Estimation of water momentum and propeller velocity in bow thruster model of autonomous surface vehicle using modified Kalman filter

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 6, December 2022, pp. 5988~5997
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i6.pp5988-5997  5988
Journal homepage: http://ijece.iaescore.com
Estimation of water momentum and propeller velocity in bow
thruster model of autonomous surface vehicle using modified
Kalman filter
Hendro Nurhadi1
, Mayga Kiki2
, Dieky Adzkiya2
, Teguh Herlambang3
1
Department of Industrial Mechanical Engineering, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
2
Department of Mathematics, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
3
Department of Information Systems, Universitas Nahdlatul Ulama Surabaya, Surabaya, Indonesia
Article Info ABSTRACT
Article history:
Received Jul 12, 2021
Revised Jul 22, 2022
Accepted Aug 12, 2022
Autonomous surface vehicle (ASV) is a vehicle in the form of an unmanned
on-water surface vessel that can move automatically. As such, an automatic
control system is essentially required. The bow thruster system functions as
a propulsion control device in its operations. In this research, the water
momentum and propeller velocity were estimated based on the dynamic bow
thruster model. The estimation methods used is the Kalman filter (KF) and
ensemble Kalman filter (EnKF). There are two scenarios: tunnel thruster
condition and open-bladed thruster condition. The estimation results in the
tunnel thruster condition showed that the root mean square error (RMSE) by
the EnKF method was relatively smaller, that is, 0.7920 and 0.1352, while
the estimation results in the open-bladed thruster condition showed that the
RMSE by the KF method was relatively smaller, that is, 1.9957 and 2.0609.
Keywords:
Autonomous surface vehicle
Bow thruster
Ensemble Kalman filter
Kalman filter
This is an open access article under the CC BY-SA license.
Corresponding Author:
Teguh Herlambang
Department of Information Systems, Universitas Nahdlatul Ulama Surabaya
Jl. Raya Jemursari 51-57, Surabaya, Indonesia
Email: teguh@unusa.ac.id
1. INTRODUCTION
Indonesia is an archipelago country consisting of 17,508 islands, with sea area of about two-thirds
of its territory and consisting of several main island groups [1]. This can provide income opportunities for the
country, especially in the marine tourism sector. Along with the rapid development of modern technology in
various fields, it also has an impact on the development of marine transportation, namely unmanned water
surface vehicles that can move automatically, i.e., autonomous surface vehicle (ASV) or unmanned surface
vehicle (USV). USV is controlled automatically by commands such as waypoints [2]. ASV can be used either
as a research or survey vessel for river or lake area inspection, seismic survey, rescue operation, and others.
The use of ASV as a research vessel has been carried out in several countries, most of which carry out
research in either rivers or offshore automatically.
In the transportation sector, especially the marine transportation sector, a ship is required to work
optimally. One way to support the smooth operation of a ship voyage requires a supporting device to support
the ship when maneuvering, a bow thruster as a propeller installed on the ship bow. Ship maneuvering is the
ship’s ability to turn and turn around when the ship is about to dock or set off the port. This ability greatly
determines the safety of the ship, especially when the ship operates in confined waters or operates around the
port. The bow thruster installation can also increase the maneuverability of a ship. By utilizing the rotational
energy of the propeller in the tunnel thruster of a ship, the direction of the ship can be turned faster than a
ship without a bow thruster. By relying on the bow thruster’s ability, it can be developed by adding an

Int J Elec & Comp Eng ISSN: 2088-8708 
Estimation of water momentum and propeller velocity in bow thruster model … (Hendro Nurhadi)
5989
additional part to the tunnel thruster. It works by providing a cover on the bow thruster that can be opened
and closed. The purpose of this tunnel cover is not only for opening and closing, but rather leads to an
increase in the maneuverability of ships utilizing the bow thruster [3].
During the ship voyage, maneuvering process can be interrupted if a bow thruster does not operate
properly. It is undeniable that (noise) interference may come from the bow thruster system itself. Disturbance
can occur when the bow thruster functioning to smooth motion is damaged, so that the operating system of
the bow thruster, which provides a transverse thrust on the bow of the ship is disturbed [4]. To overcome any
disturbance to the ship voyage, its control system is equipped with an estimator. The estimator is used to
provide predictions for the variables on the ship due to the disturbance that occurs. One of the algorithms for
estimating a state system of a dynamic model was introduced by Kalman [5]. This algorithm is called the
Kalman filter (KF), which is an algorithm that can be implemented in a stochastic linear dynamic model.
In the previous research, researches on estimation have been carried out. The study [6] conducted a
research with the aim of designing a KF estimator on noise conditions by measuring instruments, noise by
ship systems and inaccuracy in modeling. The ship dynamic variables estimated for steering purposes are
sway-yaw dynamics with variables of angular velocity, angular position, and sway direction velocity. The
results of the application of the KF algorithm are the estimated values of the three dynamic variables of the
ship with the absolute integral percentage of error of the system-on-system noise and measurement noise.
Ataei and Koma [7] investigated the navigation and guidance control system of autonomous underwater
vehicle (AUV). Then Miller et al. [8] discussed estimation and control of AUV by using acoustic. In 2018,
Wang et al. [9] described estimation of steam temperature in drum boiler. Then Schoniger et al. [10]
estimated parameter of hydraulic tomography using ensemble Kalman filter (EnKF). In Nurhadi et al. [11]
conducted a research related to the estimation of ASV position and motion due to the influence of wind speed
and wave height by applying the EnKF. The results of the application of the EnKF algorithm showed the
smallest position error and a high degree of accuracy [11]. Then [12], [13] used EnKF in blood transfusion
management and crude oil price estimation, respectively. Recently, studies [14]–[20] discussed the
application of KF in pneumatic artificial muscles, mobile robot, real-time RSSI based outdoor target tracking
and autonomous underwater vehicle. Regarding researches on the bow thruster modeling, there are numerous
references in the literature, such as [21]–[27]. In this research, we use the model proposed by Healey
et al. [21] which produces a motion control system on the thruster, namely water momentum and propeller
velocity which is a dynamic thruster model.
To the best of our knowledge, the effect of noise on water momentum and propeller velocity using a
dynamic thruster model was not yet studied in the literature. That motivates the current research of the authors.
The main contribution of this paper is a numerical analysis on the comparison between the KF method and the
EnKF method for estimating the water momentum and propeller velocity on a bow thruster autonomous surface
vehicle (ASV). We compare the performance of KF and EnKF because EnKF is an extension of KF which can
be used to estimate linear and nonlinear models by generating some ensembles. In this paper, first we linearize
the bow thruster model. Then, we analyze the stability of the linearized model. After that, we discretize the
model by using the zero-order hold method. Next, we implement the KF and the EnKF to the linearized model.
Finally, we conduct some simulations and analyze the simulation results.
2. MODELS AND PRELIMINARIES
2.1. Bow thruster model of autonomous surface vehicle
The bow thruster model that was proposed in [9] is a continuous-time nonlinear model. The model
has two state variables, i.e., motor rotational rate 𝜔𝑚 and section average flow velocity 𝑈𝑎. The model has
two input variables, i.e., voltage source 𝑉
𝑠 and vehicle velocity 𝑈0. There is one output variable in the model,
i.e., thrust force 𝑇. The state equations in the bow thruster model are:
𝜔̇𝑚 = 𝑓1(𝜔𝑚, 𝑈𝑎, 𝑉
𝑠, 𝑈0) = −𝐾1𝜔𝑚 + 𝐾2𝑉
𝑠 − 𝐾ℎ𝑄
𝑈̇𝑎 = 𝑓2(𝜔𝑚, 𝑈𝑎, 𝑉
𝑠, 𝑈0) = −𝐾4𝐾3
−1
𝑈
̅𝑎|𝑈
̅𝑎| + 𝐾3
−1
𝑇
and the output equation is given by
𝑇 = 𝑔(𝜔𝑚, 𝑈𝑎, 𝑉
𝑠, 𝑈0) = 𝐿𝑖𝑓𝑡(cos 𝜃) − 𝐷𝑟𝑎𝑔(sin 𝜃)
where 𝐿𝑖𝑓𝑡 represents the lift force, 𝐷𝑟𝑎𝑔 represents the drag force and 𝜃 represents the angle of inlet to
blades. In [28], we have linearized the bow thruster model by using the parameters for tunnel thruster test and
open bladed thruster test.
When the parameters for tunnel thruster test are used, we obtain the following linear system [28]:

