Вы находитесь на странице: 1из 11

Systems and Computers in Japan, Vol. 37, No.

7, 2006
Translated from Denshi Joho Tsushin Gakkai Ronbunshi, Vol. J85-D-I, No. 1, January 2002, pp. 21–30

A Genetic Algorithm for the Optimization of Cable Routing

Xuan Ma,1 Kazuhiro Iida,1 Mengchun Xie,2 Junji Nishino,1 Tomohiro Odaka,1 and Hisakazu Ogura1

1
Graduate School of Engineering, University of Fukui, Fukui City, 910-8507 Japan
2
Wakayama National College of Technology, Gobo City, 644-0023 Japan

SUMMARY 1. Introduction

In this paper the authors propose a structural method The problem of optimizing cable route selection is a
for a genetic algorithm (GA) for the optimization problem problem of optimizing overall the different number and
of cable routing in which cables have to be laid optimally. types of cable paths that can be created between various
When there are no limits on the layout routes, the optimal devices and equipment in a building. For one cable, this is
routes for the individual cables can be found using the equivalent to the problem of finding an optimal path be-
Dijkstra method in a finite discrete graph. However, when tween two nodes in a finite discrete graph, and solutions can
there are limits on the capacity in the layout routes, optimi- be found in algebraic time using the Dijkstra algorithm [1].
zation cannot be performed using the Dijkstra method. The However, in general there are limits to the amount of cable
proposed GA is a two-level hierarchical GA that uses that can be accommodated when laying lines (wiring tray).
chromosome coding involving two levels. The routes for The problem of selecting optimal wiring paths under such
each cable and the combinations of cable routes are defined constraints for several cables is NP-hard [2, 3], and solving
respectively, and layout route selection is optimized overall the problem of selecting optimal cable routes on a practical
through genetic operations involving these levels. Block scale is at present difficult. In this paper the authors propose
crossover and block mutation are used as genetic operations an effective method using a genetic algorithm (GA) [4] for
in the first level. In addition, a method to take advantage of solving the problem of optimizing cable routing.
lethal genes, which do not satisfy the constraints and are When there are no constraints on the accommoda-
generated in the second level is used. Computer simulations tions for the cable tray, the shortest path for each cable in
confirm that the two-level hierarchical GA with these ge- the problem of optimizing cable routing can be found using
netic operations functions effectively for the optimization the Dijkstra algorithm. When the constraints are present for
problem of cable routing. © 2006 Wiley Periodicals, Inc. the cable tray, routing is performed by eliminating any tray
Syst Comp Jpn, 37(7): 61–71, 2006; Published online in that does not satisfy the constraint conditions applying the
Wiley InterScience (www.interscience.wiley.com). DOI Dijkstra algorithm sequentially. Therefore, the superiority
10.1002/scj.10250 of the results ultimately obtained is highly dependent on the
sequence in which the cables are laid. In general, people
determine this sequence using experience and intuition, and
create a final proposed layout through repeated trial and
Key words: optimization of cable routing; path error. A GA can perform searches stochastically in parallel
search; combinatorial optimization; block crossover; hier- by representing candidate solutions to the problem as chro-
archical GA. mosomes [4], and so is efficient for finding an exact solu-

© 2006 Wiley Periodicals, Inc.


