make the speed of planning for a path very slow.
Ocean Engineering 235 (2021) 109355
5
C. Cheng et al.
Fig. 4.
An illustration of one iteration in the genetic algorithm.
3.2. Genetic algorithm
Inspired by the natural selection and evolution, the genetic algo-
rithm is proposed and has been widely used to solve various opti-
mization problems. The genetic algorithm works by initializing the
population randomly, selecting the suitable individuals with a fitness
function, carrying on the selection, crossover, mutation genetic oper-
ations, and constantly updating the population (
Haupt and Werner
,
2007
).
Fig. 4
shows one iterative process of the genetic algorithm.
Through operating on the object directly, the genetic algorithm can
obtain and guide the search space automatically without defining rules.
Restricted by genetic algorithm, researchers usually modify the
algorithm in terms of the genetic operator, evaluation factor or indi-
vidual selection of population to adapt to the characteristics of AUV.
For example,
Alvarez and Caiti
(
2001
) applied genetic algorithm to
path planning of underwater vehicle. In their method, a new ge-
netic operator is added to the algorithm to make it converge to the
global minimum when there are different minimum values in the
ocean flow field. To complete the obstacle avoidance task under strong
currents,
Alvarez et al.
(
2004
) proposed an improved genetic algorithm.
In their method, an iterative operator is added to change the initial
population, and a random migration operator is added to improve the
mutation rate. Their results show that AUV can obtain the path with
minimum energy consumption.
Zhang
(
2006
) designed a hierarchical
path planning method based on the genetic algorithm. Firstly, the
AUV workspace is decomposed into obstacle region and free region by
octree. Then the genetic algorithm is used to search the path of the free
region. According to the current information provided by electronic
charts, an adaptive function is established for genetic algorithm
Sun
and Zhang
(
2012
). In their method, the ocean current is used as an
evaluation factor of the adaptive function in the algorithm, which saves
the energy consumption of AUV.
Li et al.
(
2013
) show that adding
a node deletion operator and a smoothing operator into the genetic
algorithm can get a better path. In addition,
Cao et al.
(
2016
) improved
the initial population generation method of the genetic algorithm and
designed tangential operators to greatly improve the convergence speed
of AUV path planning.
Yan and Pan
(
2019
) proposed an improved
genetic algorithm to solve the premature problem in the traditional ge-
netic algorithm. In their method, the population entropy is used as the
diversity evaluation standard, and an adaptive method is used to adjust
the probability in crossover and mutation. In addition, considering the
kinematic characteristics of AUV as shown in Section
2
, the individuals
who are close to obstacles or do not conform to the actual attitude of
AUV are eliminated in the genetic algorithm.
The biggest advantage of genetic algorithm is that it does not need
to know how to find the optimal solution before planning. The genetic
algorithm has been successfully applied to AUV path planning in en-
vironments with known static obstacles. However, if the parameters
of the algorithm were not adjusted properly, some problems might
happen, e.g., individual peak value or the planning speed might be very
slow.
Do'stlaringiz bilan baham: