This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access

Received January 14, 2020, accepted February 9, 2020, Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2020.Doi Number

Application of Wireless Network Control to

Course-Keeping for Ships
1,2 1
RIFAN JIANG AND XIANKU ZHANG
1
Navigation College, Dalian Maritime University, Dalian 116026, China
2
Network Information Center, Dalian Polytechnic University, Dalian 116034, China
Corresponding author: Xianku Zhang (zhangxk@dlmu.edu.cn)
This work is partially supported by the National Science Foundation of China (Grant No.51679024), the Fundamental Research Funds
for the Central Universities (Grand No.3132016315), and the University 211 Project of China (Grant No.B08046).

ABSTRACT In view of the network security and reliability of the steering control system, redundant wired
control networks are adopted in some ships. To solve the problems of redundant wired networks in ships,
the paper designs a wireless network as a redundant network for a ship’s control system. The paper applies
wireless network control to course-keeping for ships in order to validate the effectiveness of the proposed
ship wireless communication network. However, there are some problems, such as time delay and packet
loss, in the wireless network control system (WiNCS). In view of the above problems, the codesign of a
communication protocol and control algorithm is studied in the paper. From the aspect of communication,
the design is based on the concurrent and disjoint multipath routing ZigBee network, which can ensure the
effective application of the control algorithm. From the aspect of control, the design of the ship course-
keeping controller combines an improved gray model with the nonlinear robust controller based on the
nonlinear decoration and integrator backstepping method, which can compensate for time delay and packet
loss. In this paper, a simulation platform is designed to study the application of a wireless network control
to course-keeping for ships. The experimental results also prove that the scheme is feasible and effective on
a simulation platform. It is a beneficial attempt to apply the WiNCS to course-keeping for ships.

INDEX TERMS Ships, course-keeping control, wireless network control system (WiNCS),
codesign of communication and control, closed-loop gain shaping, backstepping, gray model.

I. INTRODUCTION monitoring system based on wireless network sensors

Ships’ wired networks face problems, such as the difficulties
in ship wiring and fitting because of the confined spaces,
high cost of wired cable, and harsh conditions of the engine
room that make wired cable easily damaged. In addition, the
informatized and electronic equipment are more portable,
more mobile, and even smaller than the original equipment.
Therefore, various micro intelligent terminals and sensors,
instead of the original equipment play important roles in
various aspects of ship functioning because in intelligent
terminals, sensors can be installed flexibly, operated easily
and cheaply. However, the wired network places great
restrictions on the mobility and flexibility of the equipment.
To solve the above problems, a wireless network is adopted
to replace the wired network in a ship because it can not only
achieve the on-demand deployment of a variety of intelligent
terminals and sensors but also solve the problems associated
with a wired network. Recently, the wireless network has
mainly applied to ship monitoring and communication,
identification, positioning, etc. In [1] and [2], a ship
collected data through sensor nodes and then sent them to the
upper computer. The test results proved that the ship
monitoring system based on wireless network can carry out
real-time, continuous and reliable monitoring. In [3], an
automatic identification system based on a wireless sensor
network was used for ship positioning. In [4], a wireless
sensor network was used for communication and positioning
in ship networking, and the positioning algorithm in ZigBee
protocol was studied.
The steering control system is a navaid for controlling the
steering engine according to the command signal and
navigating a ship on the given course or track. In the system,
real-time data are transmitted between the two pieces of
equipment through the field bus network. In view of the
network security and reliability of the steering control system,
a redundant network needs to be adopted in the ship (for
example, dual redundant CAN bus). However, there are no
redundant control networks in some ships. Moreover, it is
difficult to install a redundant wired network in the confined