61
tion or an approximate solution for problems that require of lethal genes. In Section 4 the experimental results of
this form of trial and error. computer simulations using the proposed GA are given.
In power generation facilities and other massive in- Section 5 presents a discussion of the experiments.
stallations, several thousand to tens of thousands of power
cables and communications cables of various ratings are
used between different types of equipment and devices. The
trays used to lay out cable are planned during the design 2. The Cable Routing Optimization
phase for the building and then set up. The route for each Problem
cable is created using arrays of trays with cables laid out. If
a tray is considered an edge and the intersection (bifurca-
The cable routing optimization problem is as follows.
tion) of a tray and a device or piece of equipment is
There exists a network of cable trays that link equipment
considered a node, then the overall routes consisting of the
and facilities already set up. Cables are laid out in the tray
trays and the equipment or devices can be taken to be a
network with facilities and equipment as points of origin
discrete graph. This will be referred to as a tray graph below.
and destination. Each tray has allowed capacity for cables,
The cable routes can also be represented as arrays of nodes
and no more cables than that capacity can be laid in a tray.
in the tray graph. Each cable has a rated thickness based on
Given these constraint conditions, a route from the point of
its classification, and each tray has a layout capacity that
origin to the destination for each cable is selected, and a
matches the ratings. As a result, the cable routes have
routing plan that minimizes the total routing length for all
limitations. Depending on the class of the cable, it may be
cables is found. This represents the optimization problem.
placed in a tray of a particular type. The overall cost for the
Here, for the sake of simplicity, the types of cable and tray
cables used, that is, their overall length, can be minimized
ratings are unified, and the constraints are deemed to be
in this kind of cable routing based on the given tray setup
conditions related to the number of routes.
and capacity limitations.
A tray network is represented using a tray graph as
For an exact solution in the optimal cable routing
discrete graph, with the junctions (branching) and end
problem, the shortest route for each cable or the next shorter
points of the trays and the connecting points of equipment
route can be considered. In the two-level hierarchical GA
and devices as nodes, and the trays as edges. Figure 1 gives
proposed in this paper, a chromosome is represented using
a model of the tray graph used in the computer simulations
a two-level structure so as to integrate representations of all
in this research. In this figure, a circle represents a node,
route selections using each cable’s route and route combi-
and the number in the circle represents the number of the
nations. Then the GA is applied to each level. These GAs
node. The number on an edge represents the length of the
are referred to as the first-level GA and second-level GA.
tray. Here, for the sake of simplicity, the cost of a cable
In the first-level GA, several good routes (shortest or ap-
layout is deemed to be the length of the tray along the layout
proximately shortest routes) are found for each cable, and
route.
then in the second-level GA, the optimum combination of
the good routes for each cable is found. In the first level GA,
in which good routes for each cable are found, block
crossover and block mutation methods are used in order to
improve the efficiency of the route search. Next, the routes
are combined optimally using the second-level GA. Chro-
mosomes that do not satisfy the constraints for the tray are
referred to as lethal genes, and such lethal genes may be
formed in the second-level GA. Lethal genes are eliminated
from the next generation in general. However, some good
genes (building blocks) may exist in the lethal genes. The
authors also evaluated a method to take advantage of the
building blocks that these lethal genes have. In this paper
the effectiveness of the proposed GA is confirmed through
computer simulations.
In Section 2, the authors briefly explain the tray graph
use to address the optimizing problem of cable routing.
Section 3 explains the two-level hierarchical GA used for
the problem of optimizing cable router, genetic operations
proposed for the first time here, a method for block cross-
over and block mutation, and a method to take advantage Fig. 1. Example of the tray graph model.

62
The cable routing optimization problem in question operations for genes at each level are defined. The GA for
here can be readily formalized. Let G = (V, E) be a particular the cable route optimization problem using the authors’
undirected discrete graph. V represents a set of nodes, and method is referred to as a two-level hierarchical GA.
E represents a set of undirected edges. A positive cost is In the first-level GA, the route representation in the
assigned to each edge. The cost C is the mapping from E to representation of the chromosomes uses a list of the label
a positive real number such that C : E D ℜ+. n routes are (number) for the node in order. However, this list repre-
set up in G. The k-th route Tk is the route from the starting sentation requires genetic manipulation of the chromo-
node ak to the ending node bk. The route is represented in somes. The authors used a method for block crossover and
the order of the edges comprising the route (or in node block mutation proposed in the next section. In the second-
sequence). The proposed routes for all cables, that is, the level GA, lethal genes, the chromosomes that do not satisfy
set of n routes in G, is given by P = {(ak, bk, Tk)|k = 1, . . . , the constraints, are sometimes generated. The number of
n}. The cost C(Tk) for the route Tk is the sum of the costs lethal genes generated depends closely on the tray capacity
for the edges comprising the route: B. If the number of lethal genes generated is too high, then
(1) the efficiency of the GA drops, and, moreover, evolution
performance deteriorates. However, even if there are lethal
genes in a particular generation, the genes inherit genetic
Let the set S be the constraint conditions for G, the cable information (building blocks) for the traits that are superior
routing optimization problem becomes a matter of deter- over several generations. In order to make the most of such
mining P for F where the total cost of P is minimized under genetic information, the authors used lethal genes without
the conditions S: discarding them in their GA.
(2)
3.2. GA for selecting optimized cable routes

