JPH11143938A - Resource assignment plan making method and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To effectively make a resource assignment plan of high quality. SOLUTION: A plan making part 100 searches an optimum resource assignment plan with an evaluation function and outputs this searching result as a proposal plan. A plan correction part 200 corrects the proposal plan acquired at the part 100 and stores the correction information showing the processes including up to the one where the final proposal is prepared into a correction information data base 1600. A plan feature value calculation part 300 calculates the feature value of an assignment plan showing a variance element when a plan is made with the specific environment information based on the correction information stored in the base 1600. An evaluation function generation part 400 generates an evaluation function of the assignment plan based on the trend of the feature value of the assignment plan that is calculated at the part 300.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a resource allocation plan creating method and a resource allocation plan creating system, and more particularly, to a resource allocation plan creating method and a resource allocation plan creating system for supporting a scheduling task in a transportation field or a production field by a computer. .

[0002]

2. Description of the Related Art In the transportation field, for example, there is used a computer-assisted system for preparing a vehicle arrangement plan at a railway depot as a resource allocation plan. Also,
Even in a production field in which a variety of products are produced on a plurality of production lines, a computer-assisted creation system for assigning products to be produced to the production lines as a resource allocation plan is used.

[0003] In this way, a plurality of jobs are
As a search method for a (sub) optimal solution for a resource allocation plan creation problem for allocating resources such as equipment and equipment, linear programming (LP), local search, and local search have been conventionally used. A taboo search method, an annealing method, a genetic algorithm (GA), and the like, which are improved algorithms, are known. Each of these methods has different details,
This is an iteratively improved search algorithm that repeats partial changes to the initial solution created by some means and gradually approaches the optimal solution under the set evaluation function.

Here, as the evaluation function E (X), a linear weighted sum of the evaluation values for each evaluation index shown in equation (1) is used.

[0005]

(Equation 1)

Here, ei (i = 1,..., N) is
A (partial) evaluation function of a single evaluation index, wi (i = 1,
.., N) are weights representing the importance of each evaluation value.

An advantage of using the iteratively improved search method is that it has a high followability to a change in the evaluation index. When a certain evaluation index becomes unnecessary or a new evaluation index becomes necessary, a new evaluation function can be created simply by deleting or adding the corresponding partial evaluation function, and a plan according to the change can be created. Planning is possible. Even when the trade-off between the indices changes, it can be dealt with just by changing the weight of the partial evaluation function. However, since the evaluation index and its weight vary in a complicated manner depending on the plan creation situation and the plan creator, it is difficult for the system developer or the user to set the evaluation index and the weight each time the change occurs.

Therefore, the applicant of the present application has previously filed Japanese Patent Application No.
Japanese Patent Application No. 09713 has filed an application for a method of setting weights, in which the weights of evaluation indices are displayed using a radar chart, and a plan creator graphically corrects the radar chart to adjust the weights.

[0009]

However, the weight adjustment method using the radar chart proposed above has a problem that a plan creator who is not yet accustomed to the operation of the computer system imposes an extra burden. Graphical radar charts make it easier to intuitively understand the trade-offs between evaluation indices than numerical and verbal definitions of the weights, but the weight that changes depending on the creation situation can be set to a satisfactory value with a single adjustment Is rare.
Usually, trial and error are repeated many times, and it is almost always settled to a satisfactory value to some extent, so that the creator requires a great deal of time and effort.

[0010] Furthermore, the creator can formulate a high-quality resource allocation plan using knowledge gained through many years of experience, but it is difficult to formally define the knowledge about plan creation without contradiction. is there. If the plan creator can look at the resulting plan to judge its goodness or point out the corrections to get a good solution, but cannot explain the reason why it is orderly There are many. This fact is the paradox of knowledge engineering,
It is well known that "the better a problem domain expert, the poorer it is at describing the knowledge used to solve his problem." (Waterman, D.M.
A. : A guide to expertsystem
ems, Addison-Wesley, 198.
6).

Therefore, there is a problem that it is difficult for a plan creator to directly operate and appropriately adjust the weight of the evaluation index, which is the plan creation knowledge.

Further, in the conventional method, the evaluation index has no ambiguity, and only an evaluation value that can be uniquely calculated can be used. Therefore, flexibility in changing the evaluation index is low. Therefore, since only fixed evaluation indices can be adopted, it is necessary to arbitrarily change the evaluation function each time the plan creation situation or plan creator changes. However, it is difficult to arbitrarily define the necessary evaluation function due to the paradox of knowledge acquisition described above.

An object of the present invention is to provide a resource allocation plan creating method and a resource allocation plan creating system which realize efficient creation of a high quality resource allocation plan.

[0014]

(1) In order to achieve the above object, the present invention provides a method for creating a resource allocation plan for allocating resources to a plurality of jobs. A first step of searching for a resource allocation plan and outputting the result as a plan, and B) a second step of storing in a correction information database correction information until the plan is corrected and a final plan is created. C) a third step of using the correction information to calculate a characteristic amount of an allocation plan indicating a variable element at the time of planning under specific environmental information; and D) calculating a characteristic amount of the allocation plan. And a fourth step of generating an allocation plan evaluation function based on the tendency. By such a method, an appropriate evaluation function is constructed based on the history of past correction operations of the plan by the plan creator according to the plan creation situation and the difference of the plan creator, and the efficient allocation of the high-quality resource allocation plan is performed. Creation can be realized.

(2) In the above (1), preferably,
The modification information has information on a change in attribute value of a plan attribute including a job attribute and an assigned resource.

(3) In the above (1), preferably,
The environment information includes identification information of each plan creator. In the third step, different environment information is given to each plan creator even when other variables are the same at the time of planning. In addition, a feature amount of an allocation plan that reflects a plan creation method and a preference of each plan creator is calculated from the corresponding environment information and correction information under environment information in the vicinity thereof.

