See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332834984

Collision avoidance, guidance and control system for Autonomous Surface

Vehicles in complex navigation conditions

Chapter · May 2018

DOI: 10.1201/9780429505294-15

CITATIONS READS

11 975

2 authors:

M.A Hinostroza Carlos Guedes Soares

Norwegian University of Science and Technology University of Lisbon
58 PUBLICATIONS 730 CITATIONS 2,886 PUBLICATIONS 61,706 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by M.A Hinostroza on 03 May 2019.

The user has requested enhancement of the downloaded file.

Collision avoidance, guidance and control system for Autonomous
Surface Vehicles in complex navigation conditions
M.A. Hinostroza & C. Guedes Soares
Centre of Marine Technology and Ocean Engineering (CENTEC), Instituto Superior Técnico, Universidade
de Lisboa, Lisbon, Portugal

ABSTRACT: This paper presents a collision avoidance, guidance and control system for operation of auton-
omous surface vehicles in critical navigation conditions. The collision avoidance unit is based on fuzzy logic
intelligent decision-making algorithm, the guidance unit uses the line-of-sight algorithm and, the control unit,
is composed by a PID heading controller and a speed controller. A set of numerical simulations were carried
out for different collision scenarios to validate the effectiveness and feasibility of the system. Also a case
study involving ship traffic in the estuary of Tagus river, Lisbon was performed and a good performance of
the system was found.

