 3rd IEEE International Conference on Recent Trends in Electronics, Information & Communication Technology (RTEICT),
Bangalore, India, 18–19 May 2018; pp. 260–265.
45. Peng, J.; Tang, Q. Application of NARX Dynamic Neural Network in Quantitative Investment Forecasting System. In International
Symposium on Intelligence Computation and Applications, Proceedings of the 11th International Symposium, ISICA 2019, Guangzhou,
China, 16–17 November 2019; Springer: Berlin/Heidelberg, Germany, 2019; pp. 628–635.
46. Pisoni, E.; Farina, M.; Carnevale, C.; Piroddi, L. PForecasting peak air pollution levels using NARX models. Eng. Appl. Artif.
Intell. 2009, 22, 593–602. [CrossRef]
47. Raptodimos, Y.; Lazakis, I. Application of NARX neural network for predicting marine engine performance parameters. Ships
Offshore Struct. 2020, 15, 443–452. [CrossRef]
48. Liu, J.; Li, T.; Zhang, Z.; Chen, J. NARX prediction-based parameters online tuning method of intelligent PID system. IEEE Access
2020, 8, 130922–130936. [CrossRef]
49. Rangel, E.; Cadenas, E.; Campos-Amezcua, R.; Tena, J.L. Enhanced Prediction of Solar Radiation Using NARX Models with
Corrected Input Vectors. Energies 2020, 13, 2576. [CrossRef]
50. Zameer, A.; Muneeb, M.; Mirza, S.M.; Raja, M.A.Z. Fractional-order particle swarm based multi-objective PWR core loading
pattern optimization. Ann. Nucl. Energy 2020, 135, 106982. [CrossRef]
51. Muhammad, Y.; Khan, R.; Ullah, F.; ur Rehman, A.; Aslam, M.S.; Raja, M.A.Z. Design of fractional swarming strategy for solution
of optimal reactive power dispatch. Neural Comput. Appl. 2019, 11, 1–18. [CrossRef]
52. Akbar, S.; Zaman, F.; Asif, M.; Rehman, A.U.; Raja, M.A.Z. Novel application of FO-DPSO for 2-D parameter estimation of
electromagnetic plane waves. Neural Comput. Appl. 2019, 31, 3681–3690. [CrossRef]
53. Lodhi, S.; Manzar, M.A.; Raja, M.A.Z. Fractional neural network models for nonlinear Riccati systems. Neural Comput. Appl. 2019,
31, 359–378. [CrossRef]