JPH07230443A - Method and device for assigning chromosome - Google Patents

Abstract

PURPOSE:To provide the chromosome assigning method and device which sufficiently speeds up convergence on an optimum solution when a combinational problem is solved by genetic algorithm. CONSTITUTION:When the correspondence relation between request elements (distribution destination) having request quantities respectively and process elements (truck) whose processable quantities (maximum loading capacity) are defined is represented in a chromosome 30, the operation numbers of the tracks are allocated as values of respective genetic seats 31 of the chromosome 30 without overlapping. Then the tracks are assigned from the 1st distribution destination in order until the request quantity of the distribution destination is satisfied. In this case, the tracks are assigned by the distribution destinations by searching for genetic seats 31 in order from one end side of the chromosome 30 until the sum of the maximum loading quantities exceeds the request quantity of the distribution destination and reading the maximum loading capacity of the track corresponding to the value of each genetic seat 31 out of a maximum loading capacity table 11.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for allocating chromosomes when solving a combination problem using a genetic algorithm.

[0002]

2. Description of the Related Art A combination problem is a kind of optimization problem, and when an objective function (evaluation function) whose value is determined by a combination of variables and elements is given, the objective function of the objective function is given under a predetermined constraint condition. It tries to find a combination with the maximum or minimum value. A typical example is the case of a traveling salesman problem that seeks the order of visits that results in the shortest travel distance when visiting multiple cities once each, or when there are multiple items with different weights and values. There is a knapsack problem in which an item is selected so that the sum of values is maximized within a predetermined total weight range.
Problems such as communication network design, job scheduling, and graphic placement often result in solving this combination problem. In many cases, combinatorial problems cannot always be solved in a short time and with a small amount of calculation, for example, when trying to solve mathematically, it is necessary to check the case by brute force.

In recent years, a solution search method called a genetic algorithm has been developed. The genetic algorithm is one of the probabilistic search methods, and is modeled on the evolution process (selection and mutation) of living things. FIG. 1 (a) is a diagram for explaining the outline of the genetic algorithm, and FIG. 1 (b) is a diagram showing an example of conformation on a chromosome. The outline of the genetic algorithm will be described below.

First, consider a "chromosome (gene) 200" having a plurality of loci and each variable or each element assigned to each locus, and generating a plurality of such chromosomes (step 201). , First generation population 2
02. Each chromosome 200 corresponds to a particular combination determined by the variables or elements contained therein. The chromosomes belonging to the first generation are generally generated using random numbers, but a predetermined initial value may be used.

Next, regarding the first generation population 202,
The fitness according to the value of the objective function is calculated for each chromosome 200, a chromosome having a high fitness is selected, and further, the selected chromosome is crossed (corresponding to crossing between biologous homologous chromosomes) or mutation ( An operation such as changing a variable or an element randomly mapped to a locus with a certain probability) is added (step 203), and the second generation population 2 is added.
04 is generated. In this process, chromosomes (individuals) with low fitness are eliminated. The number of new chromosomes generated by crossing or mutation is generally the same as the number of selected chromosomes, so that the total number of individuals is generally kept constant. The same evolution process including selection, crossover (mating), mutation, and selection is applied to the second-generation population 204 to generate a third-generation population, and the evolution process is repeated. By sequentially generating the next-generation population, the population 205 of the nth generation is generated.

Since individuals having a low fitness are eliminated in the evolution process, the population 205 of the nth generation is composed of chromosomes having a high fitness as compared with the population 202 of the first generation. In this way, a higher fitness chromosome will be obtained for each generation,
As described above, since each chromosome corresponds to each combination, a more optimal combination can be found. Also, in this genetic algorithm, since the mutation is added during the evolution process, the solution space is searched throughout, so that the global optimum solution can be accurately reached. In many real-world problems, it is known that the optimal solution can be reached in a much shorter time by using the genetic algorithm than in the case of performing a brute force search (as an example, “genetic algorithm”, Hiroaki Kitano, Industrial Books, 1993 (I
SBN4-7828-5136-7)).