1 INTRODUCTION with the U.S. Navy. If tests are successful, future

such craft may be armed and used for anti-submarine
Human error contributes to more than 80% of and counter-mine duties, operating at a small frac-
ship collision accidents, which pose a rather severe tion of the cost of operating a destroyer.
threat at sea. Research on autonomous collision Recently, the World’s first autonomous cargo
avoidance at open sea is important for mitigating the vessel, was announced, YARA Birkeland is planned
overall risk of collisions in intelligent navigation, to sail in the Norwegian fjords, and is scheduled for
(Perera et al. 2011). These systems are designed to fully autonomous operation in 2020 (Kongsberg-
support decision-making for navigators and isolate Maritime, 2017). This will be a giant step for mari-
human errors during collision avoidance. However, time traffic, and sets the start for unmanned marine
in near future, autonomous surface vessels will enter vehicles. Nevertheless, there are still challenges re-
in maritime transportation, and it will needs to in- lated to these operations.
corporate an automatic collision avoidance system to Other examples is small scale the family of ASVs
ensure safe navigation. which includes scaled fishing trawler type vessel,
With the increasing number of successful projects ARTEMIS, the catamaran models, ACES and Auto-
in autonomous surface vehicles (ASVs), there is an Cat , and the SCOUT vessels. Of all these proto-
increased interest in its concept application in the types, the kayak type, SCOUT vessels have success-
maritime industry. This paper refers to the applica- fully implemented COLREGs at a basic level for
tion of the concept of unmanned navigation, involv- head-on situations whilst maintaining wireless
ing the reduction and minimization of crew on-board communication (Benjamin and Curcio, 2004).
vessels, with the ultimate goal in mind of a ship car- Moreira et al. (2008) performed successfully
rying out its functions without man on-board over- model tests using a scaled model of the tanker “Esso
time. In this scenario, Rolls-Royce (2014) has an- Osaka”, which was instrumented for autonomous
nounced that remotely operated local vessels are operation and different guidance and control ap-
expected by 2020 and ocean-going ships for mari- proaches. Perera et al. (2015), Ferrari et al. (2015),
time transport by 2030. These unmanned ships aim Hinostroza et al. (2017) and Xu et al. (2018) have
to increase safety of operations at the sea, reduce conducted model tests with an autonomous surface
fuel consumption, and transform the work roles in vehicle of 2.5 m for intelligent navigation and colli-
the maritime domain. The Defense Advanced Re- sion avoidance, manoeuvring tests and marine ex-
search Project Agency (DARPA) a developed Sea ploration and system identification, respectively.
Hunter, built as part of its Anti-Submarine Warfare Many techniques have been proposed for avoid-
Continuous Trail Unmanned Vessel (ACTUV) pro- ance of collision situations, Sato & Ishii (1998);
gram, began sea trials in April of 2015. Sea Hunter Statheros et al. (2008), but in general those tech-
is a 132 feet long trimaran (a central hull with two niques ignore the law of the sea as formulated by the
outriggers), manufactured deep in the Silicon Forest International Maritime Organization (IMO) in 1972.
of Portland, Oregon, Njus (2016). It is expected to These rules and regulations are expressed in the
undergo two years of testing before being in service Convention on the International Regulations for Pre-
venting Collisions at Sea (COLREGs). In recent lit- grid map, where both static and dynamic obstacles
erature new formulations including COLREGs regu- are well represented. By using such a map, a colli-
lation were addressed, Zhang et al. (2015) has pre- sion free path may be generated which can be direct-
sented a distributed anti-collision decision support ly used as a guidance trajectory for practical naviga-
formulation in multi-ship encounter situations under tion.
COLREGs ; He et al. (2017) has performed a Quan-
titative analysis of COLREGs rules and seamanship
for autonomous collision avoidance at open sea. The 2 MATHEMATICAL FORMULATION
present convention was designed to update and re-
place the Collision Regulations of 1960, which were In this section, the mathematical formulation of col-
adopted at the same time as the International Con- lision avoidance and guidance system for ASVs,
vention for Safety of Life at Sea (SOLAS) Conven- showed in Figure 1, is presented. The collision de-
tion. tection block uses the close distance of approach be-
Fuzzy-logic based systems, which are formulated tween vessels to determine the collision risk from
for human type thinking, facilitate a human friendly COLREGs rules. The collision avoidance unit uses
environment during the decision making process. the fuzzy logic algorithm to compute the desired
Hence, several decision making systems in research heading and speed in order to avoid the obstacles.
and commercial applications have been presented The Guidance and control block is where the trajec-
before. Automatic collision avoidance systems for tory tracking is performed based in a desired heading
ship systems using fuzzy logic based control systems and ASV speed. Notice that the collision avoidance
have been proposed by Hasegawa (1987). The con- unit is based in the work proposed by Perera et al.
junction of human behavior and the decision making (2011). However, this paper includes a collision de-
process has been formulated by various fuzzy func- tection module, which based in the distance to clos-
tions in several works. Fuzzy logic based decision est point of approach and the time to the closest
making system for collision avoidance of ocean nav- point of approach detects potential conflicts; an iter-
igation under critical collision conditions was pre- ative fuzzy membership function in order to reduce
sented in Perera et al. (2011).However, the simula- the number of fuzzy rules, and the ship motions are
tion results are limited to the two-vessel collision simulated using a realistic, 3DOF nonlinear mathe-
avoidance situations. matical model. The present system, also, incorporate
a path-planning system and controller plant.

2.1 COLREGs rules and regulations

The COLREGs, IMO (1972), include 38 rules
that have been divided into Part A (General), Part B
(Steering and Sailing), Part C (Lights and Shapes),
Part D (Sound and Light signals), and Part E (Ex-
emptions). In this study the COLREGs Part B, con-
cerning Steering and Sailing rules are considered.
It is a fact that the COLREGs rules and regula-
tions regarding collision situations in ocean naviga-
tion have been ignored in most of the recent litera-
ture. The negligence of the IMO rules may lead to
conflicts during ocean navigation. As for the report-
Figure 1 Block diagram of collision avoidance, guidance and con- ed data of maritime accidents, 56% of major mari-
trol system time collisions include violations of the COLREGs
rules and regulations. Therefore, the methods pro-
This paper focuses on a development of a novel posed by the literature ignoring the COLREGs rules
complete navigation system for ASVs, including a and regulations should not be implemented in ocean
collision avoidance unit, guidance and control sys- navigation. On the other hand, there are some practi-
tem, Figure 1. This system combines a practical col- cal issues regarding implementation of the
lision avoidance unit based on fuzzy logic intelligent COLREGs rules and regulations during ocean navi-
decisions and a guidance and control unit based in gation. The Own vessel Head-on and Overtake situa-
Line-of-sight algorithm and heading and speed con- tions and crossing are presented in Figure 2. In
trollers. It is a work specifically solving the collision Crossing situations where the Own vessel is in
avoidance problem navigating in complex environ- ‘‘Give way’’ situations or, ‘‘Stand on’’ situations
ments, including dynamic obstacles. The algorithm where there are velocity constrains in implementing
designed in this paper is able to extract information COLREGs rules and regulations of the ‘‘Give way’’
from a real navigation map to construct a synthetic and ‘‘Stand on’’ vessel collision situations when the
Target vessel has very low or very high speed com- the TCPA to determine the collision risk (CR),
pared to the Own vessel.
(1)
 DCPA  dthreshold
CR  
 TCPA  tthreshold

where dthreshold and tthreshold are defined by COLREGs

rules and regulations, i.e. dthreshold  6 miles and
tthreshold  20 min , He et al. (2017).
2.3 Fuzzy-logic collision avoidance unit
The fuzzy logic collision avoidance unit is re-
sponsible to compute the desired heading and veloci-
ty of own vessel in order to avoid potential conflicts
in a desired trajectory. This study uses the Mandami
formulation, the relative collision situation for two
vessels are showed in Figure 4.

2.3.1 Collision conditions

Figure 2 COLREGs rules and regulations, Perera et al. Figure 4 presents a relative collision situation in
(2011) ocean navigation that is similar to a Radar plot. The
Own vessel ocean domain is divided into two circu-
lar sections, the ship domain and the actual approx-
2.2 Collision detection module imate distance to the target vessel.
The collision detection module is responsible to The Own vessel Collision Regions are divided in-
determine if there is a potential conflict in route as to eight regions from I to VIII. These regions are
shown in Figure 2, where two vessels have the pos- separated by dotted lines that are coincident with the
sibility of collision at the trajectory crossing point. Collision Regions as formulated in the Fuzzy mem-
To effectively eliminate the conflict, it is necessary bership function (FMF), presented in subsection
to compute the distance to the closest point of ap- 2.3.2. It is assumed that the Target vessel will be lo-
proach (DCPA) and the Time to the closest point of cated within one of these eight regions and the colli-
approach (TCPA). As shown in Fig. 3, vessel 1 is sion avoidance decisions are formulated in accord-
travelling with the velocity V1 and vessel 2 has the ance to each region.
velocity of V2. V12 is the relative velocity of the Target vessel positions have been divided into eight
vessel1 with the respect to vessel 2, D is the distance divisions of vessel orientations regarding the relative
between two vessels and γ is the angle between the course (II-a, II-b, II-c, II-d, II-e, II-f, II-g and II-h).
relative motion line and the bearing angle of vessel These divisions are separated by dotted lines that are
1. coincident with the Relative Collision Angle FMF.

Table 1 present the summarized collision risk as-

sessments and decisions of the two-vessel collision
situation in Figure 3. The first column represents the
Collision Regions (Reg.) with respect to the Own
vessel, and the second column represents the Divi-
sions (Div.) of the Target vessel orientations. The
third column  rel represents relative collision angle,
fourth column is the DCPA and fifth is TCPA, using
these inputs is possible to assessment the Collision
Risk (Risk) with respect to each of the Collision Re-
gions. From this collision risk the actions to avoid
collision are in column sixth and seventh, according
COLREGs rules and regulations.
Figure 3 Calculation of DCPA and TCPA

The collision detection module uses this DCPA and

 Figure 4 Relative collision situation for two vessels, Perera et al. (2011)

Table 1 Rules for fuzzy logic algorithm set of all elements of the universe typical to A that
Region Div.  rel. DCPA TCPA Decision Decision are associated with the membership value of 1.
II f
2 medium Moderate   0  V0  0 The support of the fuzzy set is defined as the set
  0 of all elements of X that have nonzero membership
II f
1 medium Moderate N.A
degree in A. The FMF for inputs, collision distance
II f
3 medium Moderate   0 N.A
(R), DCPA, TCPA and relative collision angle (v),
I e
1 medium Moderate   0 N.A are presented in Figs 5,6, and 7 respectively. Figure
8 are formulated for the output FMFs of speed (dψ)
I e
2 medium Moderate   0  V0  0 and course (dV) change of the Own vessel. The core
I e
3 medium Moderate   0 N.A and support variables are listed on the respective
figures of inputs and outputs FMFs.
VIII d
1 medium Moderate   0 N.A