The constraints S are 3.2.1. Representing chromosomes


(3) The layout path P for all of the n cables is coded
genetically by dividing it into the following two hierarchi-
where the capacity for e with respect to the edge e ∈ E is
cal levels. There are normally several candidate routes for
Be, and the number of routes passing through the edge e in
route Tk of the k-th cable, from the starting node ak to the
the proposed routing P is b(e).
ending node bk. A codon pool Mk of route candidates is
Each cable is represented as a route between two
created by selecting M candidates for the k-th cable route
nodes, a starting node and an ending node, in the tray graph.
in P. The elements (candidate routes) in Mk represent the
A single route is represented as a series of edges or a series
route from ak to bk as a series of nodes. Each element in Mk
of nodes. Note that here, for the sake of simplicity, the
is given a route number of 1 ~ M. Q elements in Mk with
capacity of an edge is B for all trays.
good evaluation values are selected in sequence, and then
(4) comprise the candidate list Lk. Lk is a first-level chromo-
some. Next, the layout route P for all of the n cables is
3. Using the Genetic Algorithm for the represented issuing the chromosome g for the length n, thus
creating a second-level chromosome. The k-gene locus for
Cable Routing Optimization Problem
g represents the route Tk for the k-th cable. The value gk for
the k-gene locus of g represents the route number in the
3.1. Hierarchical GA for the cable routing candidate list Lk, and the gk route in Lk is Tk. The route Tk
optimization problem shall be referred to as the codon for the k-gene locus. Lk is
the k-th codon candidate list. Therefore, the chromosome g
The authors used a strategy consisting of two levels corresponding to the second-level chromosome, that is, the
for the cable routing optimization problem. Finding several layout path P with n routes, takes the numerical string of 1
optimal solutions (shortest or almost shortest routes) for ~ Q for n. A particular chromosome g represents a different
each cable involves the first-level GA, and then finding an layout route with n routes when the elements in the candi-
optimal combination of the routes for all the cables requires date list Lk are varied.
the second-level GA. A two-level coding method is used for In the authors’ method, the chromosome repre-
the coding of the chromosomes. The genes in the first level sentation for the problem of cable route optimization in-
represent the cable routes, and the genes in the second level volves two levels. Below, the first-level chromosome h
represent the cable combinations. In the GA here, genetic consisting of the set of the candidate route (candidate

63
codon) list Lk for the cables is referred to as the codon
chromosome.
For instance, Fig. 2 shows an example of a repre-
sentation of a chromosome where the number of cables is
n = 5 and the number of elements in Lk is Q = 5. This
example was created based on the graph model in Fig. 1.
In this representation, the gene locus for the upper
chromosomes corresponds to the cable number, and the
value for the gene locus represents the number of candidate
codons (routes) in the k-gene locus in the lower codon
chromosome. The length of the individual chromosomes is
five, the same as the number of cables. The value of the first
gene locus in the individual chromosome in the figure is Fig. 3. Example of a codon.
two, and as a result the first cable has the route 13-12-3-2-1,
which is represented by the second candidate codon. In the
same fashion, the fact that the value of the second gene
genetic operations in the list Lk of candidate codons in
locus is four means that it has the route 17-36-35-16-15-4-
first-level codon chromosomes. These genetic operations
3-2, which the second cable represents with the fourth
are applied to all of Lk(k = 1 ~ n).
codon. The candidate codon, which represents the route for
the cables in the codon chromosome, is represented using (1) Block crossover
a node label (number) series. For instance, Fig. 3 shows an
example of a candidate codon for a route with node 17 as The fitness value v(Ti) of the candidate codon Ti, i =
the starting point and node 2 as the ending point in the graph 1, . . . , M, in the candidate codon pool Mk corresponding to
in Fig. 1. Lk is determined as follows:
The candidate codon for the k-gene locus is evaluated
(5)
using the route length C(Tk) for the route Tk it represents.
The evaluation of the individual chromosome g is the F(P)
[Eq. (2)] for the sum of length of the route represented by Note, however, that CTav is an average value of the evalu-
each candidate codon. Unless otherwise stated, the candi- ation values for the candidate codons in Mk in that genera-
date codon representing the value of the k-gene locus in the tion. Mating, in which the parent pair used for crossover is
individual chromosome g and the corresponding route are selected from Mk, is performed using a roulette method
both represented with Tk. The evaluation value for the using the fitness value. Here, the mating method allows
chromosome is better when it is smaller. repeated selection. In this mating m1 parent pairs are se-
lected.
3.2.2. First-level genetic operations Block crossover is fundamentally a single-point
crossover. It is performed as follows. The explanation is
For the coding method described above, the authors given with reference to Fig. 4.
define the following block crossover and block mutation as

Fig. 2. Example of a representation of a chromosome. Fig. 4. Block crossover.