VOLUME XX, 2020 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships
space of a ship, and the cost is high. To solve these problems,
a redundant wireless network is adopted to replace the
redundant wired network in the paper, which forms a hybrid
network including wired network and redundant wireless
network in the steering control system.
The feedback control system is formed by a wireless
network, which is called a wireless network control system
(WiNCS). If a wireless network is applied to the steering
control system with high reliability and high real-time
character, problems such as time delay and packet loss must
be solved effectively in WiNCS. Therefore, the codesign of a
communication protocol and control algorithm is applied to
the control of ship course-keeping. Research about ship
course-keeping control has been conducted on an improved
simulation platform [5]. From the aspect of communication,
the design is based on the concurrent and disjoint multipath
routing ZigBee network, which ensures the reliability of data
transmission and effective implementation of the ship course-
keeping control algorithm. From the aspect of control, the
design of the ship course-keeping controller adopts a gray
model to make a real-time prediction of time delay and ship
course deviation. The predictive value is applied to the
nonlinear robust controller, which can compensate for time
delay and packet loss and provide appropriate control quality.
This paper provides four main contributions. First, a
wireless network is used as the redundant network of a ship
control system. In addition, wireless network control is
applied to ship course-keeping. Second, the collaborative
design of communication protocol and control algorithm is
studied to solve problems such as time delay and packet loss
in WiNCS. Third, the ship course-keeping controller is
designed to combine the improved gray model with the
nonlinear robust controller based on nonlinear decoration and
integrator backstepping, which compensate for time delay
and packet loss. Finally, a simulation platform based on the
concurrent and disjoint multipath routing ZigBee network is
designed to study the application of wireless network control
to course-keeping for ships. ZigBee nodes run on the
Mindows operating system and use the coulombmeter
method to monitor the battery level.
II. DESIGN OF MULTIPATH ROUTING ZIGBEE
NETWORK FROM THE PERSPECTIVE OF
COMMUNICATION
A. ZIGBEE NETWORK
ZigBee is a kind of wireless network protocol based on the
IEEE802.15.4 protocol and has the advantages of low cost,
low power consumption, a high level of security and
confidentiality, large network capacity and high reliability.
Therefore, it is widely used in low-rate wireless sensor
networks.
The application of the ZigBee network on a ship is of great
significance. Because the ZigBee network is an ad hoc
network, any terminal equipment on the ship can act as a
network node to join the ZigBee network at all times and
places, which allows the wireless network to cover the whole
ship.
2 VOLUME XX, 2020
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
B. THE SCHEME OF CONCURRENT AND DISJOINT
MULTI-PATH ROUTING
When there is something wrong with a certain node on the
path, a single-path wireless network cannot ensure that the
data are transmitted reliably; however, multipath routing can
transmit the data between the source node and the destination
node by multiple paths, which can improve the reliability and
real-time ability of data transmission. At present, many
researchers have studied multipath routing schemes. In [6],
an energy-efficient trusted cluster-based routing protocol for
WMSNs was proposed to overcome multiobjective problems,
which can eventually maximize the network lifetime. In [7],
a routing method with a load balancing algorithm was
proposed. The proposal uses the accumulated sending time of
each terminal as a weight in calculating each path quality
metric to decide multihopping routes with load balancing in
the network. In [8], a novel heuristic routing algorithm
known as Quality of Service (QoS) was proposed to provide
a multiobjective hybrid routing algorithm (Q-MOHRA) for
heterogeneous WSN. Q-MOHRA considers the link (energy,
hop count, link quality indicator, etc.) and path metrics for
optimal path selection.
By referring to the above research results and considering
the reliability and real-time ability of the steering control
system, the paper adopts the concurrent and disjoint
multipath routing ZigBee network and monitors node energy
using a coulombmeter.
1) ROUTING DISCOVERY
Multipath routing can be divided into intersecting path
routing and disjoint path routing. Each path of disjoint path
routing is independent. When errors exist in one path, there
is no impact on other paths, which can provide greater
network fault-tolerance and ensure reliable data
transmission [9].
FIGURE 1. Topological graph of concurrent multipath transmission
network.
The topological graph of the disjoint multipath routing
and concurrent transmission network is shown in FIGURE
1. S is the source node, D is the destination node, and P1,
P2 and Pn are the disjoint paths in the wireless network.
The steps to find the disjoint paths in the wireless network
are as follows.
i. First, S sends a RREQ to an adjacent node. When the
adjacent node receives the RREQ, it will reply to it if it is D.
If it is not D, it will forward the RREQ to another adjacent
node until D is found. In this way, the first path between S
and D is found, and all nodes on the path are marked.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships
ii. Repeat the first step, and the second path is found. The
marked nodes are not selected in search of the path, and other
nodes that are not marked can be selected. That is, each node
can only belong to one path.
The above steps are repeated to find the path between S
and D, which is not finished until all disjoint paths between
the two nodes are found. The disjoint multipath routing is
established according to the hop count. The path with the
smallest number of hops is P1, and the path with the largest
number of hops is Pn. After multiple paths are found and
marked, S and D can communicate with each other even
though one node may fail, which improves the stability of the
network.
2) PATH SELECTION STRATEGY
When S communicates with D in the wireless network, S