(4) In the above (1), preferably,
The above-mentioned correction information has environmental information at the time of the plan correction,
In the third step, while gradually expanding the range of the neighborhood, correction information in which the current environment information or the vicinity environment information matches the environment information at the time of the correction is collected until the number exceeds a prescribed number, and the collected information is collected. Using the obtained correction information, a feature amount of an allocation plan indicating a variable element at the time of creating a plan under specific environmental information is calculated.

(5) In the above (4), preferably,
The range of the neighborhood is specified by the distance between the pieces of environment information, and the distance function for calculating the distance is a linear sum of partial distance functions corresponding to each item in the environment information.

(6) In the above (1), preferably,
The modification information has a plan attribute value change before and after the plan modification including a plan attribute name, a pre-correction attribute value, and a corrected final attribute value, and the feature amount of the allocation plan is The third step is to search for a change in the planned attribute value of the set of correction information extracted from the correction information database, and to determine the change in the attribute value before correction and the final The values appearing in both of the typical attribute values are counted, and the importance of the plan attribute feature is calculated according to the number.

(7) In the above (6), preferably,
The fourth step calculates the weight of the partial evaluation function for each plan attribute according to the ratio of the importance of all the plan attribute features included in the feature of the allocation plan calculated in the third step. It is something to do.

(8) In the above (1), preferably,
The correction information has a corrected final attribute value, and the feature amount of the allocation plan is a list of pairs of a plan attribute value and a relative frequency of occurrence of the value in the past performance. A set of attribute features of the planning attributes including the value trends,
The third step is to search for a set attribute value change of a set of correction information extracted from the correction information database,
The relative frequency of appearance of the final attribute value excluding the value appearing in the attribute value before correction is statistically calculated and output as a tendency of the value of the attribute feature amount.

(9) In the above (8), preferably,
The fourth step is to create an attribute evaluation function for the planned attribute value so as to output an evaluation value according to the relative frequency of the attribute value that appears in the tendency of the value in the corresponding planned attribute feature value, The evaluation function of the allocation plan is generated using all the attribute evaluation functions.

(10) In order to achieve the above object, in a resource allocation plan creating system for allocating resources to a plurality of jobs, an optimum resource allocation plan under an evaluation function is searched for, and the result is proposed as a plan. A plan drafting section that outputs as a plan draft, a plan drafting section that corrects a draft created by the planning drafting section, and stores correction information until a final draft is created in a correction information database; Utilizing the modification information, a plan feature amount calculating unit that calculates a feature amount of an allocation plan indicating a variable element at the time of creating a plan under specific environmental information, and the plan feature amount calculating unit that calculates the feature amount. An evaluation function generating unit that generates an evaluation function of the allocation plan based on the tendency of the feature amount of the allocation plan. With this configuration, based on the history of past correction operations of the plan draft by the plan creator, an appropriate evaluation function is configured according to the plan creation situation and the difference of the plan creator, and the efficient allocation of a high-quality resource allocation plan is achieved. Creation can be realized.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A resource allocation plan creation method and a resource allocation plan creation system according to one embodiment of the present invention will be described below with reference to FIGS. In the following example, a method for creating a resource allocation plan and a system for creating a resource allocation plan according to the present embodiment will be described taking a system for creating a vehicle arrangement plan in a railway depot as an example. The vehicle layout plan creation system according to the present embodiment is a system that supports a vehicle operation task at a railway depot, and creates a layout plan that determines a layout number line in the base of a vehicle that returns to the depot after commercial operation. Is what you do.

First, the overall configuration of the vehicle layout plan creation system according to the present embodiment will be described with reference to FIG. The detailed processing of each processing unit and database (DB) constituting the vehicle layout plan creation system shown in FIG.
The configuration will be described in detail with reference to FIG.

The planning unit 100 includes a resource data DB 11
Resource data and job data DB stored in 00
A plan is created using the job data stored in 1200, and the created plan is evaluated using an evaluation function stored in the evaluation function DB 1300, and a highly evaluated plan is created as a plan. Then, the plan is stored in the plan DB 1400.

The details of the process of the planning unit 100 will be described later with reference to FIG. The resource data DB 11
The resource data stored in 00 is described later with reference to FIG. The job data stored in the job data DB 1200 will be described later with reference to FIG.

Further, the evaluation function stored in the evaluation function DB 1300 is generated by the evaluation function generation unit 400, and will be described later with reference to FIG. , Through the processing in the plan feature amount calculation unit 300 and the evaluation function generation unit 400,
From the history of past corrections made by the plan creator,
The evaluation function is sequentially modified to an appropriate evaluation function according to the plan preparation situation and the difference between plan creators. In addition, plan DB1
The plan stored in 400 will be described later with reference to FIG.

The plan correction unit 200 includes a plan DB 140
0 is displayed on the display device 2100, and the input device 220
It can be corrected using 0. In modifying the plan, resource data and job data are used, and the environment information stored in the environment information DB 1500 is also used. The revised plan is output to the printer 2300, and is used for arranging vehicles at the depot as a vehicle arrangement plan. The modification information when the plan is modified by the plan modification unit 200 is the modification information DB1.
600 is stored as correction information.

The environment information stored in the environment information DB 1500 will be described later with reference to FIG. Also, regarding the correction information stored in the correction information DB 1600,
This will be described with reference to FIG.

The plan feature calculating section 300 calculates a plan feature based on the environment information and the correction information, and stores the calculated plan feature in the plan feature DB 1700. The planned feature amount will be described later with reference to FIG.

The evaluation function generation unit 400 corrects the evaluation function based on the planned feature quantity, and converts the evaluation function into an evaluation function D
B1300.

That is, in the present embodiment, for plan evaluation,
The plan attribute is introduced as a more primitive unit, and the evaluation of the plan is determined using the value of the plan attribute. The judgment result is given to the plan as an evaluation value. The evaluation function for calculating the evaluation value is generated by the evaluation function generation unit 400 in FIG.
It is configured dynamically based on the modification information according to the plan creation situation and the difference between plan creators.

