|
1. Introduction
With recent advances in technology, Autonomous Underwater Ve-
hicle (AUV) is becoming more and more important in the exploration
and exploitation of ocean resources. AUV can complete a variety of
subsea tasks in civil and military fields, such as ocean pollutant moni-
toring (
Ramos et al.
,
2001
), mine hunting (
Hagen et al.
,
2003
), marine
biology exploration (
Sagala and Bambang
,
2011
), pipeline following
and inspection (
Liu et al.
,
2018
) and anti-submarine warfare, etc. The
autonomy of AUV is essential for it to operate in complex environments,
which has attracted wide attentions of researchers from all over the
world. In recent years, higher requirements are put forward for the
autonomy of AUV to meet the future development of the ocean. The
path planning and obstacle avoidance is the core technology to realize
the autonomy of AUV and will determine the application prospect of
AUV.
The path planning of AUV is considered as a series of translations
and forward angle changes from the starting position to the destination
according to certain optimization standards (
Zeng et al.
,
2016
). When
planning a path, most of the time AUV will operate in the environments
full of obstacles. There are many cases when the location and contour of
static obstacles can be measured or obtained beforehand. In these cases,
there is usually a global map about everything including obstacles
before planning a path. AUV can use this information to find a collision-
free path between the starting point and the target point in advance
∗
Corresponding author.
E-mail address:
guangliangli@ouc.edu.cn
(G. Li).
with global path planning methods. These global path planning meth-
ods can also be used in the barrier-free environments, while taking
the path length, energy consumption and current into account instead
of obstacle avoidance. In dynamic and uncertain underwater environ-
ments, it is difficult or even impossible to obtain the information of
various obstacles before path planning. In this case, there is usually
planned path from a global path planner and AUV still needs a local
path planner to avoid these unknown and dynamic obstacles, such as
ships, reefs, and moving animals or fish etc. Therefore, in this paper, we
divide the path planning methods of AUV into two categories: global
path planning with known static obstacles and local path planning with
unknown and dynamic obstacles, as shown in
Fig. 1
.
For global path planning of AUV with known static obstacles,
researchers have proposed many commonly used methods, such as
𝐴
∗
algorithm (
Dechter and Pearl
,
1985
), genetic algorithm (
Cobb
and Grefenstette
,
1993
), particle swarm optimization (
Eberhart and
Kennedy
,
1995
), differential evolution (
Storn and Price
,
1997
) and ant
colony optimization (
Dorigo et al.
,
1991
). With the increasing complex-
ity and uncertainty of the environments, the requirements for AUV path
planning are also becoming higher (
Cai et al.
,
2020
;
Mac et al.
,
2016
).
In this case, the obstacles encountered by AUV might be unknown and
even dynamic. Taking dynamic constraints into account, the rapidly-
exploring random trees algorithm is a suitable method proposed for
https://doi.org/10.1016/j.oceaneng.2021.109355
Received 16 December 2020; Received in revised form 9 June 2021; Accepted 16 June 2021
Ocean Engineering 235 (2021) 109355
2
C. Cheng et al.
Fig. 1.
Main path planning algorithms for AUV.
local path planning in high dimensional unknown environments (
Tan
et al.
,
2004
). In addition, the artificial potential field method (
Khatib
,
1986
) and fuzzy logic algorithm (
Cordón et al.
,
1996
) can use airborne
sensors to sense the surrounding environmental information in real-
time and deal with unknown static or dynamic obstacles. In recent
years, based on the development in deep neural network, some self-
learning methods like neural network (
Hansen and Salamon
,
1990
)
and reinforcement learning (
Sutton and Barto
,
1998
) were introduced
into the local path planning of AUV. Deep reinforcement learning
(
Mnih et al.
,
2015
), a combination of deep neural network and rein-
forcement learning, has also been proposed and applied in AUV local
path planning. These self-learning methods does not require any priori
knowledge of the environments and can deal with the problem of real-
time dynamic path planning. In addition, their planning time can be
greatly reduced after training.
Because of recent advances in technology and new breakthroughs
in the field of path planning for AUV, it is necessary to make a
comprehensive survey on it though several surveys already exist (
Zeng
et al.
,
2015
;
Li et al.
,
2018
). A most recent survey by
Okereke et al.
(
2020
) only focused on machine learning path planning methods for
AUV in terms of Underwater Internet of Things (UIOT). The objective of
this survey is to give an overview of work on path planning for AUV and
technical details of some representative algorithms, and also to discuss
some open problems to be solved in this area. The focus of this article
is put on these path planning algorithms that deal with constraints and
characteristics of AUV and the influence of marine environments. It also
provides an in-depth discussion and comparison between different path
planning algorithms.
The rest of this article starts with an introduction of the kinematic
and dynamic model of AUV in Section
2
, and an analysis of chal-
lenges and differences of path planning for AUV compared to other
mobile robots is also provided. Section
3
surveys and discusses global
path planning algorithms in environments with known static obstacles.
Section
4
describes and discusses local path planning methods with un-
known and dynamic obstacles. In Section
5
, we propose some potential
future research directions that are worthy to investigate in this field.
Finally, Section
6
concludes.
Do'stlaringiz bilan baham: | 