VIII d
2 medium Moderate   0  V0  0
VIII d
3 medium Moderate   0 N.A

II f
2 Small Short   0  V0  0
II f
1 Small Short   0  V0  0
II f
3 Small Short   0  V0  0 Figure 5 Relative region input fuzzy function
I e
1 Small Short   0  V0  0
I e
2 Small Short   0  V0  0
I e
3 Small Short   0  V0  0
IV h
1 medium Moderate   0 N.A

IV h
2 medium Moderate   0  V0  0
IV h
3 medium Moderate   0 N.A

*N.A, no action

2.3.2 Fuzzy membership function (FMF)

Figure 6 Distance to closest point of approach input FMF
FMF describes fuzzy sets that map from one giv-
en universe of discourse to a unit interval. This is In this work the FFM relate to the relative regions
conceptually and formally different from the funda- v1, v2, v3 are dependent on the relative region value,
mental concept of probability, (Pedrycz & Gomide θ, Figure 7 present the FMF for   0 ;
,2007). The core of the fuzzy set A is defined as the
 Figure 7 Variable fuzzy membership function

Figure 9. Coordinate frames for 3DOF marine surface vehicle.

As presented in figure 9, u c is the current’s magni-

tude,  is the current’s direction,  is the ship’s
heading angle, u is the forward component of veloc-
ity over ground, and v is the transverse component
of velocity, the relative forward velocity and trans-
verse velocity are given by
ur  u  uc cos(   )
(2)
vr  v  uc sin(   )

The time derivatives of u and v are given:

u  ur  uc r sin(   )
(3)
v  vr  uc r cos(   )
where the accelerations of the motion in 3 degree of
freedom (surge, sway and yaw) are given by
(m  X ur )ur  mvr r  mxG r 2  f1
(m  Yvr )vr  (mxG  Yr )r  mur r  f 2 (4)
(mxG  N vr )v  ( I z  N r )r  mxG ur r  f 3
Where m the
of the mass
ship is,
 X ur Yvr Yr  N vr  N rare the added mass and
moment, respectively. The dimensionless forces are
defined as multi-variety third-order regression poly-
nomials depending on the non-dimensional veloci-
Figure 8 Fuzzy membership function ties.
f1  11ur2  22 nur  33n2   CR CR   X v v 
2.4 Guidance and control module r r
(5)
 X v 2 vr2   Xee X e' 2 e2   Xrr X r' 2 r 2   Xv r X v' r r vr r    v2r 2 X v' 2r 2 vr2 r 2
r r r r

2.4.1 3DOF Mathematical model

 
In order to simulate the 3DOF planar ship motions.   Y'
f 2   vr Yv'r vr   Y Y'  c  c0  vr   Yr Yr' r    Y   c  c0  r  
The Abkowitz (1980) model is modified in order to  2 
make the modelling more flexible and realistic phys-  Y Y'   Yrrv Yr'2v r 2 vr   Yeee Ye'3 e3   Y0 Y0'
r r
ically. In this study, the current effect is considered
as the main external excitation, because the ship (6)
model has a small above water structure. Figure 9
presents the coordinate frames for 3DOF ship mo-  r