64
a. Block formation at random from the node series in the selected codon. Next,
a route linking the two selected nodes is created, and the
The parent pair for crossover selected from Mk for node series for it is designated as the mutation codon. The
mating is designated P1 and P2. First, a block length is node series between the two nodes in the original candidate
selected at random for the parent pair, and block 1 (block11 codon is replaced with the mutation codon.
and block21) are identified, as shown in the figure. Next,
when the node adjacent to the final node 15 in block 1 (3) Selection
(block11) in P1 is searched using the next node in block 1
The next generation Mk is created by adding Hk
(block21) in P2, 4 results. Consequently, from the found
obtained in (1) and (2) and selecting M individuals ran-
node position to the last node is deemed block 2 (block22)
domly from Mk. However, when more than one of the same
in P2. In the same fashion, when the node adjacent to the
candidate codon exists, only one is allowed. The method
final node 5 in block 1 (block21) in P2 is searched using the
for the alternation of generations follows the continuous
next node in block 1 (block11) in P1, 4 results. Conse-
generation model. Of the elements in Mk, Q are selected
quently, from the found node position to the final node is
from among those with better evaluation values, and the
deemed block 2 (block12) in P1. Crossover is then per-
candidate list Lk is created.
formed using the codon chromosomes obtained in block 2
as the parent pair. When block 2 is not obtained, a parent
3.2.3. Second-level genetic operations
pair is created using a different codon chromosome. When
nothing is found after searching the entire pool, a parent Next, genetic operations are performed on the indi-
pair is selected by altering the block length. Selection is vidual chromosomes representing the combinations of sec-
repeated using the codon chromosome pool until m1 parent ond-level chromosomes, that is, routes. Let G represent the
pairs are obtained. set of individual chromosomes consisting of N elements. G
is then the group (pool) of individual chromosomes.
b. Block crossover
(1) Crossover and mutation
The child codon S1 is created by connecting block 1
in P1 and block 2 in P2. In the same fashion, the child codon The fitness v(Pi) of the candidate chromosomes Pi,
S2 is created by connecting block 1 in P2 and block 2 in P1. where i = 1, . . . , N in G can be determined using the
following equation in the same form as Eq. (6) based on the
c. Eliminating duplicate gene loci evaluation value C(Pi) for Pi:
The existence of duplicate nodes in a codon after
(6)
crossover means that there is a closed loop in the route
represented by the codon. As a result, one of the duplicate
nodes and the node in between must be eliminated from the Note that CPav is the average value in the pool of the
codon. candidate chromosomes in G for that generation.
The child codon pool Hk is created using all of the For crossover in the individual chromosomes, first
child codons obtained from the above procedure as candi- the parent pairs for crossover are selected using the roulette
date codons in the next generation Mk. method taking advantage of the fitness of the individual
chromosomes defined in Eq. (6). In this mating, m2 parent
(2) Block mutations pairs are selected. A two-point crossover is performed on
the selected parent pairs. The individual chromosomes of
Next, block mutations will be explained while mak-
all of the resulting children comprise the pool G1. Next,
ing reference to Fig. 5.
point mutation is used with a mutation probability of Pm2
First, one codon is selected from the child codon pool
for the individual chromosomes in G1. In this point muta-
Hk using the mutation rate Pm1, and two nodes are selected
tion, the gene loci for the individual chromosomes are
chosen at random, and the value for the gene loci are
replaced by selecting an allele (candidate codon number 1
~ Q in Lk) at random. The individual chromosomes with
mutations in this fashion are added to G1.

(2) Lethal gene processing


Lethal genes are processed for rebirth in order to take
advantage of the individual chromosomes that do not satisfy
the constraint conditions, that is, the good portion of the
Fig. 5. Block mutation. lethal genes. After the gene pool G1 for the individual

65
chromosomes in the next generation of candidates, the (4) Hk is added to Mk and evaluated. Then, by random
object of selection in the second-level genetic operations, selection, the candidate codon pool Mk for the generation t
is created, the lethal gene pool is created by collecting only + 1 consisting of M elements is created. Next, the uppermost
the lethal genes included in G1. Crossover and mutation are Q elements are selected from Mk, and they comprise the
repeated in the lethal gene pool. As a result, individual candidate list Lk.
genes that are not lethal are returned to the pool G1 as The above processes in (1) through (4) are repeated
individual chromosomes in the next generation of candi- for k = 1 ~ n, and the next-generation codon chromosome
dates. In order to promote rebirth, the mutation probability set {Mk} is created.
Pmh is set high, and is set for multipoint mutation.
[Step 3] Evolution of the set of individual
(3) Selection chromosomes G

