Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

- Unmanned Surface Vehicle (USV)
Research and Implementation of USV by Adaptive performance test in Bintan Island Waters
R Fauzi, I Jaya and M Iqbal
Collision Avoidance Algorithm in Mixed Scenarios - Dynamic modelling and controlling
Unmanned Surface Vehicle
F A Setiawan, R E A Kadir, N Gamayanti
To cite this article: Pei Lin Hu et al 2023 J. Phys.: Conf. Ser. 2493 012003 et al.

- Research on Redundant Design

Technique for USV Autonomous Control
System
Wenbin Huang, Hao Xu, Jiaxiang Feng et
View the article online for updates and enhancements. al.

This content was downloaded from IP address 152.39.246.92 on 16/05/2023 at 14:47

AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

Research and Implementation of USV by Adaptive Collision

Avoidance Algorithm in Mixed Scenarios

Pei Lin Hu1,2,3,4, Jin Chao Xiao4, a, Jun Feng Xiong4, and Jin Qing Liu1,2,3, *
1
College of Photonic and Electronic Engineering, Fujian Normal University, Fuzhou,
350007, China
2
Key Laboratory of Optoelectronic Science and Technology for Medicine, Ministry of
Education, Fujian Normal University, Fuzhou, 350007, China
3
Fujian Provincial Key Laboratory of Photonic Technology, Fujian Normal
University, Fuzhou, 350007, China
4
Guangzhou Institute of Industrial Intelligence, Guangzhou, 511458, China
a
Email: [email protected]
*
Corresponding author：[email protected]

Abstract. During the mission of the USV (Unmanned Surface Vehicle), there may be some static
or dynamic obstacles on the pre-planned track that are not displayed on the ground station. To
avoid these obstacles in time, a model of DCPA (distance to the closest point of approach)
between the USV and the obstacle is proposed, which is used to plan the speed of the USV, and
adaptively accelerate and decelerate can safely avoid dynamic obstacles. We solved the problem
of target unreachable and heading change optimization by planning the heading at IAPF
(improved artificial potential field) in the process of sailing, which can effectively avoid static
obstacles and respond to collision avoidance in emergencies. The collision avoidance planning
outputs a set of behavior vectors (speed, heading), and hands it over to the USV controller to
control the navigation of the USV. In this study, an adaptive speed and heading planning method
is used to improve the collision avoidance ability of the USV in the mixed scene of dynamic and
static obstacles. Through multiple launching experiments, the actual collision avoidance effect
of the method is verified, and the safety and reliability of the USV application requirements are
met.

1. Introduction
In recent years, with the increasing application requirements of the USV, such as marine resource
detection, reconnaissance, hydrometeorological detection, environmental monitoring, maritime search,
and rescue, etc [1], which has played a significant role in scientific research, military, and civilian fields
[2]. During the mission in real and complex sea areas, there may be some unknown static and dynamic
obstacles on the pre-planned track. However, there is a problem faced by the USV to perform the mission
smoothly, that is how to reach the target point safely and smoothly.
To this end, domestic and foreign researchers have done some related research. The literature [3]
proposes an improved method of APF (artificial potential field) to solve the problem that the trajectory
swings when avoiding obstacles. The literature [4] proposes a global static obstacle path planning
method with an improved heuristic function, which improves the algorithm's accuracy. In [5], a method

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

is proposed to calculate an optimal and collision-free path from a set of path harnesses, which improves
the comfort and safety of USV navigation. Wu et al. [6] proposed a speed obstacle method to get a
reasonable navigation area by calculating the angular relationship between the relative speed and relative
position of the USV and the obstacle. In [7], the error and covariance of the tracking filter which can
estimate the motion information are used to model and calculate the collision probability and set the
safe collision area. Kuwata et al. [8] ensured that the multi-rules under the multi-objective apply to
judgment and avoidance at the same time.
In the above studies, Zhou et al. [3] [4] [5] were only suitable for avoiding static obstacles and lacked
the avoidance of dynamic obstacles. Wu et al.[6] [7] [8] have flaws in track optimization in actual tests.
This research aims to consider both static and dynamic obstacles and proposes a fusion decision-making
method based on speed planning with an adaptive DCPA model and heading planning with IAPF, which
solves the unreachable target [9] and excessive change of heading problems of TAPF (traditional
artificial potential field). The collision avoidance planning outputs a set of behavior vectors (speed,
heading), and hands it over to the USV controller to control the navigation of the USV through LOS
(line of sight) guidance [10]. It has been verified by multiple launching tests that it is always kept a safe
distance from the USV to the obstacle, which improves the navigation safety, track optimization of the
USV, and collision avoidance generalization in mixed obstacle scenarios.