Here, the constitution of the chromosome will be described.
FIG. 1B shows an example of the configuration of the chromosome 200 when the genetic algorithm is applied to create a truck allocation plan. Here, using 25 trucks, 22
Consider a case where a package is delivered to a delivery destination (company A to company V). It is assumed that the specified load capacity is set for each truck, the amount of luggage is set for each delivery destination, and each truck can make three round trips per day. It should be noted that the amount of luggage to a certain delivery destination is not always the amount that can be carried in one truck, and in such a case, a number of trucks (the same truck may make multiple round trips) may be used. Shall be transported. By the way, when the number of combinations is calculated in consideration of waiting, the total is 75 (= 25 × 3) round trips, and each round trip has one of 22 delivery destinations or a total of 2 waiting spots.
There are three cases, so if you simply calculate,
^{There are 75} (≈ 1.35 × 10 ¹⁰² ) ways.

The chromosome 200 indicates the destination of each truck in each round trip.
75 loci 210 are provided on the top. Further, the number assigned to each locus 210 represents the number of the delivery destination, and the number 0 represents the standby. That is, the m-th (1 ≦ m ≦ 25) th track of the n-th track (1 ≦ n
The round trip destination of ≤3) is represented by the 3m + n-3rd locus. Chromosome 20 illustrated in FIG. 1 (b)
With 0, the 1st truck will go to the 8th delivery destination (Company H) at the first time, the second will wait and the 22nd delivery destination (Company V) will go to the 3rd time, and the 25th truck Is
First and second time, go to No. 10 delivery destination (company J) and
The second time is waiting.

In order to obtain the optimum combination (= optimum vehicle allocation plan) by the genetic algorithm using the chromosomes designed in this way, a delivery destination (including waiting) is randomly set at each locus 210. Form a first-generation population, and, for example, an evaluation function f represented by the following equation
Selection may be performed using. In this evaluation function f, the smaller the value, the better, that is, the higher the fitness.

[0010]

[Equation 1] Of course, it is a prerequisite (restriction condition) that all the packages will be delivered to each delivery destination, so the amount of packages that can be actually delivered is calculated for each delivery destination in consideration of the specified loading capacity of each truck. It is checked whether the amount of the package is larger than the amount of package set for each delivery destination, and for the chromosomes that do not satisfy this condition, an extremely large value is given to the evaluation function f as something to be removed. It should be given an extremely bad fitness.

[0011]

When trying to solve a combinatorial problem using a genetic algorithm, the quality of chromosome design is an important factor that determines the speed of convergence to an optimal solution and the amount of calculation. There is. Even when solving the same problem, the convergence speed to the optimal solution may differ by several orders of magnitude depending on the chromosome design. In an extreme case, if the chromosome design is poor, the convergence speed may be similar to that of the random search.

To solve the above-mentioned truck allocation problem, a plurality of trucks include two or more trucks of the same type (all trucks having the same vehicle type, maximum load capacity, transportation cost, etc.). There are cases. When multiple trucks of the same type are included, the value of the evaluation function is exactly the same even if the variables are interchanged among the loci corresponding to the trucks of the same car type in the method of designing the chromosome shown in Fig. 1 (b). Therefore, there is a problem that the convergence to the optimum solution becomes slower. Similarly, assuming that the value of the evaluation function is not affected by the delivery time zone such as morning or afternoon, the value of the evaluation function is the same even if the first destination and the second destination are exchanged for each truck. Therefore, the convergence to the optimal solution becomes slow. Even if the delivery time zone has an effect on the evaluation function, the effect is considered to be slight. Therefore, even if the first and second destinations are exchanged, the difference between the values of the evaluation function is small. Yes, the value of the evaluation function does not vary,
Convergence will be slow. The reason for the slower convergence is that when the chromosomes in a population are similar to each other, it tends to fall into a local optimal solution, and the global optimal solution can be reached only by mutation.