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 6, December 2022: 5988-5997
5990
(
𝜔̇𝑚
𝑈̇𝑎
) = (
−70.7 1.2
−2.1 −14.8
) (
𝜔𝑚
𝑈𝑎
) + (
1133.2 0
0 1.9
) (
𝑉
𝑠
𝑈0
)
𝑇 = (−2.03 −12.2) (
𝜔𝑚
𝑈𝑎
) + (0 0) (
𝑉
𝑠
𝑈0
)
Next, we check the stability of the linear system by computing the eigenvalues. The eigenvalues of the state
matrix are 𝜆1 = −70.6549 and 𝜆2 = −14.8451. Since the real parts of all eigenvalues are negative, the
linear system is asymptotically stable. Furthermore, the linear system is observable because the rank of
observability matrix is 2.
If we use the parameters for open-bladed thruster test, we obtain the following linear system [28]:
(
𝜔̇𝑚
𝑈̇𝑎
) = (
−523.7 −1519.8
−1.15 −0.13
) (
𝜔𝑚
𝑈𝑎
) + (
0.065 0
0 2.73
) (
𝑉
𝑠
𝑈0
)
𝑇 = (−4.62 −10.4) (
𝜔𝑚
𝑈𝑎
) + (0 0) (
𝑉
𝑠
𝑈0
)
Then, as before, we determine the stability of the linear system. The eigenvalues of the system matrix are
𝜆1 = −527.0172 and 𝜆2 = 31872. The linear system is unstable because there exists an eigenvalue where
the real part is positive. Moreover, this linear system is observable because the rank of observability matrix
equals 2.
2.2. Kalman filter algorithm implementation
In this section, we discuss the KF algorithm. In the next section, the algorithm will be applied to the
linearized bow thruster model. As mentioned before, the KF algorithm can be applied to discrete-time
systems. As such, the model needs to be discretized first. The steps of the KF algorithm were as [29], [30]:
− Determine the system model and measurement model. The general form of system model was represented
as:
𝑥𝑘+1 = 𝐴𝑥𝑘 + 𝐵𝑢𝑘 + 𝐺𝑤𝑘
where 𝑥𝑘 is the state at time 𝑘, 𝑢𝑘 is the input at time 𝑘 and 𝑤𝑘 is the system noise at time 𝑘. We assume
that the system noise at time 𝑘 is normally distributed with mean 0 and variance 𝑄𝑘, i.e., 𝑤𝑘 ∼ 𝑁(0, 𝑄𝑘).
The general form of measurement model was represented as:
𝑧𝑘 = 𝐻𝑥𝑘 + 𝑣𝑘
where 𝑧𝑘 is the measurement at time 𝑘 and 𝑣𝑘 is the measurement noise at time 𝑘. We assume that the
measurement noise at time 𝑘 is normally distributed with mean 0 and variance 𝑅𝑘, i.e., 𝑣𝑘 ∼ 𝑁(0, 𝑅𝑘).
− Initialization stage. Determine the initial state, the initial covariance for system noise and the initial
covariance of measurement noise. The estimated initial state 𝑥
̂0 was generated by a normally distributed
random variable with mean 𝑥̅0 and covariance 𝑃0.
− Time update. After the system model became a discrete-time linear system, then the estimation and
covariance of the estimation could be calculated using the following equation:
State estimation: 𝑥̅𝑘+1
−
= 𝐴𝑘𝑥̅𝑘 + 𝐵𝑘𝑢𝑘
Covariance of estimation: 𝑃𝑘+1
−
= 𝐴𝑘𝑃𝑘𝐴𝑘
𝑇
+ 𝐺𝑘𝑄𝑘𝐺𝑘
𝑇
Notation 𝑥
̂𝑘+1
−
represents the estimation of state at time 𝑘 + 1 before receiving the measurement data. The
covariance of the estimation is denoted by 𝑃𝑘+1
−
.
− Measurement update. After receiving the measurement data, we find out the Kalman gain, the updated
estimation and updated covariance of estimation:
Kalman gain: 𝐾𝑘 = 𝑃𝑘+1
− [𝐻𝑘+1𝑃𝑘+1
−
𝐻𝑘+1
𝑇
+ 𝑅𝑘+1]−1
Update covariance of estimation: 𝑃𝑘+1 = [(𝑃𝑘+1
− )−1
+ 𝐻𝑘+1
𝑇
𝑅𝑘+1
−1
𝐻𝑘+1]−1