The resulting G1 is added to G, and N individual (1) m2 parent pairs are selected from G using the
chromosomes are selected using the rank selection method roulette selection method.
from among the individual chromosomes in the nonlethal (2) Two-point crossover is performed on each parent
genes in the pool G of individual chromosomes, and the pair, and the resulting child chromosomes are placed into
next generation G is created. When more than one of the the next-generation candidate pool G1.
same candidate codon exists, only one is allowed. The (3) Point mutation is performed on the individual
method for the alternation of generations follows the con- chromosomes selected from G1 at probability Pm2, and then
tinuous generation model. the resulting individual chromosomes are placed into G1.
(4) The individual chromosomes in G1 are evaluated.
The individual chromosomes that do not satisfy the con-
3.3. The hierarchical GA algorithm straints are designated as lethal genes.
(5) The lethal genes in G1 are transferred to the lethal
The hierarchical GA algorithm created in this paper gene pool, and then rebirth processing is performed on the
is summarized below. lethal genes. The reborn individual chromosomes are evalu-
[Step 1] Initialization ated and added to G1.
(6) G1 is added to G. Using the rank selection method,
(1) The final generation for evolution is set with the N individual chromosomes are selected from G, and the t +
candidate list number Q for the codons in the first-level GA, 1 generation of individual chromosome set G is created.
the group size M for the codons, the mating pair number Steps 2 and 3 are repeated until the final generation
m1, the mutation rate Pm1, the second-level GA group size is reached.
N, the mating pair number m2, the mutation rate Pm2, and The value of a gene locus in an individual chromo-
the number n of cables. some represents the route number for a cable. As a result,
(2) M candidate codons representing the routes for the same route number can represent a different route.
the k-th cable are created, and the initial codon pool Mk is Consequently, the individual chromosomes are evaluated as
created. The uppermost Q elements are selected from Mk pairs of codon chromosomes in a generation.
and designated Lk. This is repeated for all of the cables, and
the set {Mk} of codon chromosomes is created.
(3) N individual chromosomes consisting of length n 4. Simulations
and gene locus values of 1 ~ Q are created, and comprise
the initial set pool G.
In this section the authors describe the results of
(4) The number of generations is set so that t = 0.
performing experiments using computer simulations to
[Step 2] Evolution of the codon chromosome set evaluate the performance and effectiveness of the proposed
{Mk} GA method for the problem of cable route optimization.
First, the authors explain setting up the problem of
(1) m1 parent pairs are selected from Mk using the cable route optimization. The purpose of this experiment is
roulette selection method. to verify the effects in a small-scale problem in order to
(2) Block crossover is performed on each parent pair, evaluate the two-level hierarchical GA proposed for the
and the resulting child codons are put into the candidate problem of cable route optimization. As can be seen in Table
codon pool Hk for the next generation. 1, the authors created three types of graphs with shapes
(3) Block mutation is performed on the candidate similar to what is found in Fig. 1. |V| in Table 1 represents
codons selected from Hk at probability Pm1, and the result- the number of nodes, and |E| represents the number of
ing candidate codons are placed in Hk. edges. The length ei of an edge is generated at random

66
Table 1. Graph and data lower limit for the evaluation value in the problem of cable
route optimization. When there are constraints conditions,
the n cables are laid in sequence using the optimal routes
found using the Dijkstra method. At this point, the tray for
which the number of cables laid represents the upper limit
of the constraint condition are excluded, and the Dijkstra
method is applied. In this fashion, the total length of the
resulting layout routes depends on the sequence in which
within a numerical range given in Table 1. Three sets of data the cables are laid. The complete sequence for n cables is
are created for each tray graph. searched for while pruning using the branch and bound
Next, the method for generation of the initial set and (BB) method, and the resulting shortest total length repre-
the parameters for the GA are explained. In accordance with sents the optimal cable route. Here, this approach is called
the coding method described in Section 3.2.1, first the set the BB-Dijk method.
of n codon chromosomes {Mk|k = 1, 2, . . . , n} correspond- The search time in the BB-Dijk method rises dramati-
ing to the route candidate list Lk is generated at random in cally as n increases. To represent the problem of finding an
accordance with the tray graph. Next, the individual chro- optimal solution, in this experiment the number of cables
mosome set corresponding to the n layout routes P is was set to 10, and 3 as the constraint conditions were used
generated at random. When the restriction are not satisfied, for all the trays. For the selected 10 cables, the shortest
this is repeated until all chromosomes in the initial set routes were found using the Dijkstra method when there
satisfy them. were no constraint conditions with respect to the tray nodes
The mutation rate and the set size, both GA parame- in Fig. 1. The total length of the cables was 3292. With
ters, are usually set experimentally. In this research, the respect to the routes, Table 2 lists the trays which the various
authors determined them by performing preparatory experi- cables pass through given a capacity of at least three cables.
ments. In first-level genetic operations, trials were run In the BB-Dijk method, the total length was just 3500.
several times over a range of 0.01 ~ 0.1 in increments of The two-level GA was run through 10 separate trials
0.01 for the mutation probability. As a result, the authors on this problem. Figure 6 shows the generational changes
determined that 0.05 was a good average evaluation rate for for an average of 10 evaluation values for the elite individu-
the elite in each generation. Consequently, 0.05 was used. als in each generation. During the 10 trials, the best evalu-
In the second-level genetic operations, because the individ- ation value for the elite individuals was 3500, obtained
ual chromosomes are multivariate oppositional genes, a seven times. The worst evaluation value among the elite
slightly higher mutation rate was used. In the same fashion, individuals was 3527, and the average value for the 10 trials
trials were run over a range of 0.05 ~ 0.2, and as a result of was 3503.3.
a similar determination, 0.1 was chosen. The set size has a Computation time under the BB-Dijk method was
substantial influence on the results [5]. Preparatory experi- about 1 hour to search all of the cable layout sequences. In
ments using a set size of 30, 50, 100, and 200 yielded results the GA, the calculations for the 10 trials took under 4
similar to when the set size was above 50. As a result, in minutes (the computer used was a Pentium III 800 MHz
this experiment the set size M for the codon chromosomes with 512 MB of memory). However, because the computa-
and the set size N for the individual chromosomes were both tional burden rises by n! under the BB-Dijk method where
set to 50. The final generation was set to 200. The continu- n is the number of cables, the computational time increases
ous generation method was used for the alteration of gen- exponentially. In the GA, the trial time depends on the set
erations. size and the number of generations used, and is fundamen-
tally a polynomial time algorithm. Therefore, for a larger n,
the GA method may only produce a near-optimal solution,
4.1. Effects of the two-level GA based on but its computational time is vastly better. Next, the authors
preparatory experiments describe an experiment involving 60 cables. The problem
here involves a total cable length of 16,112 found using the
An optimal solution for the problem of cable route Dijkstra method when there are no constraint conditions.
optimization can be found using the following method.
When there are no constraints, the sum of the minimum
lengths for each cable represents an optimal solution, and
so this yields a lower limit for an exact solution. This lower Table 2. Number of cables through a tray
limit can be readily obtained when the minimum length for
each cable is found using the Dijkstra method. The sum of
the minimum lengths for all the cables at this point is the