Next, referring to FIG.
The resource data stored in 00 will be described. FIG. 2 is a layout diagram of a virtual depot serving as a plan creation target in the present embodiment. The information defining this depot is represented by resource data DB 110 as resource data.
0 is stored.

In the present embodiment, the virtual depot has 15 track lines R1, R2,..., R15, and the vehicle moves between the outside and the outside through one entry / exit line. . Each track is classified into three types according to its use. Lines R1 and R2 are lines for performing operation a, which is one of maintenance operations such as cleaning, repair, and inspection of the vehicle, and are provided with dedicated equipment for performing the operation. Similarly, the line R3 is a dedicated line for performing the operation b, which is one of the maintenance operations. Remaining track R
4,..., R15 are indwelling lines that have no special facilities and are used only for storing vehicles until the business start time of the next day. On each track, one train, which is a unit at the time of commercial operation of the vehicle, can be arranged.

The formation is formed by connecting several vehicles.
How many vehicles are considered as one formation depends on the bus schedule for business operation, but in the present embodiment, it is assumed that there are two types of trains, two cars or four cars. Line R
The composition kept at 15 is a 4-car composition. Of course, the two types of knitting have different lengths, so some knitting lines cannot have long knitting.
Only two-car trains can be detained.

A group of several number lines in a base is called a block. The depot shown in FIG. 2 has four blocks from block A to block D. The block is a unit of work at the time of vehicle entry and exit work, and usually one worker is arranged in one block, and is responsible for the entry / exit work of the formation in each block.

Various information on the vehicle base, such as the number, type, length, and blocks of the track, is stored as resource data as
It is created in advance and stored in the resource data DB 1100. This resource data is used in the following arrangement plan creation processing.

Next, referring to FIG.
The job data stored in 200 will be described. FIG. 3 is a data table of a train set to be stored in a vehicle base to be created in the present embodiment, and information in this data table is stored in the job data DB 1200 as job data.

Each information in the job data table shown in FIG. 3 represents an example on a specific day. Item "N
"O" is the identification ID number of the composition. Item "composition number"
Is the name of the storage organization. The item “entry time” indicates the scheduled time at which the train will return to the depot. When the “arrival time” is “railing”, the vehicle has been detained at the depot from the day before the specific day indicated by the job data table. Some trains remain at the depot without commercial operation on the applicable day as a reserve train or to receive maintenance work. Enters.

The item "delivery time" indicates the scheduled time at which the train on the next day will leave the vehicle depot. "Exit time"
On the other hand, the trains that are “railing” have no schedule for commercial operation the next day and stay at the depot all day.

The item “scheduled work” indicates the type of work to be performed, and “work a” is scheduled for work a, which is one of maintenance works such as cleaning, repair, and inspection.
“Work b” indicates that work b, which is one of the maintenance work, is scheduled. The data “absent” is entered for a composition whose work is not specifically scheduled. Item "Number of both"
Is an item for setting both numbers of knitting.

For example, the formation having the identification ID number of "1" of the item "NO" has the formation ID number of "H02", and is a "reserved line" which has been retained at the vehicle depot from the previous day, and the next day is " 7:49 ". In addition, in this knitting, the work schedule of the day is “none”, and it is “4” double knitting. For example, in the knitting with the identification ID number “4” of the item “NO”, the knitting ID number is “H08” , And storage is scheduled at “19:50” on the day. The next day, “6: 5”
2 ”is scheduled to leave. Further, this knitting is scheduled for maintenance work of "work a" on the day, and the items of the knitting data table described above for "4" double knitting are:
This is the minimum information required for creating a plan. For example, when there is other information necessary for creating a plan, such as a train operation schedule and a vehicle type, the information can be freely added and set.

The information on the storage organization is created in advance as job data, and the job data DB 120
0 is stored. This job data is used in the following arrangement plan creation processing.

Next, the processing in the planning unit 100 will be described with reference to FIG. The planning unit 100 stores the resource data DB 11 relating to the line of the allocation source shown in FIG.
00 based on the resource data stored in the evaluation function DB 1300 and the job data stored in the job data DB 1200 relating to the allocation target organization shown in FIG. And outputs it as a placement plan.

The arrangement plan creation problem can be considered as a resource allocation problem in which a resource called a track is allocated to a job of arranging a warehouse organization. The following can be considered as constraints to be considered at the time of planning.

(1) The same track is never assigned to a plurality of knitting sets. (Resource Conflict Constraint) (2) A short track is not assigned to a long train (four cars). (Equipment restrictions) (3) The work line is assigned to the composition in which the work is scheduled. (Work Constraints) The constraints (1) and (2) are physical constraints, and if they are not satisfied, the placement plan cannot be implemented. Restriction (3)
Is related to the transfer efficiency. If the work organization is not directly arranged on the work line, it is necessary to perform a turn to the corresponding line when performing work later, which is a necessary constraint to avoid this. However, this is not the case when the number of work tracks is larger than the number of scheduled work sets.If the work tracks are assigned to all work tracks, the remaining work tracks can be assigned to any set. is there. Conversely, if the number of scheduled work formations is greater than the number of work number lines, a turn will always occur. In this case, the knitting arrangement at the time of warehousing changes due to the transfer line. However, in the vehicle depot targeted in the present embodiment, the number of work track lines is sufficiently large and such a thing does not occur, or Since the number of times is small in proportion to the size of the base, the influence is assumed to be so small that it can be ignored.

Although there are countless arrangement plans that satisfy the above-described constraints, the quality of the arrangement is determined based on various indices for evaluating the plan. In order to carry out the vehicle operation smoothly, it is essential to create a placement plan with high quality, that is, high evaluation. In particular, there is a demand for an arrangement plan that enables efficient entry and exit work and maintenance work, and that can easily be restored even if a trouble occurs.

As an example, an example of an evaluation index relating to the leaving work will be described. The delivery work is performed in block units. However, if the delivery order of the knitting in the block is continuous, the work can be performed one after another without interruption, which is efficient. However, if an optimal plan is used for efficiency, if a trouble such as a breakdown occurs at the time of unloading,
Since there is not enough time, it is not possible to take sufficient countermeasures. On the other hand, in order to perform optimization related to the failure countermeasures, it is necessary to arrange so that the order of delivery of the knitting in the block is not continuous and the intervals are large. That is, the two are in a trade-off relationship, and the extent to which both are to be satisfied varies variously depending on the situation at the time of creating a different plan, such as the number of trains and the number of scheduled operations. In addition, because the ideas and preferences regarding planning differ for each creator,
Even in the same situation, if a plan is created with a fixed balance between the two, even if it is a satisfactory plan for one planner, it is unsatisfactory for another. Often it is a plan.

The above two evaluation indices are intuitively easy to understand because they are directly related to the efficiency of the outgoing work and the measures to be taken in the event of a failure. On the other hand, whether or not the arrangement order in the block should be the dispatch order, whether or not the organization of a specific issue order should be placed on a certain line, and whether or not only the organization with the fastest issue should be placed in a certain block, etc. However, what should be done concretely depends on the situation at the time of preparation and the ideas of the plan creator.
Since it depends on the preference, it is difficult to arbitrarily set a fixed evaluation index in advance for them. Furthermore, it can be said that it is very difficult to define in advance what kind of evaluation target is completely included.

In the case of the warehousing work and various maintenance work, as in the case of the warehousing work, the selection of the evaluation target, the definition of the evaluation index, and the setting of the trade-off depend on the plan creation situation and the plan creator. Therefore, it is difficult to set them in advance in a fixed manner.

The solution search algorithm in the planning unit 100 is an iterative improvement type such as a linear programming (LP), a local search, and a tabu search, an annealing, and a genetic algorithm (GA) which are improved algorithms of the local search. Use an algorithm.

Here, the outline of the processing of the local search method will be described as an example with reference to FIG. In step 410 of FIG. 4, the planning unit 100 forms an initial solution X of the arrangement plan. Here, a solution that satisfies at least the preset constraint conditions (in the present embodiment, the three conditions of resource conflict, facility, and line constraint described above) is created. Next, in step 420, the planning unit 100
Calculates the evaluation value E (X) of the initial solution using the evaluation function E.

In step 430, the planning unit 100 forms the neighborhood N (X) of the initial solution X. The neighborhood is a set of solutions obtained by making a small change to a certain solution (here, X). In the present embodiment, a set of solutions obtained by exchanging the allocation number lines of any two formations for the solution X (as long as the constraint number is not violated) is defined as a neighborhood N (X).

Next, in step 440, the planning unit 100 evaluates the evaluation values E of all the elements in the neighborhood N (X).
(Y) is calculated using the evaluation function E, and the calculated evaluation value E (Y) is compared with the initial evaluation value E (X). As a result, if there is a solution Y having an evaluation value higher than X, the process proceeds to step 450, where the solution Y is replaced with the solution X again, the process returns to step 430, and the process is repeated. Neighbor N
If any solution of (X) has a lower evaluation value than X, the process goes through the loop to step 460, outputs X as a plan to the plan DB 1400, and ends the processing.

Here, referring to FIG. 5, the planning unit 100
An example of the plan drafted in the above will be described. FIG. 5 is a data table of an arrangement plan showing the allocation of the formations to be stored in each of the track lines R1 to R15 of the depot for which a plan is to be created in the present embodiment, and each information in this data table is a plan data. As a plan DB1400
Is stored in

The layout plan is in the form of a table, and the item "number line" represents the number lines R1,..., R15 in the vehicle depot. The item “composition ID number” represents the composition ID number of the composition arranged on each track. The line in which the data of the item “composition ID number” is “...” Indicates that no composition is arranged there.

The number lines R1, R2, and R3 are work number lines as described with reference to FIG. Therefore, in the job data shown in FIG. 3, the track lines R1, R2, and R3 are allocated to the compositions having the scheduled items "composition ID numbers" of "H08", "H10", and "H07". It represents that it is. Also, in the job data shown in FIG. 3, the item “composition ID number” that is scheduled to work is “H0
For the formation of "3" and "H04", the track R
7 and R9 are assigned, and are moved to the work number line after the work on the number lines R1 to R3 is completed. For other knitting, no work is scheduled, so in addition to the work line,
This indicates that they are allocated in accordance with the scheduled entry time, the scheduled exit time, and the number of trains.

Next, referring to FIG.
The method of correcting the plan in the above will be described. The plan created in the planning section 100 is a plan correction section 2
At 00, an edit process is manually performed by a plan creator. For the editing process, the plan correction unit 200 compares the plan stored in the plan DB 1400 shown in FIG.
1 is displayed on the display screen of the display device 2100 shown in FIG. 1 based on the job data stored in the job data DB 1200 shown in FIG.

FIG. 6 shows an example of a plan correction screen displayed on the display device 2100. On the screen, the ID number of the organization arranged on each track in the plan being edited and the plan attribute value are always displayed. As the planning attribute value,
"Track number", "Order of entry", "Order of exit", "Scheduled work"
Etc. are included.

Here, the plan attributes will be described with reference to FIG. Plan attributes are any information that can be read from the plan and considered by the creator when creating the plan, i.e., likely to be relevant to the quality of the plan.

In the present embodiment, the plan attribute is shown in FIG.
Two types are defined, one for each line shown in (a) and one for each block shown in FIG. 7 (b). Although both are organized in a table format, each data in the table represents one independent plan attribute in a certain plan. For example, as the planning attributes of the individual track numbers shown in FIG. 7A, the “order of entry” of the “number track” R1 is “1”, the “order of dispatch” is “3”, and the “work schedule” is "Work a" indicates that there is something. "Track number" R
The “delivery order” of No. 9 is “99”, which indicates that the detention is not scheduled to leave the next day. In addition, as the plan attribute shown in FIG. 7B, the plan attribute of “block A” indicates that “arrangement in order of receipt” is “YES”. The “arrangement in order of entry” being “YES” means that vehicles entering the tracks R1, R2, and R3 constituting the block A enter in the order of the tracks R1, R2, and R3. ing. In addition, it indicates that “delivery order” is “YES”. The “deliver order in order” is “YES”, which means that vehicles entering the tracks R1, R2, and R3 constituting the block A leave in the order of the tracks R1, R2, and R3. ing. In addition, the number of incoming trains
Is "3", and the "number of outgoing trains" is "3". If the train has been on the track from the previous day or has been on the track the next day, the "number of incoming trains" and "number of outgoing trains" are displayed as shown in the planning attributes of "block C" and "block D". Will not match.

An evaluation object can be formed by combining some planning attributes. For example, the evaluation target “order of dispatch in block A” includes the planning attributes “order of dispatch of line R1”, “order of dispatch of line R2”, and “order of order of line R2”.
3 is the order of delivery. Conversely, an attribute that can constitute an expected evaluation target is defined in advance as a plan attribute.

Here, on the display screen shown in FIG.
If the information cannot be accommodated in one screen due to a large number of track lines, only a part of the information is displayed on the screen as shown in the figure, and the scroll bar 611 is moved to the pointer 6 by the remaining part.
The image can be displayed by specifying and sliding by the user 12.

The plan creator can correct the plan by a graphical operation using a pointing device such as a mouse or a keyboard as the input device 2200 or a keyboard.

The correction of the plan is performed by repeatedly applying some correction operations. One of them is an operation of exchanging the knitting arranged on any two track numbers.
More specifically, a replacement target line is specified on the screen using a pointer. For example, the display of the designated line is inverted so that it is easy to understand that the designated line has been designated. FIG.
In the example shown in FIG. 7, if the line R3 and the line R7 are the line to be replaced, the line R3 and the line R6 are input to the input device 220.
By designating with 0, those displays are inverted. Thereafter, when the exchange button 601 is instructed and pressed by the pointer, the exchange is performed, and the display of the composition ID number and the plan attribute value changes accordingly. That is, on the track R3, the vehicle with the formation ID number “H03” is stored in the parking order “4”.
In the delivery order “10”, the work schedule is “work b”.
Further, on the track R7, the vehicle with the formation ID number "H07" is scheduled to enter the storage order "3", exits the order "9", and becomes the work schedule "operation b". However, if the exchange creates a plan that violates the constraints, the operation is invalidated and a message to that effect is displayed. For example, when the exchange button 601 is pressed with the line R2 and the line R3 as the exchange targets, the "operation b" cannot be performed on the line R2 and the "operation a" cannot be performed on the line R2. 3) It violates the work constraint, invalidates the replacement operation, and displays a message indicating that the replacement operation is invalid. The message tells you which of the constraints is violated,
Display its contents. In the above example, for example, a message such as "violates the work constraint. Work b cannot be performed on line R2. Work a cannot be performed on line R3" is displayed. The specified track is highlighted or the specified 2
The process of exchanging one line and displaying a message is performed by the plan correction unit 200.

In addition to the exchange operation described above, a release operation for designating a composition to which a line is assigned by a pointer and releasing the arrangement number, and specifying a composition to which an arrangement number has not been allocated to specify an arbitrary number. An assigning operation can also be used.

When the user wants to invalidate an operation immediately after performing the operation, the user can press the button 602 to return the plan to the state before the correction operation. By repeating the above correction operations one after another,
The plan creator can make the plan he wants.

On the plan amendment screen shown in FIG. 6, the plan creator can obtain various information relating to the planning. For example, when the button 603 is pressed, the information on the organization stored as the described job data shown in FIG. 3 can be viewed. If you press this button after specifying a specific track with the pointer,
It is also possible to acquire only the information on the composition arranged on the track.

A button 604 is for acquiring environment information stored in an environment information DB 1500 described later with reference to FIG. A button 605 is used to display a correction frequency graph indicating a transition of the number of times of plan correction by the plan creator.

Here, an example of the correction frequency graph will be described with reference to FIG. As shown in FIG. 8, the correction frequency graph in the present embodiment is a bar graph in which the horizontal axis represents the date and the vertical axis represents the number of corrections. A label 701 represents a plan creator ("T01") that is the target of this graph, and is the same person as the current plan creator. The number of corrections is
It means the cumulative number of valid correction operations performed by the plan creator in one plan creation. Here, the number of different correction operations may be distinguished and displayed on another graph. For example, one replacement operation is equal to two cancellation or allocation operations, and the number of all correction operations is weighted. It is good to put it together.

By looking at this modification frequency graph, the plan creator can visually grasp that the burden on his / her own plan creation is gradually reduced as the system proceeds with learning.

In the display screen shown in FIG.
By pressing 06, the plan creator can know the features of his / her own plan creation method estimated by the system in the same plan creation situation as this time.

More specifically, an attribute feature quantity to be described later and a plan feature quantity as a set thereof are displayed in an easily understandable manner using a graph, a table, or the like. This allows plan makers to:
You will be able to clearly understand how to create your plan, including the parts you have done unconsciously.

When the modification of the plan is completed,
An approval button 607 is pressed. Thereby, the plan is registered in the plan DB 1400 shown in FIG. 1 as a formal plan used for actual vehicle operation. At the same time, correction information summarizing information obtained from the correction process of the plan creator is created and stored in the correction information DB 1600. Details of the correction information will be described later with reference to FIG.

By pressing the print button 608, the current plan can be printed from the printer 2300 in a form in a form.

Next, referring to FIG. 9, the modification information DB 160
The correction information stored in 0 will be described. Note that FIG.
The correction information shown in FIG. 6 is correction information obtained when a correction work for exchanging the line R4 and the line R8 is performed on the display screen shown in FIG. The correction information is data in a table format, and is composed of four items as shown.

The item “environment” indicates environment information at the time of creating the plan. Specifically, a pointer “x” to a data table of environment information described later with reference to FIG. 10 is set. The item “final plan plan attribute” represents the plan attribute of the final plan draft that has been approved as a formal plan. Specifically, a pointer “v” to the data table of the plan attribute shown in FIG. 7 is set. The item “number of corrections” is the number of valid correction operations performed by the plan creator until the final plan is obtained.

The item “attribute value change” indicates a difference in plan attributes between the plan before receiving the correction process by the plan creator and the final plan. The attribute value change is a list format of information including three items of “attribute name”, “attribute value before correction”, and “attribute value after correction”. For example, as a result of exchanging the line R4 and the line R8, "the order of entry of the line R4"
Changes from "7" before the correction to "8" after the correction, and the "order of entry of track R8" changes from "6" before the correction to "3" after the correction. It represents that. The same applies to the modification of the delivery order. "Block B receiving order arrangement"
Before the correction, line R4 to line R6 for block B
Until “YES”, which is a storage order arrangement in which the lines R4, R5, and R6 are allocated in the order of storage, after correction, the lines R4 to R6 for the block B and the line R in the storage order.
4, N, N
O ".

Next, referring to FIG.
The environment information stored in 00 will be described. The environment information used in the present embodiment is information on a plan creation environment that summarizes information on a plan creation status and a plan creator.

As shown in FIG. 10, the item “creation date” of the environment information indicates the date of the date on which the plan was created. The item “creator” contains a creator ID number representing the plan creator.
The item “composition total number” is the total number of compositions for which a plan is to be created. As shown in the job data 1200 shown in FIG. 3, since 12 plans have been created, the “composition total” is “12”.

The item “number of incoming trains” is the number of incoming trains, and is “9” in the example of the job data shown in FIG. The item “number of outgoing knitting” is the number of outgoing knitting.
Is "11" in the example of the job data shown in FIG. The item “number of work a formations” is the number of formations scheduled for work a, and is “3” in the example of the job data shown in FIG. The item “number of work b formations” is the number of formations scheduled for work b. In the example of the job data shown in FIG.
It becomes "2".

The item “diamond pattern” is a combination of a business schedule on the current day and the next day. Generally, sales schedules are divided into three days: weekdays, days before holidays (Saturday), and holidays (Sunday).
They are classified into different types, and the number of trains that operate for commercial operation and the operating hours are significantly different. The example shown in FIG. 10 indicates that the current day is a “holiday” and the next day is a “weekday”.

The items of environmental information shown in FIG. 10 are the minimum necessary for creating a plan. During system development or during system operation, if it is determined that other items (for example, block staff, work required time) are significantly related to the planning, they can be freely added.

Next, the processing of the planned feature quantity calculation unit 300 will be described with reference to FIGS. The plan feature amount calculation unit 300 calculates a plan feature amount based on the environment information stored in the environment information DB 1500 and the correction information stored in the correction information DB 1600. The plan feature is a set of attribute features. The attribute feature amount is obtained by estimation from a past plan correction example by a plan creator.

In step 1110 of FIG. 11, the plan feature quantity calculator 300 initializes the value of the distance variable k to “0”. The variable k is used to specify the distance between environment information in the subsequent processing. The distance between the environment information X and Y is a distance function D represented by the following equation (2).
Given by (X, Y).

[0087]

(Equation 2)

Here, n is the number of items of environmental information, and d is
i is a (partial) distance function relating to the item i, and xi and yi are the values of the items of the environment X and the environment Y, respectively.

In the present embodiment, the partial distance function of each item is given by the following equations (3) to (8).

[0090]

(Equation 3)

[0091]

(Equation 4)

[0092]

(Equation 5)

[0093]

(Equation 6)

[0094]

(Equation 7)

[0095]

(Equation 8)

Next, in step 1120, the plan feature quantity calculation unit 300 forms a neighborhood Nk (X) of the input environment information (this is X) near the distance k. The neighborhood is a set of environment information obtained by slightly changing certain environment information (here, X). The degree of the change is designated by the distance k between the environment information. Neighborhood Nk
The formal definition of (X) is shown in the following equation (9).

[0097]

(Equation 9)

In step 1130, the plan feature quantity calculation unit 300 searches all the correction information in the correction information database 1600, and counts those whose environmental information matches elements in the neighborhood Nk (X). Then, the number is compared with a preset value. If the number of pieces of correction information is less than the specified value, the process proceeds to step 1140, where the value of the distance variable k is increased by 1 (that is, the neighborhood is enlarged), the process returns to step 1120, and the process is repeated. If the number of correction information exceeds the specified value, step 115
Go to 0.

In step 1150, plan feature quantity calculation section 300 collects all pieces of correction information that match the environment information in neighborhood Nk (X) (this is called DAT).

Next, in step 1160, the plan feature quantity calculation unit 300 sets the DAT collected in step 1150.
, The internal step is applied to all the plan attributes, and the attribute feature amount of each plan attribute is calculated.

In sub-step 1162, plan feature quantity calculating section 300 calculates the importance of the target plan attribute. Here, the details of the importance calculation step will be described with reference to FIG. In step 1162A, the plan feature amount calculation unit 300 sets a specified value common to all plan attributes as an initial value of importance. Next, step 1
In 162B, the plan feature amount calculation unit 300
Search for the attribute change item of the correction information in. Then, specific values of the plan attribute appearing in both the item before and after the attribute change are counted. For example, the planning attribute “line R
When the storage order “5” appears in the item before correction in the correction information having the DAT and the item after correction in the other correction information, the storage order “5” is counted. This means that the target plan attribute has less restrictions on the setting of the value, and therefore has less involvement in the creation of the (quasi) optimal plan (that is, its importance is low).

Next, in step 1162C, the planned feature quantity calculation unit 300 reduces the importance from the initial value in proportion to the number of the counted values. In the present embodiment, the importance is calculated using Expression (10) shown below.

[0103]

(Equation 10)

Next, returning to FIG.
In 4, the plan feature quantity calculation unit 300 calculates the tendency of the target plan attribute. Here, the details of the process of calculating the tendency of the plan attribute will be described with reference to FIG. In step 1164A, the plan feature amount calculation unit 300
The attribute change item of each correction information in T is searched. And
List all values that appear in the item before modification of the target plan attribute. Since the value appearing in the item before correction is considered to be an unnecessary value for the (sub) optimal plan, it is excluded from the target for calculating the tendency of the attribute feature amount in the next step.

Next, in step 1164B, the plan feature quantity calculating unit 300 statistically calculates the frequency of appearance of the remaining values excluding the listed values in the final plan plan attribute of the correction information in the DAT. , And outputs the list as the tendency of the attribute feature amount.

Returning again to FIG.
In 6, the plan feature amount calculation unit 300 creates an attribute feature amount by adding a plan attribute name and an environment to the importance and tendency calculated in the previous two substeps 1162 and 1164.

Under a specific plan creation environment, the values of the plan attributes of the most desirable plan show a certain tendency. In addition, the importance of which of the plurality of plan attributes is to be prioritized also depends on the plan creation environment.
Data in which such information is formally defined is referred to as an attribute feature, and a collection of attribute features related to all plan attributes is referred to as a plan feature.

Here, the attribute characteristic amount will be described with reference to FIG. As shown in FIG. 14, the attribute feature amount is defined as table data. Item "plan attribute name"
Represents the name of the target plan attribute. In the example shown in the figure, the “order of dispatch of the line R3” is “plan attribute name”.

The item “environment” indicates a plan creation environment in which the feature amount is valid. Specifically, the pointer “x” to the environment information data table described in FIG. 10 is set in this item.

The item "importance" represents the importance of the target plan attribute. The importance is determined by step 1162 in FIG.
In the case where the importance is low, the restriction on the setting of the value of the target plan attribute is loose, and therefore, the involvement in the creation of the (quasi) optimal plan is low. When the importance is high, the restriction on the setting of the value is severe, and the involvement in the creation of the optimal plan is high. In the illustrated example, the “importance” is “80”, which indicates that the importance is high.

The item “trend” indicates the tendency of the value of the plan attribute value. Here, the tendency is a list in which a possible value of the plan attribute value and a relative appearance frequency of the value in the past performance are paired. In the illustrated example, “the order of dispatch of the line R3 is 80 under the environment X, and the tendency of the values is 60% for 9; 30% for 10; and 10% for 14”.
Can be read. As for “trend”, for example, as shown in FIG.
When it is determined as the th, "9" is counted as one time. Also,
In another modification, replacing track R3 and track R7,
The delivery order of the line R3 is “10”, and in this case, “10” is once. In this manner, the “trend” represents the relative frequency of appearance of the value given as the delivery order.

Finally, in step 1170, the plan feature amount calculation unit 300 collects the attribute feature amounts of all the plan attributes, outputs them to the plan feature amount DB 1700 as plan feature amounts in the input environment, and ends the processing. I do.

Next, referring to FIG. 15, the evaluation function generation unit 4
00 will be described.

In step 1510, the evaluation function generation unit 400 extracts the importance from all the attribute features included in the plan features stored in the plan feature DB 1700, and evaluates each plan attribute based on the extracted importance. Calculate function weights. Various methods can be used for calculating the weight. In the present embodiment, the weight is calculated using the following equation (11).

[0115]

[Equation 11]

Here, wi (i = 1,..., N) is the weight of each evaluation value, n is the total number of plan attributes, and si (i = i
1,..., N) are the degrees of importance of each plan attribute.

Next, in step 1520, the evaluation function generation section 400 applies the internal step to all plan attributes. In sub-step 1522, the evaluation function generation unit 400 extracts an attribute feature amount related to the target plan attribute from the plan feature amount.

In sub-step 1524, evaluation function generation section 400 generates an evaluation function for the target plan attribute from the tendency of the attribute feature amount. Here, the evaluation function is created such that a numerical value that is higher in proportion to the relative appearance frequency of the plan attribute value is output and the maximum value matches a preset specified value α common to all plan attributes. . In the present embodiment, an evaluation function is created using the following equation (12).

[0119]

(Equation 12)

Here, e (x) is an evaluation function for the plan attribute value x, n is the number of values in the trend list,
vi (i = 1,..., n) are each value in the trend list, and p
i (i = 1,..., n) is the relative appearance frequency, and pmax is the maximum value of the relative appearance frequency.

Next, in step 1530, the evaluation function generation section 400 evaluates the evaluation function e (x) for all the planning attributes created in step 1520,
The evaluation function E (X) of the entire plan is created using the weight w of each evaluation function created at 0, and is output to the evaluation function DB 1300. In the present embodiment, the overall evaluation function E (X) is created using the following equation (13).

[0122]

(Equation 13)

The evaluation function E (X) generated by the evaluation function generation section 400 is used in the planning section 100 as described with reference to FIG.

Each time a new plan is created and corrected, the above-described processing, that is, the processing in the planning unit 100, the plan correcting unit 200, the plan feature amount calculating unit 300, and the evaluation function generating unit 400 is repeated. It is.

As described above, according to the present embodiment, the problem of resource allocation plan creation in which the evaluation criterion changes variously depending on the situation of plan creation and the plan creator (creation environment),
By estimating the importance of the attribute of the allocation plan and the tendency of its value in a specific environment and forming an evaluation function based on the result, it is possible to make a good quality plan in the relevant environment.

Also, in order to accumulate the correction information obtained from the process of correcting the plan by the plan creator and estimate the tendency of the importance and the value of the attribute by using the information, the plan creator is instructed. It is possible to acquire knowledge related to planning without extra burden.

Further, since the accuracy of estimation increases as the accumulation of correction information progresses, the burden on the plan creator of the correction work can be gradually reduced, and ultimately the optimal allocation plan can be determined without involvement of the plan creator. Automatic creation becomes possible.

[0128]

According to the present invention, efficient creation of a high-quality resource allocation plan can be realized.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram showing an overall configuration of a vehicle arrangement plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 2 is an explanatory diagram of resource data used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 3 is an explanatory diagram of job data used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 4 is a flowchart illustrating processing performed by a planning unit in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 5 is an explanatory diagram of a plan drafted in a vehicle allocation plan generation system to which a resource allocation plan generation method and a resource allocation plan generation system according to an embodiment of the present invention are applied.

FIG. 6 is an explanatory diagram of a method for modifying a plan in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 7 is an explanatory diagram of plan attributes used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 8 is an explanatory diagram of a correction frequency graph that can be displayed in a vehicle allocation plan creation system to which the resource assignment plan creation method and the resource assignment plan creation system according to one embodiment of the present invention are applied.

FIG. 9 is an explanatory diagram of correction information used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 10 is an explanatory diagram of environment information used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 11 is a flowchart illustrating a process of a plan feature amount calculating unit in the vehicle allocation plan creation system to which the resource assignment plan creation method and the resource assignment plan creation system according to one embodiment of the present invention are applied.

FIG. 12 is a flowchart showing a process of a sub-step in FIG. 11;

FIG. 13 is a flowchart showing a process of a sub-step of FIG. 11;

FIG. 14 is an explanatory diagram of attribute feature amounts used in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

FIG. 15 is a flowchart showing the processing contents of an evaluation function generation unit in a vehicle allocation plan creation system to which a resource assignment plan creation method and a resource assignment plan creation system according to an embodiment of the present invention are applied.

[Explanation of symbols]

 100 Plan Planning Unit 200 Plan Modification Unit 300 Plan Feature Calculation Unit 400 Evaluation Function Generation Unit 1100 Resource Data DB 1200 Job Data DB 1300 Evaluation Function DB 1400 Plan DB 1500 Environmental Information DB 1600 ... Modification information DB 1700 ... Plan feature DB 2100 ... Display device 2200 ... Input device 2300 ... Printer

Claims (10)

[Claims] 1. A method for preparing a resource allocation plan for allocating resources to a plurality of jobs: A) a first step of searching for an optimal resource allocation plan under an evaluation function and outputting the result as a plan B) a second step of storing, in the correction information database, correction information until the final draft is created by correcting the plan draft; and C) using the above-mentioned correction information, under the specific environmental information. A third step of calculating a feature amount of the assignment plan indicating a variable element at the time of creating the plan; and D) a fourth step of generating an evaluation function of the assignment plan based on the tendency of the feature amount of the assignment plan. A method for creating a resource allocation plan, comprising: 2. A resource allocation plan creating method according to claim 1, wherein said modification information includes information on a change in an attribute value of a plan attribute comprising a job attribute and an allocated resource. Method. 3. The resource allocation plan creating method according to claim 1, wherein said environment information includes identification information of an individual plan creator, and said third step includes the steps of: Even if the same is the same, different environment information is given to each plan creator, and from the corresponding environment information and the correction information under the environment information in the vicinity thereof, the allocation plan reflecting the plan creation method and preference of each plan creator is reflected. A resource allocation plan creation method characterized by calculating a feature amount. 4. The resource allocation plan creating method according to claim 1, wherein the correction information includes environmental information at the time of correcting the plan, and the third step includes gradually expanding a range in the vicinity. , Correction information matching the current environment information or the environment information at the time of correction and the environment information at the time of correction are collected until the number exceeds a specified number, and using the collected correction information, A resource allocation plan creating method, comprising calculating a feature amount of an allocation plan indicating a variable element at the time of creating the plan. 5. The resource allocation plan creation method according to claim 4, wherein the range of the neighborhood is specified by a distance between environment information, and a distance function for calculating the distance corresponds to each item in the environment information. A method for creating a resource allocation plan, which is a linear sum of partial distance functions. 6. The resource allocation plan creating method according to claim 1, wherein the modification information includes a plan attribute name, a pre-modification attribute value, and a modified final attribute value. Wherein the feature amount of the allocation plan is a set of attribute feature amounts of plan attributes including the importance of the plan attributes, and the third step is a set of correction information extracted from the correction information database. Search for changes in the planning attribute value of
A resource allocation plan creating method, characterized in that values appearing in both the before-correction attribute value and the final attribute value are counted, and the importance of the plan attribute feature amount is calculated according to the number. 7. The resource allocation plan creating method according to claim 6, wherein the fourth step is a step of determining importance of all plan attribute features included in the feature of the allocation plan calculated in the third step. A weight of a partial evaluation function for each plan attribute is calculated according to the ratio of the resource allocation plan. 8. The resource allocation plan creating method according to claim 1, wherein the correction information has a corrected final attribute value, and the feature amount of the allocation plan is a plan attribute value and its value. Is a set of attribute feature amounts of plan attributes including a tendency of a value which is a list of pairs of relative appearance frequencies in the past performance of the modification information, and the third step is a step of modifying the modification information extracted from the modification information database. Search for changes in the set attribute values of the set,
Resource allocation plan creation characterized by statistically calculating the relative frequency of appearance of the final attribute value excluding the value appearing in the attribute value before correction and outputting as a tendency of the value of the attribute feature amount Method. 9. The resource allocation plan creating method according to claim 8, wherein the fourth step is to calculate an attribute evaluation function for the planned attribute value by using a relative value of the attribute value which appears in the tendency of the value in the corresponding planned attribute value. A method for creating a resource allocation plan, wherein an evaluation value is generated in accordance with the frequency of the frequency, and an evaluation function of the allocation plan is generated using all the attribute evaluation functions. 10. A resource allocation plan creating system for allocating resources to a plurality of jobs, a planning section for searching for an optimal resource allocation plan under an evaluation function and outputting the result as a plan. The plan drafting section corrects the draft created by the planning section, and stores the revised information until the final draft is created in the revised information database.Using the revised information stored in the revised information database, A plan feature amount calculating unit that calculates a feature amount of an allocation plan indicating a variable element at the time of creating a plan under specific environmental information; and a trend of the feature amount of the allocation plan calculated by the plan feature amount calculating unit. A resource allocation plan creation system, comprising: an evaluation function generation unit that generates an evaluation function of an allocation plan.