5991
Update state of estimation: 𝑥
̂𝑘+1 = 𝑥
̂𝑘+1
−
+ 𝑃𝑘+1𝐻𝑘+1
𝑇
𝑅𝑘+1
−1
(𝑧𝑘+1 − 𝐻𝑘+1𝑥
̂𝑘+1
−
)
Notation 𝑥
̂𝑘+1 denotes the estimation of state at time 𝑘 + 1 after receiving the measurement data. The
covariance of the estimation is denoted by 𝑃𝑘+1.
− Once the measurement update was finished, the time update is executed again. The time update and
measurement update are executed until all measurement data are processed.
2.3. Ensemble Kalman filter algorithm implementation
In this section, we describe the EnKF algorithm [2], [31], [32]. The algorithm was applied to the
linearized bow thruster model. Before the EnKF algorithm was applied, the continuous-time model was
discretized first. The steps in the EnKF method were as:
− Determine the system model and measurement model. The system and measurement model for EnKF is
the same with the system and measurement model for KF.
− Initialization stage. Generate 𝑁 ensemble [𝑥0,1 𝑥0,2 … 𝑥0,𝑁] in accordance with the initial estimation,
where 𝑥0,𝑖 for 𝑖 = 1, … , 𝑁 is generated from a normal distribution with mean 𝑥0 and variance 𝑃0.
Determine the mean of the generated ensemble: 𝑥0 =
1
𝑁
∑ 𝑥0,𝑖
𝑁
𝑖=1
− Time update. In this stage, efforts were made to determine the estimation of ensemble, mean, and error
covariance at the next time step.
Estimation of ensemble: 𝑥
̂𝑘,𝑖
−
= 𝐴𝑥
̂𝑘,𝑖 + 𝐵𝑢𝑘−1 + 𝑤𝑘,𝑖 for 𝑖 = 1, ⋯ , 𝑁, where 𝑥0,𝑖 for 𝑖 = 1, ⋯ , 𝑁 is
generated from a normal distribution with mean 𝑥0 and variance 𝑄𝑘.
Mean of ensemble estimation: 𝑥
̂𝑘
−
=
1
𝑁
∑ 𝑥
̂𝑘,𝑖
−
𝑁
𝑖=1
Covariance of ensemble estimation: 𝑃𝑘
−
=
1
𝑁−1
∑ (𝑥
̂𝑘,𝑖
−
− 𝑥
̂𝑘
−
)
𝑁
𝑖=1 (𝑥
̂𝑘,𝑖
−
− 𝑥
̂𝑘
−
)
𝑇
− Measurement update. At this stage, the correction was done by generating N ensemble on measurement
data to determine the estimation of ensemble, Kalman gain, mean, and error covariance. Estimation of
ensemble measurement: 𝑧𝑘,𝑖 = 𝑧𝑘 + 𝑣𝑘,𝑖 for 𝑥0,𝑖𝑖 = 1, ⋯ , 𝑁 where 𝑣𝑘,𝑖 is generated from a normal
distribution with mean 0 and variance 𝑅𝑘.
Covariance between state and measurement: 𝑃𝑥𝑧 =
1
𝑁−1
∑ (𝑥
̂𝑘,𝑖
−
− 𝑥
̂𝑘
−
)
𝑁
𝑖=1 (𝑧𝑘,𝑖
−
− 𝑧𝑘
−
)
𝑇
Covariance of measurements: 𝑃𝑧 =
1
𝑁−1
∑ (𝑧𝑘,𝑖
−
− 𝑧𝑘
−
)
𝑁
𝑘=1 (𝑧𝑘,𝑖
−
− 𝑧𝑘
−
)
𝑇
Kalman gain: 𝐾𝑘 = 𝑃𝑥𝑧(𝑃𝑧)−1
Update state estimation: 𝑥
̂𝑘+1 = 𝑥
̂𝑘+1
−
+ 𝐾𝑘(𝑧𝑘,𝑖
−
− 𝐻𝑥
̂𝑘+1
−
)
− Once the measurement update was finished, the process continues to the time update for the next time
step. This process is repeated until all measurement data has been processed.
3. SIMULATION AND ANALYSIS RESULTS
In this section, we apply the KF and EnKF to the linearized bow thruster model. In each simulation,
we define the initial state, i.e., the initial propeller velocity 𝜔𝑚(0) and the initial water momentum 𝑈𝑎(0).
The covariance of system noise for all time is 0.5, i.e., 𝑄𝑘 = 0.5, for all 𝑘. The covariance of measurement
noise for all time is also 0.5, i.e., 𝑅𝑘 = 0.5 for all 𝑘.
3.1. Simulation I
In the first simulation, the parameters for tunnel thruster test are used. In this case, the duration of
the simulation is 𝑘 = 100 steps. In this simulation, we compare the real values and the estimation results
obtained by using KF and EnKF. The comparison is based on the root mean square error (RMSE). Such
comparison was shown to find out the better method to estimate the water momentum and propeller velocity
in the linearized bow thruster model of ASV. Based on Figure 1 (a) the RMSE for estimating the propeller
velocity using the KF method is 0.8678, whereas the RMSE for estimation result using the EnKF method is

 ISSN: 2088-8708
5992
0.8051. Figure 1 (b) shows that the RMSE for the water momentum estimation by the KF method is 0.8717,
whereas the RMSE for estimation result using the EnKF method is 0.1344. Based on the simulation I result,
the conclusion is that the EnKF method is more accurate than the KF, when we use the parameters for tunnel
thruster test.
(a) (b)
Figure 1. Estimation results of (a) propeller velocity (𝜔𝑚) and (b) water momentum (𝑈𝑎)
by tunnel thruster test
3.2. Simulation II
In simulation II, the simulation was carried out based on the open-bladed thruster test with a value of
𝑘 = 100 steps. Simulation II is carried out by comparing the RMSE between the real value and both the KF
and EnKF estimation results. Such results were shown to determine the better method for estimating the
water momentum and propeller velocity in the linearized ASV bow thruster model. The results of
simulation II is displayed in Figure 2(a) and (b).
(a) (b)
Figure 2. Estimation results of propeller velocity (𝜔𝑚) and water momentum (𝑈𝑎) by open-bladed thruster test
Based on Figure 2 (a) the RMSE value for the propeller velocity estimation result using the KF
method is 1.6749, whereas the RMSE value for that using the EnKF method is 4.0858. Figure 2 (b) shows
that the RMSE value for the water momentum estimation using the KF method is 1.6820, whereas the RMSE
0 10 20 30 40 50 60 70 80 90 100
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Estimation of Propeller Velocity
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-5
-4
-3
-2
-1
0
1
2
Estimation of Water Momentum
Iteration
Water
Momentum
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-100
-80
-60
-40
-20
0
20
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-10
-5
0
5
10
15
20
25
30
35
Iteration
Water
Momentum
Real
EnKF Method
KF Method

5993
value for that using the EnKF method is 0.6795. Based on the estimation results of Simulation II, it could be
concluded that the KF method has higher accuracy than the EnKF method, when the parameters for open-
bladed thruster test are used.
3.3. Simulation III
In simulation III, we observe the effect of noise covariance to the accuracy of estimation results by
using KF and EnKF, both for parameters of tunnel thruster test and open-bladed thruster test. In all
simulations, the number of iterations is 𝑘 = 100 steps. In the first case, we try the following covariance of
noises 0.6, 0.8 and 1. Then for each covariance of noises, we implemented the KF and EnKF to the linearized
bow thruster model where the parameters are tunnel thruster test. The different noise covariance values were
expected to affect the estimation results in each method.
3.3.1. Variety of noise covariance values by tunnel thruster test
According to results in simulation III in Figures 3 to 5, it is shown that the higher noise covariance
value would affect the results of estimation. The values of RMSE are represented in Table 1. According to
Table 1, EnKF has a better performance than KF in almost all cases. There is only one case where KF has a
better performance than EnKF, i.e., the estimation of propeller velocity (𝜔𝑚) when the noise covariance
equals 1.
Figure 3. Estimation of propeller velocity (𝜔𝑚) and water momentum (𝑈𝑎) by tunnel thruster test with
covariance of noise equals 0.6
covariance of noise equals 0.8
0 10 20 30 40 50 60 70 80 90 100
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-4
-3
-2
-1
0
1
2
3
4
Iteration
Water
Momentum
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-2
-1
0
1
2
3
4
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-3
-2
-1
0
1
2
3
4
5
Iteration
Water
Momentum
Real
EnKF Method
KF Method

 ISSN: 2088-8708
5994
covariance of noise equals 1
Table 1. RMSE values 𝜔𝑚 and 𝑈𝑎 for some covariance of noise by tunnel thruster test
0.6 0.8 1
𝜔𝑚 𝑈𝑎 𝜔𝑚 𝑈𝑎 𝜔𝑚 𝑈𝑎
KF 1.0751 1.0804 1,1638 1,1598 1.1969 1.8830
ENKF 0.8051 0.1344 0.9951 0.1659 1.2049 0.2017
3.3.2. Variation of noise covariance value by open-bladed thruster test
In this case, we vary the noise covariance values in 0.6, 0.8 and 1. For each noise covariance, we
implement the KF and EnKF to the linearized bow thruster model where the parameters are the open-bladed
thruster test. It was the condition under which the different noise covariance values was expected to affect the
estimation results of each method. Based on the results in simulation III in Figure 6-8, it was shown that the
higher noise covariance value affected the estimation results. The RMSE values are presented in Table 2.
From Table 2, we can conclude that KF has a better performance in the estimation of propeller velocity
(𝜔𝑚), whereas EnKF has a better performance in the estimation of water momentum (𝑈𝑎).
Figure 6. Estimation of propeller velocity (𝜔𝑚) and water momentum (𝑈𝑎) by open-bladed thruster test
where the covariance of noise is 0.6
0 10 20 30 40 50 60 70 80 90 100
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-4
-3
-2
-1
0
1
2
3
4
Iteration
Water
Momentum
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-350
-300
-250
-200
-150
-100
-50
0
50
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-20
0
20
40
60
80
100
120
Iteration
Water
Momentum
Real
EnKF Method
KF Method