2. Heading planning by IAPF

To solve the problems of unreachable targets and too large a change of heading of the USV by TAPF,
an IAPF with adaptive parameters was proposed. Therefore, an improved repulsion force related to the
target point is introduced into the traditional repulsion potential field. Meanwhile, we don't change the
traditional attractive potential field, whose function is:
 2 ( q, qg ) (1)
1
U att ( q ) =
2
The improved repulsive potential field function is:
1  1 
2

  ( q, qg ) , s   ( q, qo )   o
1
  − n

 2   ( q, qo )  o 
 (2)
U rep ( q ) =  , 0   ( q, qo )  s
 0,  ( q, qo )   o



Where  ,  are the attractive and repulsive force gain coefficient respectively, q, qg , qo are the
current position ( x, y ) , target position (x , y )
g g and obstacle position ( xo , yo ) .  ( q, qo ) is the
position vector of the USV relative to the obstacle,  o is the influence radius of the obstacle, s is the
safe distance between the USV and the obstacle, n is the adjustment factor, 0  n  1 ,  (q, qg ) is
the position vector of the target relative to the USV,  n q, qg ( ) is the position vector of the target
relative to the USV after adding the adjustment factor.
Since the potential field function grows and changes fastest along the gradient direction, so the
attractive force can be calculated from the negative gradient of the attractive potential field function:
Fatt ( q ) = −U att ( q ) =  (q, qg ) (3)
The improved repulsive force can be calculated from the negative gradient in the same way. When
the USV enters the influence of the obstacle and is within a safe distance, there is a repulsive force
component:
 1 1  1
Frep1 ( q ) =   −  2  n (q, qg ) (4)
  ( q , qo )  o   ( q , qo )
The direction is from the obstacle to the USV, and another repulsive force component is:

2
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

1 1 1 n
Frep2 ( q ) = −  ( − )2 (5)
2  ( q, qo ) o  1− n (q, qg )
The direction is from the USV to the target. If there is less than a safe distance from the USV to the
obstacle, the repulsive force should be  , and the direction is directed from the obstacle to the USV.
In other cases, the repulsion force should be zero.
 Frep1 ( q ) + Frep2 ( q ) , s   ( q, qo )  o
 (6)
Frep ( q ) =  ,0   ( q, qo )  s
 0,  ( q, qo )  o

Frep1 ( q ) and Frep2 ( q ) are shown in Figure 1. When the USV gradually approaches the target,
 n (q, qg ) → 0 , Frep1 ( q ) → 0 and the value of Frep2 ( q ) is related to n . When n = 0 , Frep2 ( q ) = 0 ,
that is, the repulsive force on the USV is the traditional repulsive force. When n =1 ,
− )2 , this is a certain value. When 0  n  1 , Frep2 ( q ) →  .Therefore, as long
1 1 1
Frep2 ( q ) = −  (
2  ( q, qo ) o
as n  0 , the USV can reach the target safely and smoothly, which solves the problem that the target
cannot be reached by the TAPF.
Compared with the traditional repulsive force, after the improved repulsion force is introduced, the
angle between the repulsion force and the attractive force is smaller, which can solve the problem of the
excessive change of the heading in actual navigation by adjusting the coefficient.

Figure 1. The improved potential field force.

3. Speed planning by adapting the DCPA

3.1. The DCPA model

Before the USV avoids the obstacle, it must first judge whether the USV collides with the obstacle.
DCPA is a good basis for judgment, it is a description of the urgency of the meeting.
As is shown in Figure 2, in the USV body coordinate system, the axis y is the positive direction of
the forward speed of the USV, the axis x is the positive direction of the lateral speed of the USV, the
USV sails at the speed vu at the position U ( 0,0 ) , and the heading is  u . The obstacle moves with
speed vo at the position O( xo , yo ) , the direction is o . The speed of the obstacle relative to the USV
is v , and the direction is   . CPA is the closest meeting point between the USV and the obstacle.
Among them, vu with its along-axis x, y component vux , vuy can be obtained through GPS, xo , yo
and vo with its along-axis x, y component vox , voy is directly obtained by radar detection, v with
its along-axis x, y component vx , vy can be deduced.  is the azimuth angle of the obstacle
relative to the USV heading, and d is the distance from the USV to the obstacle.