sends the data by the path with the smallest number of hops
in view of the time delay. On the other hand, S sends the
same data by concurrent multiple paths at the same time.
The steps of path selection are listed below when S and D
can communicate with each other.
i. When S sends the data to D, S first selects the paths with
the smallest numbers of hops in the routing table, that is, P1
and P2, and sends the same data packet simultaneously.
ii. When D receives the data, it only receives the packet
that arrives first and discards the packet that arrives later.
iii. If D finds continuous data loss in one path in the
process of receiving the data, D provides feedback about the
problem path to S. When S sends the data next time, it selects
P3 in the routing table to replace the problem path.
III. DESIGN OF SHIP COURSE-KEEPING
CONTROLLER FROM THE ASPECT OF CONTROL
A. DESIGN OF SHIP COURSE-KEEPING CONTROLLER
BASED ON THE WiNCS
To address some existing problems in the WiNCS, Chinese
and international researchers have come up with various
control methods. In [10], a static controller was designed that
takes the feedback from both state and output into account in
order to solve the problems of modeling and stabilization for
a wireless-based network control system with time delay. In
[11], a controller with explicit consideration of multipacket
transmission was investigated with a special focus on lossy
multipacket transmission in the wireless communication
context. In [12], a generalized predictive control with the
Kalman state estimator was proposed to compensate for the
wireless network-induced delay and packet loss.
Referring to the above research results, the controller in
this paper was designed by using an improved gray model to
predict system time delay and the deviation value between
the input and output of a system at a given time and applying
the predictive value to the nonlinear robust controller based
on the nonlinear decoration and integrator backstepping
method.
The structure of the wireless network control system is
shown in FIGURE 2, in which τsc is the time delay from the
sensor to the controller and τca is the time delay from the
VOLUME XX, 2020
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
controller to the actuator. If the time when the controller runs
the program cannot be neglected, it will be included into τca.
Therefore, the total time delay of network can be expressed
as τ=τsc+τca. The system acquires the time delay of a loop by
comparing the time of the output of the controller with the
time of the input of the controller without considering the
time synchronization problem [13]. In the kth period, the
controller records the time mark t1(k) when the control
quantity u1(k) is sent and the time mark t2(k) when the
feedback quantity is received, which yields the time delay of
the system loop in the kth period t= t2(k)-t1(k). At the same
time, the controller compares the time mark between the
output and input to judge whether the data packet is out of
order and discards any data packet that is out of order. If the
input time mark of the controller is not t2(k) in the kth period,
the wrong data packet is received. The controller discards the
wrong data packet, and the actuator compensates for the
packet loss. For the packet loss from the sensor to the
controller and from the controller to the actuator in the
WiNCS, the actuator uniformly compensates for packet loss
by setting a time threshold tf. When the actuator does not
receive the data in tf, which is considered packet loss, the
compensation control of packet loss is conducted in the
actuator.
FIGURE 2. Structure of the wireless network control system.
In the control structure, the two improved gray models are
the time delay prediction model and the deviation prediction
model. In the (k-1)th period, the time delay prediction model
sets up an improved GM(1,1) model by using the time delay
τ of transmission in the system loop, makes an online
prediction of time delay τ in the kth period and inputs τ into
the deviation prediction model. Another input into the
deviation prediction model is ship course error e. In the kth
period, the improved GM(1,1) model is established by using
ship course error e. The deviation prediction model online
predicts ship course error e1 at moment k+τ and ship course
error e2 at moment k+τ+tf and transmits the predictive value
e1 and e2 to the nonlinear robust controller, which determines
the controlled variable u1(k+τ) at moment k+τ and the
controlled variable u2(k+τ+tf) at moment k+τ+tf. Finally, the
actuator manipulates the controlled object based on u1(k+τ)
and places u2(k+τ+tf) in the buffer. u2 always remains
updated in the buffer. If the actuator does not receive the
controlled variable in tf, u2 in the buffer is applied to the
controlled plant.
3
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
B. THE IMPROVED GRAY PREDICTION MODEL
For the uncertainty and randomized factors of the system, the
gray prediction model can determine its change and
development trend to predict the data and control the system.
The gray prediction model only needs two parameters to
satisfy the requirement of forecasting for a system with some
unknown parameters and missing information due to its
strengths of having few original data requirements and
involving simple calculation steps for prediction [14,15].
When time delay and packet loss do not occur in the WiNCS,
the data information is known and reliable. However, if time
delay and packet loss do occur, the data information is
inaccurate or unknown. Therefore, the WiNCS is regarded as
a gray system. To predict the ship course error and time
delay by the gray model, the known and reliable information
is used to predict the unknown information and transmit the
predictive information to the controller in order to
compensate for time delay and packet loss.
The modeling process of the traditional GM(1,1) gray
prediction model is demonstrated as follows:
Assuming that the original sequence is
X(0)(k)={X(0)(1),X(0)(2),…,X(0)(n)}, its first-order
accumulating generation sequence is defined as
X(1)(k)={X(1)(1),X(1)(2),…,X(1)(n)}. X(1) can be modeled by
the first-order gray differential equation,
dX (1)
 aX (1)  b
dt
where a is the developing coefficient, b is the gray input,
respectively. Therefore, the values of parameters a and b
can be estimated by the least-squares error method as
follows:
 
aˆ  B T B
1
B T Yn
After obtaining a and b from (2), the solution of (1) can
be further defined. The initial condition is set as X(1)(1),
which is equal to X(0)(1), and the specified solution of (1)
can be then defined as
 b b integrator backstepping method has the defects of
Xˆ1 k    X 0  1  e  a(k  1) 
 a  a
By substituting k=(n+1) into (3), the one-step ahead
prediction can be calculated as
 b
Xˆ1 n  1   X 0  1  e  an 
 a
Finally, the predicted datum of X(0)(k+1) can be estimated
as
b
 

Xˆ0  k  1  1  e a  X 0  1  e  ak
a

4 VOLUME XX, 2020
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships
The GM(1,1) is the foundation of the gray prediction
model. This paper attempts to enhance prediction precision
in three directions, namely, initial condition selection, equal
dimension and new information in the gray theory and
background value optimization.
The GM(1,1) model allows new, high-precision data in
complement while removing some of the oldest data; the
original data are used to model the column vectors, with the
dimension remaining constant. This is called the equal
dimension and new information GM(1,1) model. Assuming
that the original sequence is X(0) ={X(0)(1),X(0)(2),…,X(0)(n)},
a new sequence X(0)={X(0)(2),X(0)(3), ... X(0)(n+1)} can be
obtained through the equal dimension and new information
GM(1,1) model.
For the sake of further optimizing the traditional
GM(1,1), the optimized background value is used to replace
the original background value. As an exponential function
is adopted in the gray model, the background value should
be optimized to fit the curve. By means of the integration of
exponential function, the optimized background value is
obtained, which can better fit the exponential function and
reduce the errors of the traditional GM(1,1).
C. THE IMPROVEMENT OF INTEGRATOR
BACKSTEPPING CONTROL FOR SHIPS WITH CONCISE
ROBUST CONTROL AND NONLINEAR DECORATION
Course-keeping for ships is the core of automatic navigation
(1)
in sea transportation. Much work has concentrated on
developing novel control strategies for closed-loop systems.
The closed-loop gain shaping algorithm is a concise robust
control algorithm [16]. The controller designed by the
closed-loop gain shaping algorithm has good robust
performance, simple controller structure and concise
(2) controller design procedures, but proving the stability of the
closed-loop gain shaping algorithm controller is slightly
difficult because of its constructional nature.
The backstepping control is a kind of nonlinear controller
design algorithm. Due to its unique advantages in dealing
with nonlinear control problems, backstepping method has
drawn much attention in the field of ship motion control.
However, course-keeping control through the normal
(3)
introducing many adjustment parameters and not considering
the energy cost of the actuator. To solve these problems, the
closed-loop gain shaping algorithm and backstepping method
are used to complement each other. In [17], an improved
concise backstepping controller based on a Lyapunov
candidate function were proposed, which introduces a
b nonlinear function of course error to replace the course error
(4)
a itself in the feedback loop. Compared with the pure
backstepping control, the improved algorithm proposed
reserves the nonlinear item of the system and possesses a
strong disturbance rejection ability and robustness to the
mathematical model uncertainty. In [18], to simplify the
design process of the nonlinear controller, decrease the
(5) number of undetermined parameters, improve robustness,
and reduce the energy consumption of the
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships

course-keeping controller for ships, an improved concise Considering Equation (6) and Equation (7), Equation (12)
design method is proposed by introducing an arctan is obtained.
nonlinear function passed through by course error signal K0
under the backstepping design framework, while the design
z2  
T0
 x2   x23   bu    c1 x2
process of the controller is simplified to only one step. In (12)
[19], to enhance the performance of the integrator Assuming there is no disturbance term Δ in (12).
backstepping controller, a control law formed by the closed-  
V2  c1 z12  z2 z1    a1 x2  a2 x23  bu  c1 x2  (13)
loop gain-shaping algorithm is adopted to replace the linear
K
terms in the original control law. Then, the improvement of where a1   0  , a2   K 0  .
integrator backstepping control for ships with concise robust T0 T0
control and nonlinear decoration was proposed to enhance The control law u can be obtained.
the robustness of the controller. 1
u    a1 x2  a2 x2 3    c1 x2  z1  c2 z2 
To guarantee that the ship course  tracks the desired b (14)
course  r , let e    r , x1   and x2  x1  r   ; the where c2  0 is a design parameter.
nonlinear dynamic model of the ship can be described as According to Equation (13) and Equation (14), we have
 x1  x2 V2  c1 z12  c2 z2 2  0 .

 x2  f ( x2 )  bu   (6) According to the Lyapunov stability theorem, the error
y  x z2 can be stabilized by the control law (14) without
 1

where yR is the output of the system, consideration of interference term Δ.

For the system (6) and system error definition (7), the
K0 K
f ( x2 )   H ( ) , H ( )     3 , b  0 , u   , control law of the system can be defined as
T0 T0 1
Ψ is the ship heading, K0 and T0 are the maneuverability u    a1 x2  a2 x2 3    c1 x2  z1   c2    z2 
b (15)
indices of the ship, δ is the rudder angle,  and  are
where   0 is a design parameter.
nonlinear parameters, Δ is the uncertain disturbance and u is
According to Equations (11), (12) and (15), we have
the designed control law for course-keeping.
For a system described as (6), the course-keeping V2  c1 z12  c2 z2 2   z2 2  z2 
controller design method via backstepping is as follows: The control law can be redesigned with the introduction
First, define the error of the system as of a nonlinear damping term to eliminate the uncertainty
  z2 term.
 (7) We needs to prove z   0 to achieve V  0 .
2 2
 z1  x1  r According to [19], we have
z  x  2
 2 2 
V2   c1 z12  c2 z2 2 
where  is the virtual control variable, z1 and z2 are the 4 (16)
2
state variables, and  is an integral term introduced in the 
  c2 z2 2  
controller design to eliminate the static error caused by an 4
uncertain interference term Δ. 2

Set the first Lyapunov function as Therefore, c2 z2 2   should be proved. Moreover,
1 4
V1  z12 2
2 
2
c2 z2   can be expressed as
4
V1  z1  z2    r 
(8) 
2