It is an object of the present invention to provide a method and apparatus for allocating chromosomes that can converge to an optimal solution sufficiently quickly when a combinational problem is solved by a genetic algorithm.

[0014]

According to the method of allocating a chromosome of the present invention, a plurality of request elements each having a required quantity defined, an order set, and a plurality of processable quantities defined are prepared. A method for allocating chromosomes used to solve a combinational problem that defines a correspondence relationship between and by a genetic algorithm, wherein each of the processing elements is allocated as a value of each locus of the chromosome without allowing duplication. , The processable amount or the sum of the processable amounts becomes equal to or larger than the required amount of the required element in the first rank of the order, the processing elements are sequentially read from the loci on one end side of the chromosome, and the required element is read. And a corresponding relationship between the extracted processing element and
The processing of setting the correspondence relationship is repeated for the locus from which the processing element has not been read for the request element of the next order.

The chromosome assigning apparatus of the present invention is a combination problem that defines a correspondence relationship between a plurality of request elements each having a required amount defined and a plurality of processing elements each having a processable amount defined. A method for allocating a chromosome used for solving the above problem by a genetic algorithm, wherein the combination problem is such that a processable amount that is not less than a required amount of the required element is assigned to each of the required elements. A first storage for searching for a combination that associates a plurality of the processing elements with the request element and minimizes or maximizes a predetermined evaluation function, and stores the requested amount of the request element for each of the request elements. Means, second storage means for storing the processable amount of the processing element for each of the processing elements, and a number corresponding to the number of the processing elements. A chromosomal storage means consisting of elements, wherein each storage element corresponds to each locus of the chromosome, and the processing element is stored as a value of the storage element for each storage element by assigning it without duplication. Until the processable amount or the sum of the processable amounts becomes equal to or more than the required amount of the request element of the first rank in the first storage unit, the chromosome storage unit of the chromosome storage unit is referred to while referring to the second storage unit. The storage elements are sequentially searched from one end side, the processing elements are read, a correspondence relationship is set between the request element and the retrieved processing element, and the processing elements are sequentially processed with respect to the request elements in the next order. Assigning means for repeating the process of setting the correspondence relationship with respect to the storage elements that have not been read.

[0016]

[Function] As a guideline for designing chromosomes in a genetic algorithm, completeness (all solution candidates in the solution space can be represented by chromosomes) and soundness (all chromosomes in the chromosome expression space correspond to solution candidates in the solution space) In addition to that, non-redundancy (corresponding one-to-one correspondence between a chromosome and a solution candidate) is known. It is intended to raise

In the present invention, various contents can be considered as the contents of the combination problem, but as a typical example,
One example is the preparation of truck allocation plans. Hereinafter, the operation of the present invention will be described by taking the truck allocation problem as an example, but the application range of the present invention is not limited to this.

When the truck allocation plan is considered as a combination problem, the delivery destination that has requested delivery of a predetermined amount of luggage is the request element, and the truck performing the delivery is the supply element. The required amount is the amount of luggage to be delivered to the delivery destination, and the processable amount is the maximum load amount of each truck. In the chromosome described in FIG. 1 (b) in the “Conventional Technology” section, each locus itself corresponds to a track, and the locus value indicates the delivery destination. On the other hand, in the chromosome allocation method and device of the present invention, the value of each locus corresponds to the track, and the association between the chromosome and the delivery destination is the required amount (the amount of parcel to be delivered).
And indirectly through the processable amount (maximum load amount). By allocating chromosomes in this way,
In the case where the same type of tracks exist, the conventional allocation method causes the degeneracy of the chromosomes and it takes time to search the solution by the genetic algorithm. However, the allocation of the chromosomes was performed based on the present invention. In this case, since the degeneration of the chromosome is extremely unlikely, the optimal solution can be found in a short time. Simply put, traditional allocation methods
The elements (combinations) in the solution space and the elements (chromosomes) in the chromosome expression space do not have a one-to-one correspondence, but according to the present invention, the one-to-one correspondence is achieved and the efficiency is improved. You will be able to search.

