Page 48 - Read Online
P. 48
Wang et al. Intell Robot 2023;3(4):538-64 I http://dx.doi.org/10.20517/ir.2023.30 Page 15 of 27
Prob = softmax(−w ) (32)
where w is the n-th column of W, and the softmax function is used to normalize the probabilities based
on optimizing benefits. The method above will be more purposeful to generate other feasible solutions
that are different from the local optimal solution during the iteration.
5.1.3. Improved chromosome operations
The fine-adjustment mechanism is used in the chromosome operations of traditional GA in IGAFA. Consid-
ering the characteristics of the priority-encoding method, the traditional crossover is divided into two types:
sexual and asexual crossover. The specific chromosome-operations are as follows:
(i) Sexual crossover
Sexual crossover refers to the crossover with two chromosomes as parents, which is similar to the tra-
ditional crossover in GA. Firstly, two random chromosomes are selected as parents, the chromosomal
location of the start of crossover is selected by Equation (29) with a random length , and then the
selected gene segments are replaced with each other.
{
parent ( ), ≤ < +
1
child 1 ( ) =
parent ( ), ℎ
2
{
parent ( ), ≤ < + (33)
2
child 2 ( ) =
parent ( ), ℎ
1
= 1, 2, ,
(ii) Asexual crossover
Asexual crossover refers to the crossover with only one chromosome as a parent, which is used to adjust
the access priority or location of destination cells. There are four crossover operators defined as follows:
a. Swap: Select two different genes and by Equation (29) and swap these genes to exchange the
priority.
), =
parent(
),
child( ) = parent( =
(34)
parent( ), ℎ
= 1, 2, ,
b. Insert: Selecttwodifferentgenes and , byEquation(29), theninsertthefirstgeneintothelocation
of the second gene to raise or lower the priority.
),
parent( =
parent( +1 ),
< and ≤ <
child( ) =
parent( −1 ), > and < ≤ (35)
parent( ), ℎ
= 1, 2, ,
c. Delete: Select one gene by Equation (29) and move it to the end of the chromosome to give up the
access of it.