f3   N0 N 0'   Nv N v' r vr   N N'  c  c0  vr   N N'    N rrv N r' 2v r 2vr 
r r

tions.  1 
 Neee N e' 3 e3   Nr N r' r    N N'  c  c0  r 
 2 
(7)
Detailed information about symbols used in equa- 3 NUMERICAL SIMULATIONS
tions (4),(5),(6) and (7) can be found in Moreira et
al. (2007). This section presents results of numerical simula-
tions for the collision avoidance system including
2.4.2 Line-of-Sight Algorithm the guidance and control for the own and target ves-
sels. The own and target ships are simulated consid-
According to the approach presented in Caccia ering the same vessel, based in particularities of the
(2006), the basic automatic guidance capabilities, i.e. “Esso-Osaka” tanker ship
PID auto heading and line-of-sight (LOS) guidance,
is proved to be sufficient for an unmanned surface
vehicle to satisfactorily accomplish its operational 3.1 “Esso Osaka” ship model
goal. This sub-section describes the model of the “Esso
In the LOS algorithm the desired geometric path Osaka” ship. For the verification of the guidance and
is composed by a collection of way-points pk in a control designs, a good mathematical model of the
way-point table. The LOS position is located some- ship is required to generate typical input/output data.
where along the straight-line segment connecting the The dynamics of the “Esso Osaka” tanker ship is de-
previous pk 1 and current pk way-points as shown in
scribed based on the horizontal motion with the vari-
Moreira et al. (2007). Thus, let the ship’s current
horizontal position p=[x,y] be the centre of a circle ables of surge, sway and yaw. The model was scaled
with radius of n ship lengths (nLpp). This circle will 1:100 from the real “Esso Osaka” ship. The vehicle
intersect the current straight-line segment at two main characteristics are listed in Table 2. The non-
points where pLOS   xLOS , yLOS  is selected as the dimensional hydrodynamics coefficients are present-
point closest to the next way-point. To calculate ed in Moreira et al. (2007).
pLOS , two equations with two unknowns must be
solved online. These are: Table 2 “Esso Osaka” model particulars
“Esso Osaka” MODEL
(y LOS  y) 2  (x LOS  x) 2  (nL pp ) 2 (8) Overall Length(mm) 3430
Length between perp. 3250
ylos  yk 1 yk  yk 1 (9) Breadth(mm) 530
  tan( k 1 ) Draught (estimated at the tests) (mm) 217
xlos  xk 1 xk  xk 1 Displacement (estimated at trials) (kg) 319.4
Rudder area (m²) 0.0120
The Eq. (8) is recognized as the theorem of Py- Propeller area (m²) 0.0065
thagoras, while the second equation states that the Scaling coefficient 100
slope of the path between the previous and current
way-point is constant.
Selecting way-points in the way-point table relies 3.2 Numerical simulations
on a switching algorithm. A criterion for selecting In this, sub-section numerical simulations of the col-
the next way-point, located at pk 1  [x k 1 , yk 1 ] , is lision avoidance system for two different collision
for the ship to be within a circle of acceptance of the scenarios. The grid map is a rectangular map of
current way-point pk. Hence, if at some instant of 500x700 [m], where 1 pixel is equal to 1m, It is as-
time t the ship position p(t) satisfies, sumed that the position and speed of the target ves-
sel is always available.
(x k  x(t)) 2  (y k  y(t)) 2  Rk2 (10)
Table 3 shows the model configuration for the colli-
sion scenarios. The own vessel starts at the origin
the next way-point is selected from the way-point with a constant heading pointed the North. The tar-
table. Rk denotes the radius of the circle of ac- gets vessel starts at an arbitrary position in order to
ceptance for the current way-point. It is imperative simulate collision, with a constant speed and head-
that the circle enclosing the ship has a sufficient ra- ing.
dius such that the solutions of Eq. (10) exist. There-
fore, nLpp  Rk , for all k is a necessary bound. Table 3 Model configurations for collision avoidance tests
Case Vessel 1 Vessel2
Two independent PID controllers with the gains A Start point (m) (0,0) (250,550)
obtained with the pole placement method were used Initial heading (ᵒ) 0 225
for controlling the speed and the heading, as was Des. target (m) (0,500) (–100,100)
presented in Moreira et al. (2007). Speed (m/s) 0.5 0.6
B Start point (m) (0,0) (-350,350)
Initial heading (ᵒ) 0 90
Des. target (m) (0,500) (250,350)
Speed (m/s) 0.5 0.6
Fig. 10 presents the trajectories of the own vessel and regulation, the own vessel pass through the
and target vessel during the execution of the colli- portside.
sion avoidance scenario in case A. In this figure, the
own vessel trajectory is on blue line and the target
vessel in orange. The desired trajectories are plotted
in red dotted lines, for each vessel. From the plot is (a)
possible to see the modification of the ship heading
of vessel 1 when a collision situation is detected.
According to COLREGs rules and regulation, the
own vessel pass through the portside.

(b)

Figure 11 Distance and speeds of vessels in simulation A.

Figure 10 Simulation of collision situation A

Figure 11a presents the time series of distance be-

tween own and target vessel during the execution of
the collision avoidance task in case A. From this plot
is possible to see a constant decrease of distance be-
tween ships at begin of simulation, it is because the
trajectories are in an imminent collision situation,
However, after the collision situation is detected, the
ship changes in order to avoid collision and the dis-
tance increases. Figure 11b, presents speeds of each
vessel, the speed of the own vessel varies in order to
avoid a potential collision, according to the
COLREGs rules and regulations. The target vessel Figure 12 Simulation of collision situation B
has a constant speed during whole simulations.

Fig. 12 presents the trajectories of two vessel during Figure 13a presents the time series of distance be-
the execution of the collision avoidance scenario in tween own and target vessel during the execution of
case B. In this figure, the own vessel trajectory is on the collision avoidance task in case B. From this plot
blue line and the target vessel in orange. The desired is possible to see a constant decrease of distance be-
trajectories are plotted in red dotted lines, for each tween ships at begin of simulation, it is because the
vessel. From the plot is possible to see the modifica- trajectories are in an imminent collision situation,
tion of the ship heading of vessel 1 when a collision However, after the collision situation is detected, the
situation is detected. According to COLREGs rules ship changes in order to avoid collision and the dis-
tance increases. Figure 13b, presents speeds of each
vessel, the speed of the own vessel varies in order to
avoid a potential collision, according to the
COLREGs rules and regulations. The target vessel
has a constant speed during whole simulations.

(a)

Figure 14a Place for simulations

(b) Lisbon

Figure 13 Distance and speeds of vessels in simulation B.

Figure 14b Place for simulations

4 CASE STUDY

In this case study, the marine traffic in estuary of

Tagus river in Lisbon is chose as place for simula-
tions. The objective is test the capabilities of the col-
lision avoidance and guidance system in a grid map
with real dimensions, considering a real marine envi-
ronment, For this purpose the marine routes of two
“Transtejo” vessels are simulated.
Figure 15 (a) Esso-osaka model (b) “Transtejo” vessel
4.1 Place of tests
The place chosen for the numerical simulation is 4.2 Results
the estuary of the Tagus river in Lisbon. The area
In order to study the performance of collision avoid-
chosen for the simulation is a square area of 4Km
ance, guidance and control system in a real collision
side length, Figure 14a,b. This area includes the
scenario, an aerial picture of the estuary of Tagus
“Cais do Sodré” dock in Lisbon, Portugal and two
river is pre-processing to create the grid map. Fig-
routes of river transport, i.e. “Cacilhas-Cais do
ure16 presents a bit map of the grid map in Rhinoc-
Sodré” and “Seixal-Cais do Sodré”.
eros software. Figure 16 also shows the scaled map
Figure 15a,b shows the “Esso-Osaka” model, used in
for the simulations. In the grid map 1 [pixel] is equal
for numerical simulations and the “Transtejo” ves-
to 10 [m].
sel, this vessel is used for river transport. In this
Table 4 presents the model configuration of the nu-
study the “Transtejo” vessels are modelled as “Esso
merical simulation. In this configuration, the starting
Osaka” scaled model
point of the “Transtejo” vessel is located in “Cais do
Notice that in this case study the influence of
Sodré” dock, Lisbon and the target point are a locat-
wind speed, river waves and others environmental
ed Dock in “Cacilhas” and “Seixal” docks. The start-
disturbances are neglected.
ing point of the own vessel is an arbitrary point in
the middle of “Tagus” river and the target point is in
a Navy dock in Lisbon.
 Figure 16 Generation of the grid map from aerial picture