The combination problem targeted by the method and apparatus for allocating chromosomes of the present invention is not only the scheduling problem in the transportation industry typified by the above-described truck allocation plan but also optimum conditions in various scheduling and production processes. There are combinatorial problems in the fields of computing, real-time control, and adaptive control of robots.

[0020]

Embodiments of the present invention will now be described with reference to the drawings. FIG. 2 is a block diagram showing the configuration of a chromosome assignment device according to an embodiment of the present invention. This allocation device
According to the method of the present invention, the chromosome is assigned and the evaluation value of each chromosome (objective function, value of the evaluation function) is calculated, and selection, mating and selection by genetic algorithm are performed according to the evaluation value. ing. Here, a case of creating a truck allocation plan will be described as an example.

As the problem setting, similarly to the case described in "Prior Art", a dispatch plan is created when 25 trucks are used to deliver parcels to 22 delivery destinations. Is to make three round trips to the same delivery destination and different delivery destinations, and in each round trip, deliveries the package to only one delivery destination or waits. In this embodiment, the number of round trips of which truck is identified by the operation numbers from 1 to 75. The n-th round trip (n = 1 to 3) of the m-th truck is operation number X = 3.
It is represented by m + n-2. For example, 2 of the 15th track
The second round trip is represented by the operation number 44. The maximum load amount corresponding to the processable amount is set for each truck, and the amount of parcels to be sent to the delivery destination, that is, the required amount is set for each delivery destination. Some of the requested amounts exceed the maximum value of the maximum load amount, and it is necessary to deliver to such a delivery destination a plurality of round trips (the same truck or different trucks may be used). Further, an optimum load capacity is set for each truck within the range of the maximum load capacity, and when carrying a load of this optimum load capacity, the evaluation value in the genetic algorithm is set to be the best.

Next, the assigning device of this embodiment will be described. This allocation device includes a maximum load capacity table 11 that stores the maximum load capacity of each truck, a delivery quantity table 12 that stores the required quantity of each delivery destination, a chromosome storage unit 13 that stores chromosomes, and a maximum load capacity. Based on the calculated evaluation value, an evaluation value calculation unit 14 that calculates the evaluation value of each chromosome by allocating the chromosome based on the method of the present invention with reference to the table 11, the delivery amount table 12, and the chromosome storage unit 13. A GA processing unit 15 that calculates the fitness of each chromosome and performs operations such as selecting, intersecting, and mutating the chromosomes by a genetic algorithm to replace the chromosomes in the chromosome storage unit 13 with the next-generation chromosomes, and the initial value chromosomes ( An initial value generation unit 16 that generates a chromosome of the first generation population) by random numbers and stores it in the chromosome storage unit 13. In a genetic algorithm, populations of each generation are generally formed by 100 or 200 chromosomes, and correspondingly, the chromosome storage unit 13 has a capacity enough to store the entire chromosomes of the population of one generation. have. Further, each chromosome is composed of a plurality of (75 in this case) loci, and a storage element of the chromosome storage unit 13 is assigned to each locus. Here, the storage element stores one unit of storage, and typically corresponds to each storage address in the chromosome storage unit 13.

Next, the assignment of chromosomes in this embodiment will be described with reference to FIG.

As shown in FIG. 3A, 75 loci 31 are provided on the chromosome 30, and each locus 31 holds the operation number of the truck as a value. The 75 loci 31 correspond to the service numbers 1 to 75. In this case, the same operation number does not appear in different loci 31. That is, each locus stores the operation number as a value without allowing duplication. Since the value of the leftmost locus (first locus) in the figure is 43, this locus represents the first round trip of the fifteenth track, and the second locus from the left ( The value 59) represents the second round trip of the 20th track. Hereinafter, the mth truck (the truck whose truck number is m) is referred to as truck #m, and the nth delivery destination (the delivery destination with delivery destination number n) is referred to as the delivery destination #n. I will decide.

