|
6. Conclusion
In this article, we have surveyed most recent work on path planning
and obstacle avoidance for AUV. We introduced the kinematic and
dynamic model of AUV and give an overview of work on path planning
for AUV and technical details of some representative algorithms that
deal with constraints and characteristics of AUV and the influence
of marine environments. We also provide an in-depth discussion and
comparison between different path planning algorithms, and propose
some potential future research directions that are worthy to investigate
in this field.
Path planning methods of AUV are mainly divided into two cat-
egories: global path planning with known static obstacles and local
path planning with unknown and dynamic obstacles. When there is
a global map about everything including obstacles before planning a
path, e.g., the location and contour of static obstacles can be mea-
sured or obtained beforehand, AUV can use this information to find
a collision-free path between the starting point and the target point
in advance with global path planning methods. Global path planning
for AUV mainly includes
𝐴
∗
algorithm, genetic algorithm, differential
evolution algorithm, particle swarm optimization algorithm and ant
colony optimization algorithm.
𝐴
∗
algorithm is suitable for working in
a simple and small-scale marine environment, since a large number of
nodes need to be calculated in large-scale environments, resulting in
a low searching efficiency. AUV path planning with genetic algorithm
needs to have a large storage space and powerful computing system
since a large number of evolutionary operations to be carried out and
more parameters to be updated. The differential evolutionary algorithm
is similar to the genetic algorithm and is verified to be more robust
than the genetic algorithm in a wide range of AUV path planning
applications. The ant colony optimization algorithm and particle swarm
optimization algorithm can adapt to the environment in a shorter time.
The particle swarm optimization algorithm has fewer parameters and
certain memory functions, greatly reducing the searching time.
In dynamic and uncertain underwater environments, local path
planning methods are needed to avoid unknown and dynamic obsta-
cles by obtaining the local environmental information with sensors
in real time, on top of a planned path from a global path planner.
Common local path planning methods include RRT, artificial potential
field, fuzzy logic algorithm, neural network, reinforcement learning
and deep reinforcement learning etc. The RRT algorithm can be easily
extended to the unexplored area, which is very suitable for solving path
planning problems in high-dimensional space. However, the real-time
performance of RRT is not too high compared with other methods.
Artificial potential field plays an important role in real-time obstacle
avoidance for AUV. However, when the distance between two obstacles
is too close or there are obstacles near the target point, it may fail
to find the direction to travel and easily fall into the local minimum
point since it does not consider the constraints of dynamics of AUV and
the obstacle size. The fuzzy logic algorithm was shown to be robust in
dealing with practical problems and has been widely used in AUV to
avoid unknown and dynamic obstacles. It does not need an accurate
mathematical model, and is suitable for solving highly complex and
nonlinear problems. However, fuzzy logic algorithm requires human
experts to be fully familiar with the operation mechanism of AUV and
establish an appropriate rule base, and might not work well in unknown
and uncertain environments. AUV path planning with neural network
can store empirical knowledge and deal with nonlinear mapping prob-
lems by learning autonomously with simple rules. Traditional neural
networks need to collect samples before learning, which is a very time-
consuming process and might be difficult or even impossible in many
situations for AUV path planning. Bio-inspired neural network does
not need any pre-training process and is very suitable to deal with
unknown dynamic environments. AUV with reinforcement learning
(RL) can plan an optimal path without any prior knowledge and work
well in complex and fully uncertain environments. Deep reinforcement
learning (DRL) further implements an end-to-end learning to map the
original sonar image to the action of AUV, which allows AUV to learn
to plan an optimal path in high-dimensional and uncertain marine
environments. RL and DRL can make use of samples in simulation
and human knowledge to solve the sample inefficiency problem for
applying to actual AUV platforms.
Ocean Engineering 235 (2021) 109355
15
C. Cheng et al.
Do'stlaringiz bilan baham: | 