Table 4 Model configurations for case study simulation

ASV “Transtejo” “Transtejo”
1 2
Start point (pixel) (25,250) (235,418) (436,44)
Initial heading (ᵒ) 60 100 340
Des. target (pixel) (425,436) (180,190) (235,418)
Des speed (pix./s) 0.5 0.6 0.6

Figure 17 presents the sequence motions of collision

avoidance manoeuvres of an ASV in a complex ma-
rine environment including static and dynamic ob-
stacles. In this plots the trajectory of ASV is plot in
red, the “Transtejo” 1 ship in brown and “Transtejo”
2 ship in orange. The desired trajectories are plotted
in red dotted lines, the desired waypoints are repre-
sented by a black circle, and the static environment
is plotted in blue asterisks. From this sequence it is
clear to see the heading angle changes in order to
avoid potential collision situation. At the begin of
simulation the ASV is tracking the desired path,
however, after some minutes the first “Transtejo” 1
ship is detected and a potential collision is identified,
Thus, according to the COLREGs rules and regula-
tions, the ASVs starts the collision avoidance ma-
noeuvre. Later, a second collision situation is detect-
ed, with the “Transtejo” 2 ship and the ASV, once
more; modify his heading angle in order to avoid the
imminent collision. Finally, when the collision risk
disappears, the ASVs continues to tracking the de-
sired path.

Figure 18a presents the time series of distance be-

tween ASV and “Transtejo” 1, in blue, ship and
ASV and “Transtejo” 2 ship, in green, during the ex-
ecution of the collision avoidance task. From this
plot is possible to see a constant decrease of distance
between ships at begin of simulation, it is because
the trajectories are in an imminent collision situa-
tion, However, after the collision situation is detect-
ed, the ship changes in order to avoid collision and
the distance increases. Figure 18b, presents speeds of

Figure 17 sequence motion of collision avoidance and path-

following algorithm
each vessel, the speed of the own vessel varies in or- tre for Marine Technology and Ocean Engineering
der to avoid a potential collision, according to the (CENTEC).
COLREGs rules and regulations. The target vessel This work was performed within the Strategic Re-
has a constant speed during whole simulations. search Plan of the Centre for Marine Technology and
Ocean Engineering (CENTEC), which is financed by
Portuguese Foundation for Science and Technology
(Fundação para a Ciência e Tecnologia-FCT). The
authors are grateful to H. Xu for his help and advises
(a) in guidance of vessels.