5995
where the covariance of noise is 0.8
where the covariance of noise equals 1
Table 2. RMSE values 𝜔𝑚 and 𝑈𝑎 for some covariance of noise by open-bladed thruster test
0.6 0.8 1
𝜔𝑚 𝑈𝑎 𝜔𝑚 𝑈𝑎 𝜔𝑚 𝑈𝑎
KF 1.7762 1.7748 1,8369 1,8449 2.2645 2.2752
ENKF 4.1879 0.6980 5.0252 0.8365 5.8559 0.9749
4. CONCLUSION
Based on the research results, the estimation results by the tunnel thruster test for the propeller
velocity (𝜔𝑚) and the water momentum (𝑈𝑎) was more accurate when we use the EnKF method due to the
relatively lower RMSE value, whereas in open-bladed thruster test for the estimation of propeller velocity
(𝜔𝑚) and water momentum (𝑈𝑎), KF was more accurate than EnKF.
REFERENCES
[1] H. Nurhadi, E. Apriliani, T. Herlambang, and D. Adzkiya, “Sliding mode control design for autonomous surface vehicle motion
under the influence of environmental factor,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10,
no. 5, pp. 4789–4797, Oct. 2020, doi: 10.11591/ijece.v10i5.pp4789-4797.
[2] H. Nurhadi, T. Herlambang, and D. Adzkiya, “Trajectory estimation of autonomous surface vehicle using square root ensemble
Kalman Filter,” in 2019 International Conference on Advanced Mechatronics, Intelligent Manufacture and Industrial Automation
0 10 20 30 40 50 60 70 80 90 100
-300
-250
-200
-150
-100
-50
0
50
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
0
10
20
30
40
50
60
70
80
90
Iteration
Water
Momentum
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-140
-120
-100
-80
-60
-40
-20
0
20
Iteration
Propeller
Velocity
Real
EnKF Method
KF Method
0 10 20 30 40 50 60 70 80 90 100
-10
0
10
20
30
40
50
Iteration
Water
Momentum
Real
EnKF Method
KF Method