67
found for the three tray graphs given in Table 1 using a GA
consisting of block crossover and block mutation. Table 4
lists the best, worst, and average values for the evaluations
obtained as a result of 10 trials for each tray. The best values
all match those found under the Dijikstra method. More-
over, several routes that were almost the shortest were found
using the same approach for the cables. The example given
in Table 5 represents the top five routes found for routes
between node 31 and node 43 with the edge length shown
in Fig. 1. These routes match the exact solutions.
The Dijkstra method is an algorithm for finding the
shortest route, but it cannot find the next shortest route or
Fig. 6. Comparison of solutions. other shorter routes as is. GA is better at finding near-opti-
mal solutions. By taking advantage of comparatively sim-
ple block crossover and block mutation, several solutions
that are almost the shortest can very likely be found.
The best numbers found using the Rand-Dijk method and In the experiments in the next sections, the authors
the Rand-BB-Dijk method are compared. The Rand-Dijk created three separate problems for the three tray graphs
method involves finding the cable routes using the Dijkstra given in Table 1. The experiments were run with 40, 60, or
method by selecting at random cable layout sequences for 90 cables. Below, the results of an experiment in a tray
all of the cables. In this experiment, within the time of 10 graph with 49 nodes and 84 edges, and involving 60 cables,
trials of the GA, the 10,000 trials of Rand-Dijk are per- are given. Similar results were found for the other cases.
formed to find the best value. The Rand-BB-Dijk method
involves finding solutions for some cables using the Rand-
4.3. Effects of the number of list candidates on
Dijk method, and then finding solutions for the remaining
GA performance
cables using the BB-Dijk method. In the experiment the
best values were found by running 20 trials in which the The number Q of list candidates in the proposed GA
Rand-Dijk method was used for 50 cables, and the BB-Dijk significantly impacts the GA search performance. The
method was used for the remaining 10 cables. The BB-Dijk authors analyzed the effects of Q where the constraints
method requires a considerable amount of time for trials, included an allowed number of cables in each tray of B =
and when the number of cables for the BB-Dijk method is 18. The two-level GA was run through 10 trials for a cable
reduced, the best solution is often a local solution. Note that number of 10, 20, and 30. Lethal genes were not used in
in the 20 trials, more than 10 hours was needed. Table 3
summarizes the results of the experiment. The time for
10,000 trials under the Rand-Dijk method was roughly the
same, 4 minutes, as the time for 10 trials in the two-level
GA method. It is clear that the two-level GA method yields Table 4. Resolution of best, worst, and average
better layout routes in 10 trials up to 200 generations. The
execution time is also superior under the two-level hierar-
chical GA.