Define the virtual control parameter as f 3  z2    (17)

  c1 z1   r 4
(9) According to Equation (17), we have
where c1>0 is a controller design parameter. Substituting
  2
Equation (9) into Equation (8), Equation (10) is obtained. z2  f 3 1    (18)
V1  c1 z12  z1 z2 (10)  4 
 
Second, define the second Lyapunov function as Substituting Equation (18) into Equation (16), we
 1
V2  V1   2  z2 2 achieve V2  0 .Because     , when the condition
2 2
V2  c1 z12  z2  z1    z2    
(11) R   z 2 : z2   is satisfied, the stability of the
 2  c2 
where  is a constant larger than 0.

VOLUME XX, 2020 5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships

system error z2 incorporating the uncertain disturbance can network module, ship model and controller module. A photo
be guaranteed by the definition of the control law (15). of the simulation platform is shown in FIGURE 3.
According to [17-20], the designed control law u can be
used directly when designing a nonlinear controller based
on integrator backstepping methodology.
1 1 T
u   (a1  a2 3 )  v  H ( )  0 v (19)
b b K0
where v  k p e1  ki e1dt  kd e1 .

The design scheme of the closed-loop gain shaping
algorithm controller is used to design the linear PID
controller, and the sensitivity function T of the closed-loop
system is given as FIGURE 3. Photo of simulation platform.
GK
T  1  T1s  1  (20)
1  GK The network module consists of a wired network and a
where the controller K can be easily deduced based on the redundant wireless network. The wired network uses Internet
selection of the bandwidth frequency of the closed-loop as a carrier to transmit the data. The redundant wireless
system 1/ T1 and the controlled plant G. network employs the ZigBee module as the communication
v 1 module. The program of the ZigBee module is written by C
K  (21) language. The components of the ZigBee network are the
e1 GT1s
terminal node, routing node and coordinator. The terminal
The ship model, a standard Nomoto model, is expressed node is responsible for collecting and processing the
in Equation (22). information and communicating with the routing node. The
routing node establishes a channel that is used to transmit the
K0 communication between the coordinator and the terminal
G  s   (22)
node in order to facilitate the transmission of information and
s T 0s  1
expand the scope of the network. The coordinator establishes
the ZigBee network, sets the relevant configuration of the
To eliminate the steady state error using the closed-loop network, receives data information from each node and
gain shaping algorithm, a minor constant ɛ(0.01) is added transmits the computer’s orders to the corresponding nodes.
into the denominator of the Nomoto model [20]. Then, To enhance the coordinator’s real-time capability, stability
Equation (23) is obtained: and task management, the STM32 microcontroller is used as
the core of the coordinator. It runs the real-time Mindows
K0 operating system and communicates with the ZigBee module
G s   2
(23) through the USART interface. The operating system run is
T0s  s  
called Mindows, a small and real-time embedded operating
system written in the development environment of Keil
According to [17-20], the nonlinear decoration control MDK4.70 and used in embedded devices. The user entry of
law u can be obtained as follows, where 0<ω <1. the Mindows operating system is the MDS_RootTask
T0 1  T function rather than the main function, and the user can write
u (   0 s )sin(e1 ) the user code directly without considering the initialization
K 0 K 0T1 K 0T1s K 0T1 (24) process after entering the user program. Each C file of the
The stability of the proposed nonlinear robust control law operating system has two h header files, and the inner header
is proved based on the Lyapunov theorem. The controller file is the internal header file of the operating system, which
design procedure is definite and simple without reducing contains only internal information and is not provided to the
the robustness and control performance. Only two tuning user. A header file without an inner header file is an external
parameters (ω and T1) need to be selected, and T1 can be header file that comprises information available to the
determined according to the system bandwidth. operating system and the user. FIGURE 4 shows the
corresponding relationship of the Mindows operating system
IV. SIMULATION PLATFORM AND SIMULATION file. Mindows supports a multitasking preemptive system,
STUDY which achieves the real-time switching of tasks, implements
the semaphore mechanism and message queue, and can cut
A. SIMULATION PLATFORM off unwanted functions. The detailed information was shown
To study the application of the redundant control network to in [21].
the steering control system, a simulation platform based on a
realistic hybrid network was established, which includes a

6 VOLUME XX, 2020

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships

Start

Accessing battery
voltage register

Vbat>3.1v No

Yes

No Starting first
time？

Yes
Configuring register
and resetting
voltameter

Estimating battery level(Cref)

according to battery voltage Delay=5s
value

Correction Cref Correction error

Accessing battery SOC=Cref+REG_Cha

voltage register rge

FIGURE 4. Calling relationships of the Mindows operating system file. Vbat>3.1v Yes

For the problems regarding the power of the ZigBee nodes, No End
the power of nodes is supplied by two dry batteries.
STC3100, a battery monitoring chip with a coulombmeter, is FIGURE 5. Flow chart of battery level monitoring task.
adopted in each ZigBee node to prevent node failure in case
of a dead battery, accurately monitor battery voltage, The ship model and controller module are written by VB
temperature and current and ensure the real-time output of language and run respectively in two upper computers. They
battery power. When the battery power is low, the node establish the wired network based on TCP/IP protocol
sends a warning message to the upper computer and informs through the Winsock, use the RS232 serial port protocol to
users of the need for timely replacement of the battery in the establish the connection communication with the wireless
node. In addition, auto reboot at regular intervals is set in the network by the MSComm and concurrently send the same
ZigBee node to avoid the abnormality of a node for a long data to keep the real-time data communication through the
period. Internet and ZigBee network. The function structure diagram
The ZigBee nodes visit the STC3100 register and read and flow diagram of the simulation platform are shown
data regarding aspects such as the change in battery level, separately in FIGURE 6 and FIGURE 7.
battery voltage, electric current and temperature. When the
operating system processor starts, it sets the parameters of
STC3100 register and launches its tally function of the
battery level. If the operating system processor is starting for
the first time, the gateway first forecasts the battery level
through the battery voltage. Then, it calculates the battery
level in real time. The flow chart is shown in FIGURE 5. The
battery level varies with the temperature, the numbers of
cycles of charging and discharging and the battery life. In
addition, the accumulative errors caused by many factors
such as temperature, aging of the battery and resistance
accuracy should be considered to obtain an accurate battery
level. Therefore, the gateway regularly eliminates
accumulative errors to maintain the accuracy of the battery FIGURE 6. Structure diagram of the function of the simulation platform.
level monitoring.

VOLUME XX, 2020 7

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships

that the ship model and controller send and receive, the
packet loss rates are worked out when the metal door of
engine room is closed, as shown in TABLE 1. The table
shows that the maximum packet loss rate is 0.29%, and the
average packet loss rate is 0.22%. The experiment results
prove that the packet loss rate is low, which can meet the
needs of the ship communication and verify the reliability of
the data communication.
TABLE 1 PACKET LOSS RATE WHEN THE DOOR CLOSES
Module Sending Receiving packets Packet Loss
packets rate
Ship model 10000 9971 0.29
Controller 10000 9986 0.14

Taking the new launched training ship YUPENG of

Dalian Maritime University as an example, the main ship
parameters are shown in TABLE 2. Hence, the model
parameters can be determined, as listed in TABLE 3.
TABLE 2 SHIP DETAILS OF TRAINING SHIP YUPENG
Length between perpendiculars L (m) 189
(LBPs)
Molded Breadth B (m) 27.8
Molded draught d (m) 11.0
FIGURE 7. Flow diagram of simulation platform. Ballast draught d (m) 6.313
3
Volume of displacement ▽ (m ) 42293
When the simulation platform runs, the parameters Trial speed V (kn) 17.26
regarding time delay and packet loss are first set in the 2
Rudder area AR (m ) 38
network module. Then, the communication is set in the ship
model and controller module. Finally, the network Rudder area (ballast) AR (m2) 31.67
connection is established. After that, the ship course is set, Block coefficient Cb 0.72
and the control algorithm is chosen in the controller interface.
Block coefficient (ballast) Cb 0.661
In the ship model interface, the ship model is selected, and
the ship data is set. The experiments begin. The data are Longitudinal center of gravity xc (m) -4.043
(ballast)
processed by the algorithm model. The simulation curves of
the ship course and rudder angle are drawn in the ship model TABLE 3 SHIP PARAMETERS UNDER BALLAST CONDITION
interface. Turning ability index K0 (1/s) 0.21
Following index T0 (s) 107.76
B. SIMULATION STUDY
The ship model and controller are simulated and studied in In the simulation experiments, the setting course is 20°.
two upper computers. They communicate with each other The sampling period T is 1 s, and the simulation time t is
through a hybrid network composed of an Internet network 2200 s. When the simulation time is 200 s, the Internet is
and the redundant ZigBee network. This study emphasizes disconnected. In the simulation time from 200 s to 600 s,
the simulation of the control of the ship course-keeping based wireless network experiences no interference. Since the
on wireless network control in the redundant ZigBee network simulation time is 600 s, the wireless network experiences
in case the Internet network cannot be connected. The ten interference. The time delay [0, 4T] is set in the network
network nodes are placed in the redundant ZigBee network module from 600 s to 900 s. The time delay 10T is set in the
of 200 m×5 m. According to the construction method of the network module from 1000 s to 1200 s. When the simulation
path proposed in this study, there are three concurrent and time is 1600 s, the packet loss rate is set at 100% in the
disjoint paths between the two upper computers. network module. The simulation results are presented in
When the metal door of the ship engine room is closed, the
FIGURE 8 and FIGURE 9. Dfai is the course error, and  is
wireless signals can be obstructed. To solve this problem and
rudder angle. The two schemes about the application of
ensure the reliability of wireless communication, the nodes
wireless network control to ship course-keeping control are
are placed inside and outside the door of the engine room,
as follows.
and they communicate with each other through RS232 serial
In the first scheme, the robust PID control is used to
ports. The experiments of data collection that put emphasis
simulate ship course-keeping control on the simulation
on the packet loss rate are conducted in the simulation
platform. The simulation result without considering the
environment of the ship engine room to prove the feasibility
WiNCS problems is shown in FIGURE 8.
of this design. By comparing and analyzing the data packets

