JPH01114979A - Scheduling method - Google Patents

Abstract

PURPOSE:To obtain a proper scheduling solution within a practically short time by using required restricting conditions or the like as input data and sequentially improving a proper initial state for job assignment to obtain the state of the solution. CONSTITUTION:In case of executing scheduling based upon an input, a specified restricting condition is eased and the set of properly selected jobs is assigned to respective machines to form 1 the initial state. When the restricted condition is completely satisfied after checking the state, the solution of scheduling is outputted 6. When the condition is not satisfied, a partial set to be replaced is retrieved 3 as a swapping candidate in different machines. When the improvement of an evaluation value obtained by executing the replacement of the swapping candidate is decided, the current state is updated 5 to a new state. The retrieval and replacement of swapping candidates are repeatedly executed until the new state reaches the scheduling solution.

Description

[Detailed Description of the Invention] [Summary] Regarding computer processing for solving a scheduling problem in which jobs are assigned to machines so as to satisfy required constraints, it is possible to obtain a reasonable scheduling solution in a practical processing time. The purpose of the present invention is to provide a scheduling processing method in which a condition representing a machine, a condition representing a job, and a constraint condition are input, and a state in which the job is assigned to the machine is generated so as to satisfy the constraint condition. An evaluation value and a judgment condition are set, a state that does not necessarily satisfy all of the constraints is generated as an initial state, and when the state satisfies all of the constraints, the state is used as a scheduling solution. output, and if the state does not satisfy all of the constraint conditions, search for a part of the job assignment for the state that can be exchanged between different machines as a swap candidate. and execute the exchange if it is determined based on the determination condition that the evaluation value as a result of executing the exchange of the swap candidate is improved from the evaluation value before executing the exchange. and generates a new state, and repeats the search for the swap candidate and the execution of the swap until the new state becomes the scheduling solution.

[Industrial application field]

The present invention relates to a computer-processed scheduling processing method for solving a scheduling problem of allocating jobs to machines so as to satisfy required constraints.

In this case, a machine is an object that must be bound for a certain period of time by a job, and in most cases is an object that executes an assigned job. In this case, a truck or the like that transports cargo is a machine.

[Problems to be solved by the prior art and the invention] For example, given n business offices and m trucks, a truck operation plan that realizes the required number of flights between these business offices. Consider the processing of a scheduling problem that involves

In that case, the geographical relationship of the business offices (routes connecting the business offices,
(e.g., the time required for the operation, etc.), the operation of each office, the time required for loading, unloading, etc. at a certain location, business hours of the office, requests for the arrival and unloading times of cargo, etc. for each office, etc. is given as a constraint.

The solution to this problem is, for example, at what time does each truck T1, T2, etc. leave office A, arrive at office B, and leave office B after working there for how many minutes? The schedule for one day is such that the driver departs from office C and heads to office C, and such a solution must be found so as to satisfy the above-mentioned constraints.

In general, such a scheduling problem can be viewed as a problem in which the truck is considered a "machine" and the operation from office A to B is a "job" to be processed by that machine, and the job is assigned to the machine. can.

To solve such problems, there are analytical methods such as so-called operation research that convert the problem into a formula and find the solution mathematically, but in real-life problems like the one mentioned above, unless the problem is greatly simplified, Since it is extremely difficult to express it mathematically, there are many cases where it cannot be applied.

Furthermore, even if formulating is possible, if constraints change, the structure of the formula generally needs to be reconsidered, and it is often difficult to respond to changes in the situation.

Therefore, such analytical methods are abandoned, and job assignments for machines are generated and checked to see if they satisfy the requirements and constraints, successively covering all combinations of job assignments. , if one that satisfies the conditions can be obtained, a method can be considered in which that assignment is used as a solution, but with this method, as the number of machines and jobs increases, the number of different assignment states will rapidly become enormous, and the so-called This results in a combinatorial explosion, making it impossible to reach a solution in a realistic processing time.

An object of the present invention is to provide a scheduling processing method that enables a reasonable scheduling solution to be obtained in a practical processing time in a non-analytical approach such as the latter.

[Means for solving problems]

FIG. 1 is a process flowchart showing the configuration of the present invention.

The figure shows the flow of processing for generating a scheduling solution indicating a state in which the job is assigned to the machine so as to satisfy the constraint conditions by inputting conditions representing a machine, conditions representing a job, and constraint conditions. -6 are processing steps.

[For production]

A condition representing a machine, a condition representing a zip, and a constraint condition are used as input data, and a definition of an evaluation value and a judgment condition for evaluating the status of zip allocation to the machine are given, and the scheduling process is started, and processing step 1 is performed. Then, under the relaxed constraint conditions, a state indicating the allocation of all zips to each machine is appropriately generated and set as an initial state.

In processing step 2, it is checked whether the state satisfies all of the constraint conditions, and if all of the constraint conditions are satisfied, in processing step 6, a scheduling solution is created and output based on the state.

If the state does not satisfy all of the constraints,
Regarding the state in processing step 3, if a part of the job assignment that can be exchanged between different machines is searched and set as a swap candidate, and in processing step 4 the swap candidate is exchanged, the above state Check whether the evaluation value is improved from the current evaluation value according to the judgment conditions.

If it is determined that the evaluation value will be improved by exchange,
After performing the exchange and generating a new state in process step 5, the process returns to process step 2 and repeats from checking the constraint conditions.