FIG. 3B shows an example of the specific contents of the maximum load capacity table 11, in which the maximum load capacity is shown for each truck. In addition, in the maximum load capacity table 11,
An operation number column is provided to indicate which operation number of which round trip of which truck corresponds. Also,
FIG. 3C shows an example of specific contents of the delivery amount table 12. The amount of parcels to be delivered to the delivery destination is stored as a requested amount for each delivery destination, and further, up to the delivery destination. The time required for the round trip is stored. The time required for the round trip is used when calculating the evaluation value of the chromosome.

In the present embodiment, the assignment of chromosomes is performed when the evaluation value for each chromosome is calculated.
In the process of gene manipulation such as selection, mating, selection, and mutation, it is sufficient to manipulate only the value of each locus 31 on the basis of the fitness corresponding to the already evaluated value. It is not necessary that the relationship between the truck operation number and the delivery destination is associated, but it is premised that this association is fixed when the evaluation value is calculated.

On the chromosome 30 shown in FIG. 3A, the association between the truck and the delivery destination is started from the locus on the left end side in the figure. Referring to the delivery amount table 12, since the requested amount of the delivery destination # 1 is 0, the vehicle allocation to the delivery destination # 1 is not performed, and subsequently, the requested amount of the delivery destination # 2 is read. The required amount of the delivery destination # 2 is 3 tons, and the locus 31 of the chromosome 30 remains until the sum of the maximum loading amounts becomes 3 tons or more.
Is searched from the left end side in the figure, and a track is assigned. The locus at the left end of the figure is 43, which corresponds to the second round trip of truck # 15, which has a maximum loading capacity of 3 tons. Therefore, the second round trip of truck # 15 to destination # 2. Shall be delivered.

Next, looking at the requested amount of the delivery destination # 3, 2.9
Since it is ton, the locus 3 and the second locus from the left end of the chromosome 30 until the sum of the maximum load is 2.9 tons or more.
1 is sequentially searched and a track is assigned. The second locus (value 59) is truck # 20 (max load 3 tons)
This is the second round trip, so this round trip will go to delivery destination # 3.

Although the requested amount of the delivery destination # 4 is 1.9 tons, the third locus (value 73) is the first round trip of the truck # 25 having the maximum loading capacity of 1 ton, and thus the request is sufficient. We cannot satisfy the quantity of 1.9 tons. Therefore, referring to the fourth locus (value 70), this is the first round trip of truck # 24 having a maximum loading capacity of 1 ton, and the first round trip of truck # 25 and the first round trip of truck # 24. The maximum load capacity is 2 tons. Therefore, it is assumed that the package is delivered to the delivery destination # 4 by the first round trip of the truck # 25 and the first round trip of the truck # 24.

The delivery destinations # 5 and # 6 are shown below as illustrated.
The trucks are allocated to the trucks, and the next delivery destination is sequentially determined by the number of round trips of which truck to deliver, and the correspondence between the trucks and the delivery destinations is completed. The general procedure of this operation is shown as a flowchart in FIG. 5 described later. In the example shown in FIG. 3 (a), when the fifth locus from the right end of chromosome 30 (the 71st locus counting from the left end) is reached, all delivery destinations (delivery destinations # 1 to # 22) are delivered. Track allocation is complete. Therefore, the four loci from the right end of the chromosome 30 represent unused tracks, that is, waiting tracks. For example, the value of the fourth locus from the right end is 1, so track # 1 is 1
You will not make the second round trip.

In this way, the evaluation value calculation unit 14 associates the delivery destination with the truck for each chromosome,
An evaluation value is calculated for each chromosome.

Next, a procedure for solving a combination problem by a genetic algorithm using this chromosome assigning device, that is, a procedure for obtaining an optimum vehicle allocation plan will be described.
This will be described with reference to the flowchart of FIG.

First, the initial value generator 16 generates n chromosomes 30 by random numbers to generate a first generation population (step 111). Each chromosome 30 is shown in FIG.
It has the structure shown in (a). Then, the evaluation value calculation unit 14 calculates the evaluation value for each of the chromosomes 30 (step 112). There are various possible evaluation values f, for example,

[0034]

[Equation 2] The one represented by can be used. Here, the smaller the evaluation value, the better the chromosome, that is, the chromosome having the higher fitness. Furthermore, the evaluation value may be set in consideration of fuel consumption, labor costs, and the like. In addition, regarding chromosomes that are supposed to be shipped by ordinary (not refrigerated) trucks for packages that require low temperature transportation, and other chromosomes that violate the constraints that are strongly required to comply, Alternatively, an extremely bad evaluation value may be given instead of the evaluation value according to Expression (2), or it may be replaced with another chromosome generated by a random number.

Then, the control is transferred to the GA processing section 15, the fitness is calculated from the evaluation value of each chromosome, and the selection roulette is prepared (step 113). The selection roulette is for selecting each chromosome with a probability according to its fitness. Then, using the roulette wheel for selection, one parent chromosome is selected for each, and n / 2 sets of parents are selected in total (step 114), and two children chromosomes are generated from each parent chromosome by crossing each other. (Step 115). Then, the offspring chromosome is mutated with an appropriate probability (step 116) to generate a second generation population, which is stored in the chromosome storage unit 15. At this time, the second
In each chromosome of the generational population, crossover and mutation are performed so that the values of the loci within the chromosome do not overlap. Then, an evaluation value is calculated for each of the second generation chromosomes in the same manner as described above (step 11).
7). Next, judge whether the end condition is satisfied,
That is, it is determined whether the vehicle allocation is completed (step 122). As the termination condition here, for example, a condition that the evaluation value of the best chromosome hardly changes over several generations, or the evaluation value reaches a predetermined value can be adopted. If the end condition is satisfied, the process by the genetic algorithm is ended, and if not satisfied, the process returns to step 113 and the same process is repeated in the next generation.

Next, the evaluation value is calculated (steps 112, 1
Details of 17) will be described with reference to FIG. In the process of calculating this evaluation value, as described above, the chromosome allocation by the method of the present invention is executed. Here, the required amount of the i-th delivery destination (delivery destination #i) is represented by D (i), and the maximum load of the truck corresponding to the value of the j-th locus is represented by A (j). As a general rule, the chromosome allocation process will be described. Here, if the above-mentioned operation numbers are different, it is assumed that the truck is a different truck, and the discussion will proceed.

First, variables i and j are initialized to 1 (step 120). It is checked whether the requested amount of the delivery destination #i is 0 (step 121). If it is 0, the process proceeds to step 131 described later to perform the process for the next delivery destination, and 1 is set to i. If it is not 0, j is substituted for the variable MIN (step 122), and the variable T representing the sum of the maximum loading amounts is initialized to 0 (step 123). Then, the maximum load capacity A (j) of the truck corresponding to the value of the j-th locus is added to the variable T (step 124). Here, the value of j is checked (step 125), and when j is equal to the total number of loci, that is, when the end of the locus is reached, step 1 described later is performed.
33, and if the end has not been reached, the process proceeds to the next step 126. When passing through the step 125 for the first time, generally, since the end of the gene locus has not been reached, the step goes to the step 126.

At step 126, 1 is added to j.
Then, it is checked whether the variable T is equal to or larger than the requested amount D (i) of the delivery destination #i (step 127), and D (i) ≦ T.
In this case, since the truck is assigned with satisfying the request amount of the delivery destination #i, the process proceeds to step 128. D (i)>
In the case of T, since the requested amount of the delivery destination #i is not yet satisfied, the process returns to step 124 in order to assign the truck corresponding to the next locus to the delivery destination #i. At the time of returning to step 124, T ≠ 0, and the sum of the maximum load capacity of the trucks already assigned to this delivery destination #i is T. Therefore, by repeating this loop, the delivery destination # Tracks will be allocated one after another until the requirement of i is satisfied.