8 VOLUME XX, 2020

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships
In the second scheme, the collaborative design of the
communication protocol and control algorithm is studied to
solve problems such as the time delay and packet loss in
WiNCS. The ship course-keeping controller is designed to
combine the improved gray model with the nonlinear robust
controller based on the nonlinear decoration and integrator
backstepping method. FIGURE 9 shows the simulation result
when considering the WiNCS problems. In the gray model of
the controller, n = 10 and time threshold tf = 10T.
FIGURE 8. Simulation result without considering the WiNCS problems
FIGURE 9. Simulation result when considering the WiNCS problems
It can be seen in FIGURE 8 that small amplitudes of
fluctuation appear in the ship course because of some
instable factors in the normal wireless network. When the
small time delay is set in the network module from 600 s to
900 s, different amplitudes of fluctuation are seen in the ship
course. When the simulation time is 1100 s, a large time
delay is set. The percentage of overshoot is 60%, and the
amplitude of rudder angle is ±10° when  is approximately
﹣2°. When the simulation time is 1600 s, the packet loss
rate is set at 100% in the sampling period 15T. The large
amplitudes of fluctuation appear in the ship course, and the
rudder angle fluctuates ±20° around﹣2°.
FIGURE 9 shows that the rudder angle fluctuates ±2°
around﹣2° in the normal wireless network. The ship course
VOLUME XX, 2020
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
control has a satisfactory effect. After the small time delay is
set, the ship course has a smaller overshoot. After the large
time delay is set, the percentage of overshoot is 10%. When
the packet loss rate is set at 100% in the sampling period 15T,
the percentage of overshoot is 25%, and the rudder angle
fluctuates ±10° around﹣2°.
Through the comparison and analysis of the results
between FIGURE 8 and FIGURE 9, it is found that the first
and second schemes are adopted to stably control the ship
course in the normal environment of the wireless network.
When the small time delay is set, the control effects in the
second scheme are better than those in the first scheme.
When the large time delay is set, the control effects in the
first scheme are unsatisfied, but they have good performance
in the second scheme. When the high rate of packet loss is set,
the first scheme adopted is unable to stably control the ship
course. However, the control effects are satisfied in the
second scheme, and the second scheme achieves the purpose
of ship course-keeping control. It can be concluded that the
ship course can be controlled stably when the small time
delay is set whether the WiNCS problems are considered or
not. On the other hand, when the large time delay and high
rate of packet loss are set, the ship course cannot be
controlled stably if the WiNCS problems are not considered.
Therefore, the ship course-keeping controller based on
WiNCS is adopted to achieve improved control performance
and robustness.
V. CONCLUSION
With the development of ship wireless network technology,
wireless networks are applied in ship control systems. For
ships in which a redundant control network needs to be
installed, this paper proposed a scheme in which a wireless
network can be used as a redundant network, which can
solve the problems of a redundant wired network and which
has the characteristics of flexible installation, low cost and
powerful extensibility. The paper applied wireless network
control to course-keeping for ships in order to validate the
effectiveness of the proposed ship wireless communication
network. For the problems concerning node power, the
coulombmeter is adopted in each node to monitor node
power in real time. In view of other problems in the WiNCS,
the codesign of a communication protocol and control
algorithm was studied to solve problems such as time delay
and packet loss. In this paper, a novel idea is presented, and
the ship course-keeping controller is designed by using an
improved gray model to predict the system time delay and
the deviation value between the input and the output of a
system at a given time and applying the predictive value to
improve the integrator backstepping control with concise
robust control and nonlinear decoration, which can
compensate for the time delay and packet loss. The
experiment results prove that the ship course-keeping
controller based on the WiNCS has improved control
performance and robustness. Furthermore, in extreme
network environments, the scheme can optimize the overall
performance of
9
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2973464, IEEE Access
R.Jiang,X.Zhang: Application of Wireless Network Control to Course-Keeping for Ships

wireless network control and yield better control effects,
which has some reference value in practical application. In
the future, we will carry out the practical implementation of
the control system.
ACKNOWLEDGMENT
The authors would like to thank all the anonymous reviewers
for their valuable comments and suggestions for improving
the quality of this paper.
[19] X.Zhang, X.Han, W.Guan, and G.Zhang, “Improvement of integrator
backstepping control for ships with concise robust control and
nonlinear decoration,” Ocean Engineering, vol. 189, pp. 1-7, 2019.
[20] G.Xu and X.ZHANG, “Network Control-based Concise Robust
Control of Ship Course-keeping,” J.Dalian Maritime Uni., vol. 40,
no.1, pp. 5-8, 2014.
[21] Q.Wang, Z.Gu, and R.JIANG, “The kernel scheduling of embedded
operating system,” Beijing: Beihang University Press, 2015.