4.2. Effects of block genetic operations

The authors evaluated the effects of block crossover


and block mutation. Several routes between nodes were

Table 3. Resolution of best, worst, and average of the


elite

68
Table 5. Example of the shortest routes Table 6. Comparison of using lethal genes

advantage of lethal genes increases the total processing time


by roughly 3% (8 seconds) when lethal genes are used
under experimental conditions that are otherwise the same,
this experiment. Figure 7 shows the generational changes
but the solutions are far better. This means that the lethal
in the average evaluation number for the elites in each
genes include superior qualities as building blocks, and by
generation. As can be seen in Fig. 7, it is clear that as the
bringing them back the building blocks can contribute to
value of Q increases, the convergence of the evaluation
the evolutionary process.
value for the elites is delayed. When the value of Q rises,
the search time for a solution increases, and when the value
of Q is too low, the number of route combinations is too
small, thus increasing the possibility of a local solution
5. Conclusion
emerging. Figure 7 makes clear that the possible selections
when Q = 5 are too limited. In the experimental below Q = In this paper the authors proposed the creation of a
10 was used. hierarchical GA for the problem of cable route optimiza-
tion. The problem of cable route optimization is a problem
4.4. Effects of the method using lethal genes consisting of two levels: searching for routes and combin-
ing routes. Because of constraint conditions in which there
The authors evaluated the active use of lethal genes is a limit to the cable capacity of a tray, the problem of cable
generated by genetic operations when the constraint condi- route optimization is different from a simple problem of
tions were severe. If the superior building blocks borne by searching for optimum routes, and instead becomes one of
the lethal genes generated can be used effectively, then selecting optimum combinations from several candidate
improvements in the efficiency of the GA can be gained. routes within the constraint conditions. In other words, this
The evaluation experiment was performed for a problem is a problem that is similar to the multiple-choice knapsack
with a cable number of n = 60 and constraint conditions of problem [5–7]. As a result, the authors proposed a hierar-
B = 18, and with a candidate codon list in which Q = 10. chical GA which focuses on chromosome coding with a
Ten trials were run with a GA that makes active use of lethal hierarchical structure for the problem of cable route opti-
genes and one that does not, and the results were compared. mization.
Table 6 lists the best and worst evaluation values for the In this paper the authors used two levels of chromo-
elites obtained in the 10 trails, the average evaluation value some coding in combination with a hierarchical structure
for the 10 elites, and the time required for the 10 trials. As for the problem of cable route optimization. In the GA, the
can be clearly seen in Table 6, the method that takes key point is having the GA operate efficiently: how to code
the problem in question as chromosomes and how to define
the corresponding genetic operations. In the past, genetic
operations corresponding to various coding methods [8, 9]
have been proposed. However, when the problem is com-
plex, it is often difficult to represent using only a single-
level (one-dimensional series) chromosome. The
chromosomes with a compound structure proposed in this
paper are therefore necessary. Depending on the format of
the problem to be solved, the problem’s structural space and
the GA’s search space may have different structures. As a
result, solutions that cannot be implemented may appear or
lethal genes may result, depending on the coding and the
crossover and mutation methods used. Because the route
representation and gene formats are different in conven-
Fig. 7. Generational changes in the average evaluation tional route searching GAs, caution is required with respect
value for the elite with respect to different candidate list to the relationship between the problem’s structural space
numbers. and the GA’s search space. In the proposed methods for