When D (i) ≤T and the allocation of the truck to the delivery destination #i is completed in step 127, j-1 is substituted into MAX in step 128. At this point, the number of the first locus associated with the delivery destination #i is assigned to the variable MIN, and the number of the last locus is assigned to the variable MAX. Therefore, next, between the variables MIN and MAX. Trucks and shipping destinations for each locus #
It is associated with i, and the process for delivery destination #i is completed (step 129). Then, it is judged whether or not any unprocessed delivery destinations remain (step 130). If no unprocessed delivery destinations remain, it means that the assignment for this chromosome has been completed, and an evaluation value is calculated (step 132). ), And the evaluation value calculation process is completed. When calculating the evaluation value, the evaluation value may be improved by adding the number of used trucks to the evaluation function. On the other hand, if the delivery destination remains in step 130, the process moves to the next delivery destination.
1 is added to i (step 131), and the process returns to step 121.

When it is detected in step 125 that the end of the locus has been reached, it means that there is no unassigned truck despite the unprocessed delivery destination.
This is the case when the prescribed delivery cannot be performed. At this time, in step 133, j is substituted for the variable MAX. At this point, the number of the first locus associated with the delivery destination #i is assigned to the variable MIN, and the number of the last locus is assigned to the variable MAX. Therefore, next, between the variables MIN and MAX. The respective trucks corresponding to the loci of the above are associated with the delivery destination #i (step 13).
4). Then, as a chromosome that cannot perform the predetermined delivery, an extremely bad evaluation value is given to this chromosome,
Alternatively, the unloaded amount is added to the evaluation value (step 13
5), the evaluation value calculation process is completed. However, when D (i) ≦ T and i corresponds to the last delivery destination at the time of step 135, it means that the predetermined delivery can be performed, and thus the normal evaluation value is set as in step 132. It shall be given to this chromosome. Alternatively, the evaluation value may be improved by adding the number of used trucks to the evaluation function.

By performing the above processing on each chromosome 30, the evaluation value of each chromosome 30 can be calculated.

Next, the case of allocating the chromosomes according to the present embodiment and creating the vehicle allocation plan by the genetic algorithm, and the case of using the chromosomes described with reference to FIG. The result of comparing the case where the vehicle allocation plan is created by the statistical algorithm (conventional example) will be described.
The problem setting is as described above in the present embodiment, and the generation change of 500,000 generations is performed based on the present embodiment to obtain the best solution for the vehicle allocation plan. As a result, according to the present example, in the evolution of 4000 generations, a solution with a weight error of 4.3 tons and a time error of 3.3 hours from the best solution was obtained, whereas with the conventional example. For example, with the evolution of the 20000 generation, the weight error from the best solution is 5.0 tons and the time error is 10.
A solution of 6 hours was obtained. In the end, it was confirmed that a better solution can be obtained in a shorter time according to the present invention than in the case of using the conventional method.

In the genetic algorithm, selection pressure is applied to each chromosome according to its fitness, but it is possible to simply select the chromosome in proportion to the fitness (fitness proportional strategy). No, for example, make sure that the best chromosomes survive in the next generation (elite conservation strategy), rank each chromosome in one of several ranks according to fitness, and leave offspring with a probability according to the rank (Rank strategy), calculate the expected value of the offspring left by each chromosome, and if that chromosome is selected, subtract 0.5 from that expected value (expected value strategy). May be. The elite conservation strategy is a strategy for surviving the best chromosomes, and the expected value strategy is a strategy for suppressing the influence of random number fluctuations when the number of chromosomes is small.