 ISSN: 2088-8708
5996
(ICAMIMIA), Oct. 2019, pp. 325–328, doi: 10.1109/ICAMIMIA47173.2019.9223354.
[3] D. A. Taylor, Introduction to marine engineering. Elsevier, 1983.
[4] T. I. Fossen, “Guidance and control of ocean vehicles,” University of Trondheim, Norway, 1999.
[5] R. E. Kalman, “A new approach to linear filtering and prediction problems,” Journal of Basic Engineering, vol. 82, no. 1,
pp. 35–45, Mar. 1960, doi: 10.1115/1.3662552.
[6] F. L. Lewis, Optimal estimation, with an introduction to stochastic control theory. John Wiley and Sons, 1986.
[7] M. Ataei and A. Yousefi-Koma, “Three-dimensional optimal path planning for waypoint guidance of an autonomous underwater
vehicle,” Robotics and Autonomous Systems Journal, vol. 67, pp. 23–32, 2015, doi: 10.1016/j.robot.2014.10.007.
[8] A. Miller, B. Miller, and G. Miller, “AUV position estimation via acoustic seabed profile measurements,” 2018 IEEE/OES
Autonomous Underwater Vehicle Workshop (AUV), pp. 1–8, 2018.
[9] X. Wang, G. Wang, H. Chen, and L. Zhang, “Real-time temperature field reconstruction of boiler drum based on fuzzy adaptive
Kalman filter and order reduction,” International Journal of Thermal Sciences, vol. 113, pp. 145–153, 2017, doi:
10.1016/j.ijthermalsci.2016.11.017.
[10] A. Schöniger, W. Nowak, and H.-J. H. Franssen, “Parameter estimation by ensemble Kalman filters with transformed data:
Approach and application to hydraulic tomography,” Water Resources Research, vol. 48, no. 4, Apr. 2012, doi:
10.1029/2011WR010462.
[11] H. Nurhadi, T. Herlambang, and D. Adzkiya, “Position estimation of touristant ASV using ensemble Kalman filter,” AIP
Conference Proceedings 2187, 2019, doi: 10.1063/1.5138367.
[12] W. H. Shanty, Firdaus, and T. Herlambang, “Prediction of availability of packed red cells (prc) at pmi surabaya city using
ensemble Kalman filter as management of blood transfusion management,” Journal of Physics: Conference Series, vol. 1211,
no. 1, 2019.
[13] P. Katias, D. F. Karya, T. Herlambang, and H. Khusnah, “Ensemble kalman filter for crude oil price estimation,” Journal of
Physics: Conference Series, vol. 1211, Apr. 2019, doi: 10.1088/1742-6596/1211/1/012032.
[14] T. Kodama and K. Kogiso, “Applications of UKF and EnKF to estimation of contraction ratio of McKibben pneumatic artificial
muscles,” in 2017 American Control Conference (ACC), May 2017, pp. 5217–5222, doi: 10.23919/ACC.2017.7963765.
[15] O. Bayasli and H. Salhi, “The cubic root unscented kalman filter to estimate the position and orientation of mobile robot
trajectory,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 5, pp. 5243–5250, Oct. 2020, doi:
10.11591/ijece.v10i5.pp5243-5250.
[16] K. Vadivukkarasi and R. Kumar, “Investigations on real time RSSI based outdoor target tracking using kalman filter in wireless
sensor networks,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 2, pp. 1943–1951, Apr.
2020, doi: 10.11591/ijece.v10i2.pp1943-1951.
[17] B. He, H. Zhang, C. Li, S. Zhang, Y. Liang, and T. Yan, “Autonomous navigation for autonomous underwater vehicles based on
information filters and active sensing,” Sensors, vol. 11, no. 11, pp. 10958–10980, Nov. 2011, doi: 10.3390/s111110958.
[18] Y. Bai, X. Wang, X. Jin, Z. Zhao, and B. Zhang, “A neuron-based Kalman Filter with nonlinear autoregressive model,” Sensors,
vol. 20, no. 1, Jan. 2020, doi: 10.3390/s20010299.
[19] M. F. M. Jusof, A. N. K. Nasir, M. A. Ahmad, and Z. Ibrahim, “An exponential based simulated Kalman filter algorithm for data-
driven PID tuning in liquid slosh controller,” in 2018 IEEE International Conference on Applied System Invention (ICASI), Apr.
2018, pp. 984–987, doi: 10.1109/ICASI.2018.8394437.
[20] A. Azwan, A. Razak, M. F. M. Jusof, A. N. K. Nasir, and M. A. Ahmad, “A multiobjective simulated Kalman filter optimization
algorithm,” in 2018 IEEE International Conference on Applied System Invention (ICASI), Apr. 2018, pp. 23–26, doi:
10.1109/ICASI.2018.8394257.
[21] A. J. Healey, S. M. Rock, S. Cody, D. Miles, and J. P. Brown, “Toward an improved understanding of thruster dynamics for
underwater vehicles,” in Proceedings of IEEE Symposium on Autonomous Underwater Vehicle Technology (AUV’94), 1995,
pp. 340–352, doi: 10.1109/AUV.1994.518646.
[22] J. Szabo, “Fully kinetic numerical modeling of a plasma thruster,” Massachusetts Institute of Technology, 2001.
[23] I. D. Boyd, “Review of hall thruster plume modeling,” Journal of Spacecraft and Rockets, vol. 38, no. 3, pp. 381–387, May 2001,
doi: 10.2514/2.3695.
[24] I. D. Boyd and R. A. Dressler, “Far field modeling of the plasma plume of a Hall thruster,” Journal of Applied Physics, vol. 92,
no. 4, pp. 1764–1774, Aug. 2002, doi: 10.1063/1.1492014.
[25] I. D. Boyd and J. T. Yim, “Modeling of the near field plume of a Hall thruster,” Journal of Applied Physics, vol. 95, no. 9,
pp. 4575–4584, May 2004, doi: 10.1063/1.1688444.
[26] J. Kim and W. K. Chung, “Accurate and practical thruster modeling for underwater vehicles,” Ocean Engineering, vol. 33,
no. 5–6, pp. 566–586, Apr. 2006, doi: 10.1016/j.oceaneng.2005.07.008.
[27] P. G. Mikellides and J. K. Villarreal, “High energy pulsed inductive thruster modeling operating with ammonia propellant,”
Journal of Applied Physics, vol. 102, no. 10, Nov. 2007, doi: 10.1063/1.2809436.
[28] D. Adzkiya, H. Nurhadi, and T. Herlambang, “Linearization of two-state thruster models,” Journal of Physics: Conference Series,
vol. 1490, no. 1, 2020.
[29] R. Garcia, J. Puig, P. Ridao, and X. Cufi, “Augmented state Kalman filtering for AUV navigation,” in Proceedings 2002 IEEE
International Conference on Robotics and Automation (Cat. No.02CH37292), 2002, vol. 4, pp. 4010–4015, doi:
10.1109/ROBOT.2002.1014362.
[30] B. Armstrong, E. Wolbrecht, and D. B. Edwards, “AUV navigation in the presence of a magnetic disturbance with an extended
Kalman filter,” in OCEANS’10 IEEE SYDNEY, May 2010, pp. 1–6, doi: 10.1109/OCEANSSYD.2010.5603905.
[31] G. Evensen, Data asimilation the ensemble Kalman filter. Springer Berlin Heidelberg, 2009.
[32] E. Purnaningrum and E. Apriliani, “Auto floodgate control using EnKf-NMPC method,” International Journal of Computing
Science and Applied Mathematics, vol. 2, no. 1, Mar. 2016, doi: 10.12962/j24775401.v2i1.1579.

5997
BIOGRAPHIES OF AUTHORS
Hendro Nurhadi received the Dipl. Ing. (FH) degree from the University of
Applied Science Georg-Simon-Ohm Fachhochshule Nuremberg, Nuremberg, Germany, in
2001 and the Ph.D. degree from the National Taiwan University of Science and Technology
(NTUST), Taipei, Taiwan, in 2009. He is currently with the Department of Mechanical
Engineering at Institute of Technology Sepuluh Nopember (ITS), Surabaya, Indonesia, in
charged as a Head of Mechatronics Laboratory, also assigned as a coordinator of national
consortium for mechatronics for defense, unmanned systems and industrial machineries. He
also assigned as researcher in Center of Excellence for Mechatronics and Industrial
Automation (PUI-PT MIA-RC ITS) Kemenristekdikti. He has authored numerous international
journal papers and international conferences, as well as reviewer and editor for various
international journal papers and international proceedings. His research interests and
consulting activities are in the areas of control system, robotics and automation, advanced
mechatronics, automated optical inspection (AOI), machine tools, dynamic systems,
automation of manufacturing processes, computer-aided design and manufacturing,
optimization applications, digital signal processing, artificial intelligence, and related fields.
He can be contacted at email: hdnurhadi@me.its.ac.id.
Mayga Kiki holds a BSc in mathematics from Institut Teknologi Sepuluh
Nopember, Indonesia. She is currently Office Development Program at PT. Bank Syariah
Indonesia, Tbk. She can be contacted at email: maygakiki2@gmail.com.
Dieky Adzkiya is an Assistant Professor in the Department of Mathematics and a
member of Mechatronics and Industrial Automation Research Center, both at Institut
Teknologi Sepuluh Nopember, Indonesia. He received the B.Sc. degree in September 2005 and
the M.Sc. degree in October 2008, both in Mathematics from the Institut Teknologi Sepuluh
Nopember, Surabaya, Indonesia. He received the Ph.D. degree in Systems and Control in
October 2014 and after that he continued as a postdoctoral researcher until June 2015, both at
the Delft Center for Systems and Control, Delft University of Technology, Delft, The
Netherlands. His research interests are in the analysis and verification of max-plus-linear
systems and in their applications. He can be contacted at email: dieky@matematika.its.ac.id.
Teguh Herlambang is currently Lecturer at the Department of Information
System, Nahdlatul Ulama Surabaya University (UNUSA) Surabaya, Indonesia. He received
his B.Sc and M.Sc. degree from Department of Mathematics at Institute of Technology
Sepuluh Nopember (ITS) in 2010 and 2012. He received his Ph.D. degree from Department of
Ocean Engineering at Institute of Technology Sepuluh Nopember (ITS). He is currently is a
Head of Research Department of FEBTD UNUSA, also assigned as researcher in Center of
Excellence for Mechatronics and Industrial Automation (PUI-PT MIA-RC ITS)
Kemenristekdikti. His area of interest is modelling, navigation, guidance and control of
dynamics system. He can be contacted at email: teguh@unusa.ac.id.

Estimation of water momentum and propeller velocity in bow thruster model of autonomous surface vehicle using modified Kalman filter

More Related Content

Similar to Estimation of water momentum and propeller velocity in bow thruster model of autonomous surface vehicle using modified Kalman filter (20)

More from IJECEIAES (20)

Recently uploaded (20)

Estimation of water momentum and propeller velocity in bow thruster model of autonomous surface vehicle using modified Kalman filter