REFERENCES

Abkowitz, M. A. (1980). Measurement of hydrodynamic

characteristics from ship maneuvring trials by system
identification. SNAME Transactions, 88, 283–318.
Caccia, M., 2006, “Autonomous surface crafts: prototypes and
basic research issues”, in Proc. 14th Mediterranean
Conference on Control and Automation, Ancona, Italy,
pp. 1–6.
Benjamin, M. R., & Curcio, J. A. (2004). COLREGs-based
navigation of autonomous marine vehicles. Autonomous
(b) Underwater Vehicles, 2004 IEEE/OES (pp. 32-39).
IEEE.
Ferrari, V., Perera, L. P., Santos, F. P., Hinostroza, M. A.,
Sutulo, S., & Guedes Soares, C. (2015). Initial
experimental tests of a research-oriented self-running
ship model. Guedes Soares, C. & Santos T.A. (Eds.)
Maritime Technology and Engineering, Taylor & Francis
Group, London, UK, 913-918.
Figure 18 Distances between AVS and “Trantejo” vessels Fossen, T. I. (2011). Handbook of Marine Craft
and Speeds Hydrodynamics and Motion Control. John Wiley & Sons
Ltd.
Hasegawa K (1987) Automatic collision avoidance system for
5 CONCLUSIONS ship using fuzzy control. In: Proceedings of 8th ship
control system symposium, pp 234–258
He Y., Jin Y., Huang L., Xiong Y., Chen P., Mou J., (2017),
A system for collision avoidance, guidance and Quantitative analysis of COLREG rules and seamanship
control for operation of ASVs in a complex marine for autonomous collision avoidance at open sea, Ocean
environment was presented. The fuzzy logic based Engineering, V.140,pp. 281-291.
intelligent decision-making algorithm was used for Hinostroza M.A., Xu Haitong & Guedes Soares C. (2017),
collision avoidance. The guidance unit employs the Path-planning and path-following control system for
LOS algorithm and, the control unit is composed by autonomous surface vessel, Maritime Transportation and
a PID heading controller plus a speed controller. Harvesting of Sea Resources, Taylor & Francis Group,
Two different collision scenarios were simulated London,pp.991-998.
in order to study the collision avoidance system per- IMO (1972) Convention on the international regulations for
preventing collisions at sea (COLREGs).
formance and a good results were found.
http://www.imo.org/conventions/
From the case-study simulation, a good perfor- KongsbergMaritime, 2017. Autonomous ship project, key facts
mance of the system was found in real marine envi- about YARA Birkeland. URL, https://www.km.kongs
ronment. berg.com/ks/web/nokbg0240.nsf/AllWeb/4B8113B707A
Future improvements are to increase the number 50A4FC125811D00407045?OpenDocument
of collision scenarios and more vessels. Expand the Moreira, L., Fossen, T.I., and Guedes Soares, C. 2007, “Path
collision avoidance, guidance and control unit for a Following Control System for a Tanker Ship Model”.
fleet of ships. Introduce disturbances i.e. wind and Ocean Engineering. 34:2074–2085.
waves. Moreira, L., Santos, F.J., Mocanu, A., Liberato, M., Pascoal,
R., and Guedes Soares, C., 2008, “ Instrumentation used
in guidance, control and navigation of a ship model”, 8th
Portuguese Conference on Automatic Control, Vila Real,
ACKNOWLEDGEMENTS Portugal, pp. 530–535.
Njus E. (2016), The military's Oregon-built drone ship is head-
The first author has been funded by a PhD schol- ed to California, Oregon Business News, 7 April 2016
arship from the University of Lisbon and from Cen- Pedrycz W, Gomide E (2007) Fuzzy systems engineering to-
ward human centric computing. Wiley, Hoboken
Perera L. P., Carvalho J. and Guedes Soares C. (2011). Fuzzy-
logic based decision making system for collision
avoidance of ocean navigation under critical collision
conditions. Journal of Marine Science and Technology,
16(1), 84–99.
Perera, L. P., Ferrari, V., Santos, F. P., Hinostroza, M. A., and
Guedes Soares, C. (2015). Experimental Evaluations on
Ship Autonomous Navigation and Collision Avoidance
by Intelligent Guidance. IEEE Journal of Oceanic
Engineering, 40(2), 374–387.
Rachel Courtland, “DARPA’s Self-Driving Submarine Hunter
Steers Like a Human,” IEEE Spectrum, April 7, 2016,
http://spectrum.ieee.org/automaton/robotics/military-
robots/darpa-actuv-self-driving-submarine-hunter-steers-
like-a-human.
Rolls-Royce, “Autonomous ships: The next step,” White
Paper,Available: http://www.rolls-
royce.com/~/media/Files/R/Rolls-Royce/
documents/customers/marine/ship-intel/rr-ship-intel-
aawa-8pg.pdf
Sato Y., Ishii H., (1998), Study of a collision-avoidance system
for ships, Control Engineering Practice, Vol.6,pag. 1141-
1149,
Statheros, T., Howells, G., and McDonald-Maier K., (2008),
“Autonomous ship collision avoidance navigation
concepts, technologies and techniques,” The Journal of
Navigation, Vol. 61, pp. 129-142
Xu, H.,Hinostroza M.A, & Guedes Soares, C. (2018).
Identification of hydrodynamic coefficients of ship
nonlinear manoeuvring mathematical model with free
running model tests., International Journal of Maritime
Engineering,Accepted for publication
Zhang J., Zhang D., Yan X., Haugen S., Guedes Soares C.,
(2015) A distributed anti-collision decision support
formulation in multi-ship encounter situations under
COLREGs, Ocean Engineering, Vol. 105, Pag. 336-348.

View publication stats