3
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

Figure 2. The DCPA model.

First, the azimuth function is introduced:
 x
 arctan , x  0, y  0
y

 x
 2 + arctan y , x  0, y  0

 x (7)
  + arctan , x  0, y  0

 ( x, y ) =  y
  + arctan x , x  0, y  0
 y

 
 , x  0, y = 0
2

 3
, x  0, y = 0

 2
From this, we get:
u =  ( vux , vuy )


o =  ( vox , voy ) (8)

  =  ( vx , vy )
Assuming the speed of the USV and the obstacle remain constant, we have:
  
 d , −      +
 2 2 (9)
dcpa = 

 d sin ( −  −  ) ,   − or   + 


 
2

2

3.2. Adaptive speed planning

3.2.1. Speed planning for static obstacles. For static obstacles, the DCPA is the lateral distance from the
USV to the obstacle. When the USV enters the influence range of the obstacle, we can divide the
navigation area into a deceleration area and an obstacle avoidance area according to the different lateral
distances. When the lateral distance is greater than or equal to the DCPA threshold, the USV enters the
deceleration zone otherwise, enters the obstacle avoidance zone, as is shown in Figure 3:

4
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

Figure 3. The USV sailing zone for the static obstacle.

3.2.2. Speed planning for dynamic obstacles. For dynamic obstacles, the DCPA threshold is set as d0 ,
( )
when d sin   −  −  = d0 , then the following relation can be obtained:
vox − vuxc
=A (10)
voy − vuyc
then:
vox − vuxc
vuyc = voy − (11)
A
 d 
Where A = tan  arcsin 0 +   , vuxc and vuyc are the critical values that the USV forward speed
 d 
and lateral speed need to reach at this time, and we assume that vuxc (lateral drift speed) remains
unchanged in the USV sailing mission, then we can get:
v −v
vuyc = voy − ox ux (12)
A
So, we can get a collision avoidance strategy by adapting the forward speed threshold.
 
In the first case:  −   + , we can get: dcpa = d , and if d  d 0 , the USV sails normally,
2 2
otherwise the USV adopts the IAPF (improved artificial potential field) to avoid obstacles urgently, and
the influence radius of obstacles is the DCPA threshold.
 
In the second case:     +
2
or    −
2
(
, we can get: dcpa = d sin   −  −  , and if )
dcpa  d0 , the USV sails normally. Otherwise, the USV must make a speed change to reach the DCPA
threshold.

4. Experimental results and analysis

The size of the USV in this study is 0.9 m * 0.5 m * 0.4 m, and the normal sailing speed is about 1.5
m/s. For the convenience of the experiment and research, the obstacle in this study is assumed to be
spherical, with a radius of about 0.5 m.
Considering that the static obstacle is in a stable motion state, after several experiments, it is more
appropriate to set the DCPA threshold of 2 m, and the obstacle influence radius of 5 m. The experimental
results of TAPF and IAPF are shown in Figure 4 and Figure 5. When the USV is less than 5 m away
from the obstacle, it starts to decelerate, and if the DCPA is less than 2 m at this time, the heading must
be changed. From the comparative experiment, we could notice that the closest distance from the USV
to the obstacle is about 2.78 m by the improved method, while the traditional method is only about 1.56
m, from the collision avoidance trajectory of the USV, the improved method has a better trajectory
smoothness, and the variation of the heading angle is reduced, which improves the safety and trajectory

5
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

optimization of the USV.

Figure 4. The traditional method for static obstacles.

Figure 5. An improved method for static obstacles.