[0044]

As described above, according to the present invention, the correspondence relationship between the request elements provided with a plurality of request amounts defined and the plurality of processing elements provided with the processable amounts defined is determined. When solving the combinatorial problem using a genetic algorithm, the processing elements are assigned as the values of each locus of the chromosome without duplication, and the loci are sequentially processed until the sum of the processable amounts exceeds the required amount of the required elements. By repeating searching and reading the processing element and setting the correspondence between the request element and the extracted processing element, the non-redundancy of the chromosome is improved, and the optimum combination can be obtained in a shorter time. The effect is that you can ask for it. Particularly, according to the present invention, it becomes possible to create a scheduling and a delivery plan in the transportation industry in a short time.

[Brief description of drawings]

FIG. 1A is a diagram for explaining an outline of a genetic algorithm, and FIG. 1B is a diagram showing a configuration example of a chromosome.

FIG. 2 is a block diagram showing a configuration of a chromosome assignment device according to an embodiment of the present invention.

3A is a diagram showing an example of chromosome allocation, FIG. 3B is a diagram showing an example of contents of a maximum load amount table, and FIG. 3C is a diagram showing an example of contents of a delivery amount table.

4 is a flow chart showing processing by a genetic algorithm in the apparatus of FIG.

FIG. 5 is a flowchart showing a process of allocating a chromosome and calculating an evaluation value.

[Explanation of symbols]

11 maximum load amount table 12 delivery amount table 13 chromosome storage unit 14 evaluation value calculation unit 15 GA processing unit 16 initial value generation unit 30,200 chromosomes 31,210 loci 111-118, 120-135, 201,203
Steps 202, 204, 205 Population

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location C12Q 1/68 Z 9453-4B G06F 15/18 550 C 9071-5L 17/60 // G05B 13 / 02 M 7531-3H G06F 15/21 C

Claims (4)

[Claims] 1. A combination problem which defines a correspondence relationship between a plurality of request elements each having a defined request amount and a set order and a plurality of prepared process elements each having a processable amount defined therein is genetically defined. A method of allocating a chromosome used for solving by an algorithm, wherein each of the processing elements is assigned as a value of each locus of the chromosome without allowing duplication, and the processable amount or the sum of the processable amounts is Correspondence between the required processing element and the processing element retrieved by sequentially reading the processing elements from the locus on one end side of the chromosome until the required amount of the required elements in the first order of the order is equal to or more than the required amount. And a process of sequentially setting the correspondence relationship with respect to the request elements in the next order, targeting loci from which the processing elements have not been read out. Repeat, allocation method of the chromosome. 2. A genetic algorithm is used to solve a combination problem that defines a correspondence relationship between a plurality of request elements each having a required amount defined and a plurality of process elements each having a processable amount defined. A chromosome allocation device used for that, wherein the combination problem is such that one or a plurality of the processing elements are assigned such that a processable amount that does not fall below the request amount of the request element is assigned to each of the request elements. First search means for searching for a combination that corresponds to a request element and that minimizes or maximizes a predetermined evaluation function, and stores a request amount of the request element for each request element, and for each processing element A second storage means for storing the processable amount of the processing element, and a number of storage elements corresponding to the number of the processing elements. The element is a chromosome storage means that corresponds to each locus of the chromosome, and the processing element is stored by assigning the processing element as a value of the storage element without duplication, and the processable amount or the Until the sum of the processable amounts becomes equal to or larger than the required amount of the request element of the first rank in the first storage means, the storage element is loaded from one end side of the chromosome storage means while referring to the second storage means. A search is performed in order to read the processing elements, a correspondence relationship is set between the request element and the fetched processing element, and the processing elements of the request elements in the next order are not read. A chromosome assigning device, comprising: an assigning unit that repeats the process of setting the correspondence with respect to an element. 3. The chromosome assignment device according to claim 2, wherein the combination problem is a scheduling problem in the transportation industry. 4. The first element corresponds to a destination, and the requested amount is a delivery amount requested from the corresponding destination,
The second element corresponds to one-time transportation by the transportation means,
The chromosome allocation device according to claim 3, wherein the treatable amount is the maximum transportable amount of the corresponding transporting means.