REFERENCES
[1] P.Mao, “The Design of a Monitoring system for Marine Engine
Room Base on Zig Bee Protocol Stack,” Logistics Eng. Management,
vol. 39, no. 6, pp. 188-190, 2017. RIFAN JIANG is studying for a Ph.D. degree
[2] G.Lian, J.Zhang, B.Chen, F.Ban, Z.Hou, and H.Li, “Design and and majoring in Traffic Information Engineering
Experimental Study of the Wireless Online Monitoring System of a and Control from Dalian Maritime University,
High-Temperature Superconducting Machine,” IEEE Tran. Applied Dalian, China. His current research interests
Super., vol. 29, no. 2, pp. 1-5, 2019. include nonlinear feedback control, wireless
[3] S.Huai, S.Zhang, J.Zhang, and K.Huang, “Holographic detection of network control and ship robust control.
AIS real-time signals based on sparse representation,” Eurasip J.
Wireless Com. Net., vol. 1, 2019.
[4] L.Wang and Y.Yang, “Application research of wireless sensor in
ship-marking network based on ZigBee,” Ship Sci. Technology, vol.
41, no.5, pp. 136-138, 2019.
[5] R.Jiang and X.Zhang, “Wireless Network-Based Platform for Ship
Course-Keeping Control,” Nav. China, vol. 40, no. 1, pp. 1-3, 2017.
[6] K.A.Ranjith and A.Sivagami, “Security Aware Multipath Routing
Protocol for WMSNs for Minimizing Effect of Compromising
Attacks,” J. Net. Systems Management, vol. 37, no. 3, pp. 573-599,
2019. XIANKU ZHANG received his Ph.D. degree
[7] T.HABARA, K.MIZUTANI, and H.HARADA, “A Load Balancing from Dalian Maritime University (DMU),
Algorithm for Layer 2 Routing in IEEE 802.15.10,” Ieice Tran. Com., Dalian, China, in 1998, where he is currently
vol. E101-B, no. 10, pp. 2131-2141, 2018. a Professor/Doctor Supervisor. He is a
[8] N.Kulkarni, N.R.Prasad, and R.Prasad, “QoS Assured Multi- Scholarship Leader of a National Key
objective Hybrid Routing Algorithm for Heterogeneous WSN,” Discipline Named Traffic Information
Wireless Personal Com., vol. 100, no. 2, pp. 255-266, 2018. Engineering and Control, and the Vice
[9] D.Wu, S.DONG, and Q.KANG, “Survey on Multipath Routing Director of Marine Simulation and Control.
Protocol for Wireless Multimedia Sensor Networks,” Tran. He has authored or coauthored 230 papers,
Microsystem Tech., vol. 33, no. 7, pp. 4-8, 2014. among which 18 papers are referred to by the
[10] J.MA, F.PAN, and L.ZHOU, “Modelling and stabilization of a Science Citation Index and 70 papers are
wireless network control system with time delay,” Tran. Ins. Meas. referred to by the Engineering Index, and 14
Control, vol. 40, no. 2, pp. 640-646, 2018. books in the fields of ship motion control, robust control, intelligent
[11] S.HASSAN, R.IBRAHIM, and N.SAAD, “Predictive PI controller control, and computer programming. His current research interests include
for wireless control system with variable network delay and robust control, nonlinear control, and their application in marine control
disturbance,” IEEE Int. Sym. Rob. Manufacturing Automation, pp. 1- systems. He has received one second-class scientific award in China and
6, 2017. two second-class scientific awards from Liaoning, China.
[12] M.F.R.LEE, F.H.S.CHIU, and H.C.HUANG, “Generalized
Predictive Control in a Wireless Networked Control System,” Inter. J.
Dis. Sensor Net., vol. 2013, no.2, pp. 870-875, 2013.
[13] J.LI, B. LIU, and Z.JIANG, “Networked Predictive Control of
Internet_based Telerobotics Based on Grey Prediction of Delay,”
Computer Simu., vol. 33, no.1, pp. 331-335, 2016.
[14] H.Zhang, J.JI, and G.Wang, “Application of grey predictor algorithm
on the synchronous control of TBM propulsion system,” J. Machine
Des., vol. 35, no.2, pp. 1-5, 2018.
[15] Z.Wang and M.LI, “Application of grey prediction fuzzy PID
technology in cooling water temperature control of marine main
engine cylinder liner,” Ship Sci. Technology, vol. 39, no.7, pp. 79-81,
2017.
[16] X.Zhang, “Simple Robust Control of Ship Motion,” Beijing: Science
Press, 2012.
[17] X.Zhang, G.Yang, Q.Zhang, and G.Zhang, “Improved Concise
Backstepping Control of Course Keeping for Ships Using Nonlinear
Feedback Technique,” J. Navigation, vol. 70, no.6, pp. 1401-1414,
2017.
[18] Q.Zhang and X.Zhang, “Nonlinear Improved Concise Backstepping
Control of Course Keeping for Ships,” IEEE Access, vol. 7, pp.
19258-19265, 2019.

10 VOLUME XX, 2020

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.