To ensure navigation safety, it is appropriate to set 4 m as the DCPA threshold considering the
complex motion state of dynamic obstacles and 1.5 s delay response of radar detection about obstacles
after several experiments, and the DCPA threshold is the influence radius of the obstacle. There are two
different obstacle avoidance actions encountered by the TAPF method, as shown in Figure 6 and Figure
7. Since the radar is installed directly in front of the USV, the radar's horizontal detection angle range is
−45 ~ 45 . If the obstacle can be detected, indicating that the USV has not avoided the obstacle at this
time. As is shown in Figure 6, when the obstacle comes, the distance is far away and the reaction is
delayed. When the obstacle avoidance action occurs, the obstacle has no collision danger. As shown in
Figure 7, when the obstacle comes, the distance is relatively close and the response lags. The USV
follows the obstacle around once before returning to the original track, which is the same as dealing
with static obstacles. And the minimum distance of the USV from the obstacle is less than the safe
distance, which is a high safety risk.

6
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

Figure 6. The traditional method for dynamic obstacles in the first case.

Figure 7. The traditional method for dynamic obstacles in the second case.

Compare with the improved method shown in Figure 8. The USV calculates that the DCPA is less
than its threshold, so it decelerates in advance until the DCPA is greater than its threshold. At the same
time, there is always kept a safe distance of more than 5 m from the USV to the obstacle, which is
improved the safety and rationality of collision avoidance.

Figure 8. An improved method for dynamic obstacles.

5. Conclusion
It can be seen from the experimental results that based on the IAPF and DCPA model, we improve the
rationality and safety of the USV to avoid dynamic obstacles. In the face of emergencies, it can also

7
AIIM-2022 IOP Publishing
Journal of Physics: Conference Series 2493 (2023) 012003 doi:10.1088/1742-6596/2493/1/012003

make timely avoidance, and at the same time solve the optimization of unreachable targets and heading
changes in the face of static obstacles. We noticed that what was planned in this study is a set of behavior
vectors (heading, speed), which is still a reactive obstacle avoidance action, without considering the
historical information and the prediction of future motion trends of USV and obstacles. To further
improve the collision avoidance safety and trajectory optimization of USV in actual navigation, this is
also an important issue to be studied and discussed in the next step.

Acknowledgments
This work was supported by the Key-Area Research and Development Program of Guangdong Province
(2020B1111010002).

References
[1] Chen H, Zhang X Y and Jiang C F 2015 A Review of Surface Unmanned Boat Path Planning
Research (in Chinese). World Shipping, 38(11): 30-33
[2] Zhang W D, Liu X C and Han P 2020 Progress and Challenges of Overwater Unmanned Systems
(in Chinese). Acta Automatica Sinica, 46(5): 847-857
[3] Zhou Z X, Chen Z, Bao T, Zhang B and Zhang A T 2022 Obstacle Avoidance Based on IAPF
(improved artificial potential field) for Under-Actuated USV (in Chinese). Ship Engineering,
44(04): 19-23
[4] Yu L W, Ye X and Qian T H 2022 Obstacle Avoidance Strategy of Autonomous Unmanned
Watercraft on the Surface Using Heuristic Search Algorithm (in Chinese). Journal of Jianghan
University Natural Science Edition, 50(01): 79-86
[5] Hu Z H, Yang Z H, Liu X C, Zhang W D and Luo. R 2021 Radar-based maritime path planning
with static obstacles in a Frenet frame (in Chinese). Science China Technological Sciences,
51(11): 1401-1409
[6] Wu B, Xiong Y and Wen Y Q 2014 Automatic collision avoidance algorithm for unmanned surface
vessel based on velocity obstacles (in Chinese). Journal of Dalian Maritime University, 40(02):
13-16
[7] Park J, Choi J and Choi H 2019 COLREGS-compliant path planning considering time-varying
trajectory uncertainty of autonomous surface vehicle．Electronics Letters, 55(04): 222-224
[8] Kuwata Y, Wolf M T, Zarzhitsky D and Huntsberger T L 2011 Safe maritime navigation with
COLREGS using velocity obstacles. In IEEE Journal of Oceanic Engineering, 39(01): 110-
119
[9] Huo F C, Ren W and Liu D H 2016 Research on Path Planning Method Based on Improved
Artificial Potential Field Method (in Chinese). Techniques of Automation and Applications,
35(3): 63-67
[10] Han P, Liu Z L, Zhou Z C, Tang H, Ban L and Hao L L 2018 Path tracking control algorithm based
on LOS method for surface self-propulsion vessel (in Chinese). Applied Science and
Technology, 45(03): 66-70