69
block crossover and block mutation, the strings repre- 2. Sedgwick R (editor), Noshita K, Hoshi M, Sato T,
senting the routes are used as is as candidate codons. Taguchi A (translator). Algorithms, Vol. 3. Kindai
Consequently, the search space for the first-level GA is the Kagakusha Publishing; 1992.
same as the solution space for the problem of finding routes 3. Iwasaki T, Ota K, Nakagawa Y, Miyashita F, Naruhisa
(candidate codons) for each cable. Thus, the appearance of Y. Approximation method for nonlinear integer pro-
candidate codons that do not represent continuous routes is gramming problems. Trans IEICE 1997;J80-
avoided. A:1439–1443.
Several GAs have been proposed [8, 9, 12, 13] for 4. Holland JH. Adaptation in natural and artificial sys-
route searches without constraints, and the GA proposed in tems. University of Michigan Press; 1975.
this paper can be used for such networks. In this paper, for 5. Ma X, Xie M, Nishino J, Odaka T, Ogura H. Appli-
the sake of simplicity the authors assumed that the con- cation of generation schemata to multiple value cod-
straint defining the capacity of a tray is the same for all ing system in MCKP. Trans IEICE 2000;J83-D-I:
trays. However, no problem results if the constraints vary 1214–1218.
with each tray. When the constraints are severe, lethal genes 6. Sinha P, Zoltners AA. The multiple-choice knapsack
that do not satisfy the constraints are readily produced. In problem. Oper Res 1979;27:503–614.
some cases, an exact solution (a solution that satisfies the 7. Armatrong RD, Kung DS, Sinha P, Zoltners AA. A
constraints) may not exist. In such cases as well, the two- computational study of a multiple-choice knapsack
level GA in the authors’ method can be used under the same algorithm. ACM Trans Math Software 1983;9:184–
framework by extending their method through fuzzy logic 198.
for the constraints and reflecting in the evaluation values 8. Leung Y, Li G, Xu ZB. A genetic algorithm for the
the extent to which the constraints are satisfied, and a multiple destination routing problems. IEEE Trans
near-optimal solution that respects the constraints to some Evol Comput 1998;2:150–161.
extent can be found. 9. Esbensen H. Computing near-optimal solutions to
In this paper the authors proposed a two-level GA the Steiner problem in a graph using a genetic algo-
with two-level chromosome coding for the problem of rithm. Networks 1995;26:173–185.
cable route optimization. This is a GA that hierarchically 10. Schultz AC. Adapting the evaluation space to im-
combines a route search GA and a route combination GA. prove global learning. Proc 4th International Confer-
Experiments involving computer simulation confirmed that ence on Genetic Algorithms, p 158–164, 1991.
the method of block crossover and block mutation is effec- 11. Kitano H. Genetic algorithms. Sangyo Tosho Pub-
tive for route searches, and that the proposed GA functions lishing; 1993.
effectively for the problem of cable route optimization. 12. Yamamura Y, Ono T, Kobayashi S. Character-pre-
serving genetic algorithms for traveling salesman
REFERENCES problem. J Japan Soc Artif Intell 1992;7:117–127.
13. Inagaki J, Haseyama M, Kitajima H. A method of
1. Eio AV, Hopcroft JE (editors), Ono Y (translator). determining various solutions for routing application
Data structure and algorithms. Baifukan Publishing; with a genetic algorithm. Trans IEICE 1999;J82-D-
1996. I:1102–1111.

70
AUTHORS (from left to right)

Xuan Ma graduated from the Department of Precision Instrument at Xi’an University of Technology (China) in 1983.
He enrolled in the Graduate School in 1989, completed his master’s course in 1992, and became a lecturer there. He went to
Japan in 1997 and completed his doctoral course at the University of Fukui in 2002. He returned to China and is now an associate
professor in the Faculty of Automation and Information Engineering, Xi’an University of Technology. He holds a D.Eng. degree,
and is pursuing research related to information processing.

Kazuhiro Iida graduated from the Department of Informatics at Fukui University in 1999 and completed the master’s
and doctoral courses at the University of Fukui in 2001 and 2005. He is now an instructor at the IB Techno College and teaches
computer science and practical information techniques.

Mengchun Xie received her M.S. degree in mechanical engineering from Xi’an University of Technology (China) in
1988 and D.Eng. degree from Fukui University in 1997. She is now an associate professor in the Department of Electrical and
Computer Engineering, Wakayama National College of Technology. Her research focuses on knowledge representation,
knowledge processing and modeling in the fields of artificial intelligence, computer simulation, and multiagent systems. She
is a member of the Information Processing Society of Japan and the China Mechanical Design Society.

Junji Nishino received his B.S. degree in control engineering and M.S. degree in systems science from Tokyo Institute
of Technology in 1990 and 1992. He has been an assistant professor in the Department of Systems Engineering, University of
Electro-Communications, since 2001. He received a research encouragement award from the Japanese Society for Artificial
Intelligence in 2001. His research interests include intelligent/information systems, linguistic fuzzy system theory, and fuzzy
intelligent controls. He is a member of the Japan Society for Fuzzy Theory and Intelligent Informatics, the Information
Processing Society of Japan, the Japanese Academy of Facial Studies, and the Japan Society of Kansei Engineering.

Tomohiro Odaka (member) received his D.Eng. degree from Waseda University in 1990. He is now a professor in the
Graduate Program of Nuclear Power and Energy Safety Engineering, Graduate School of Engineering, Fukui University. He is
a member of IEEE CS, ACM, IPSJ, IEICE, and JSMBE. His research activities focus on user modeling, network security,
biomedical signal processing, intelligent human interface, and computer-aided instruction.

Hisakazu Ogura (member) completed his doctoral course at Kyoto University in 1977. After serving as an assistant
professor and associate professor at Kochi Medical School, he moved to the Department of Information Science at Fukui
University, and is now a professor in the Department of Human and Artificial Intelligent Systems. He is engaged in research in
fields such as GA, NNW, fuzzy language, language representation, medical image analysis, and medical information processing,
with a focus on the representation, acquisition, and processing of knowledge. He holds a D.Sc. degree, and is a member of
IEICE, IPSJ, JSAI, the Japan SOFT, IEEE, and ACM.

71

Вам также может понравиться