If the evaluation value is not improved, the process returns to step 3 to search for another swap candidate, and repeats the above process until the new state becomes the scheduling solution.

This processing method gradually improves the initial state of appropriate job assignments and brings the state closer to the solution, thereby limiting the number of states that need to be tested and obtaining at least one solution in a relatively short practical processing time. becomes possible. Furthermore, since constraints are provided as data separate from processing procedures, it is easy to add, change, or delete constraints.

〔Example〕

As a scheduling problem, for example, m trucks are specified as a condition representing a machine, and then "flights from office P to office Q" are determined between n offices.
A problem for assigning jobs is input.

Scheduling constraints include the initial location of trucks, the geographical relationship of offices (routes connecting offices, time required for travel between offices, etc.), and the amount of time spent on loading, unloading, etc. for each office. The time, business hours of the office (that is, limits on departure and arrival times of flights), requirements for arrival and unloading times of cargo, etc. for each office are specified.

In addition, the definition of an evaluation function that calculates the evaluation value of the zip allocation state,
Judgment conditions for determining the direction of improvement of the evaluation value are set.

That is, as an evaluation value, for example, the excess time of each flight is defined, and as a judgment condition, for example, this evaluation value is defined as -
Conditions are set to shift the state in a direction that reduces the amount of energy.

Here, the status is represented by a set of ordered jobs assigned to each machine, as shown below, for example. That is, for example, machine 1 = (job 1. job 3. -・-・・-) machine 2
= (Job 4. Job 5, --------------・) Machine m = (Zip 2. Zip 7, ----------------------
·)become that way.

When scheduling is performed based on the above inputs, in processing step 1 in Figure 1, initialization processing such as expanding input data into a predetermined internal representation convenient for processing is performed, and then the initial state is generate. In order to easily generate the state, we have significantly relaxed the specified constraints,
For example, considering only the routes between offices under the above conditions, a set of appropriately selected jobs is sequentially assigned to each machine and set as an initial state.

Processing step 2 checks the state based on constraint conditions, and if all the constraint conditions are satisfied, the state is set as one scheduling solution, and processing step 6 outputs a solution in a predetermined format.

However, even for the state solved in this way, by further continuing the processing described below, it is generally possible to obtain another solution that is more improved based on the evaluation value, so if necessary, A bus (indicated by a broken line in the figure) is provided for returning from processing step 6 to processing step 3 and continuing the processing.

If it is determined in processing step 2 that the state does not satisfy all of the constraints, the process proceeds to processing step 3, and from the job assignment of that state, subsets of jobs that are assigned to different machines and are mutually interchangeable are determined. Search for and select it as a swap candidate.

In the above example, the interchangeable subset includes, for example, a row of jobs that are assigned to two different machines and that consist of one or more jobs that line up after jobs that have the same starting office; Alternatively, there may be a line for each job, from jobs that have the same departure office to jobs that have the same arrival office.

In addition, a job or a sequence of jobs assigned to a single machine, including a case where one subset is empty, is also a swap candidate when it is moved to another machine.

When a swap candidate is searched, in processing step 4, an evaluation value is calculated for the state obtained by exchanging the job subset of the candidate between the corresponding machines, and it is calculated from the evaluation value of the current state before the exchange. It is determined whether the improvement is achieved based on the determination conditions.

As a result, if it is determined that there is no improvement, the process returns to step 3, cancels the searched swap candidate, and searches for another swap candidate.

If it is improved, execute swapping of the swap candidate job sequence in processing step 5, update the state to the new state, and then return to processing step 2 to check whether the disconnected state is resolved. do.

In this way, based on appropriate evaluation values and judgment conditions,
By repeatedly exchanging swap candidates, it is possible to gradually improve the state and approach a solution.

Next, processing for a simplified example of the above-mentioned scheduling problem will be explained using specific data examples.

FIG. 2(a) is input data specifying a job to be delivered by truck, and FIG. 2(b) is a diagram explaining the input data using a diagram. Further, as shown in FIG. 2(C), it is assumed that three trucks (T1, T2, and T3), which are machines that execute this job, are arranged.

The constraint conditions in this case are as follows.

(a) Initial arrangement of tracks (see Figure 2 (C) above) (b)
The route and required time are Offices A-B, B-C, and C-DSD-A.

It takes 50 minutes between A and C.

(The required flight time allocated to each C1 track is 200
minutes or less.

(d) Operate all trucks.

(e) The truck deployment point will be Sales Office A, BSC.

Further, the value of the time exceeding 200 minutes of the total research required time of the flight assigned to each trunk is defined as the evaluation value, and the following condition is specified as the judgment condition.

That is, one of the current evaluation values of the swap candidates is e, and the other is 82.
and the new evaluation value after the exchange is ne.

and nez, ■ma(e++ez)>IIIax(ne+, net
) or ■ (e+=net) (ex>net) or ■
It is assumed that the condition is improved when (ez=net) (e, > ne,) is satisfied.

In this case, the exchange of a subset of jobs in swap candidates is performed by a loop swap (Figure 3 (a)) performed for a sequence of jobs that constitute a bus with the same departure and arrival points, and by a loop swap performed for a sequence of jobs that constitute a bus with the same departure and arrival points. There are two types of tail swaps that are performed on the sequence of jobs constituting , and buses with these conditions that have the same starting point are searched from two different job sets and used as swap candidates.

Under the above conditions, in order to perform scheduling, for example, first consider only the route order, and all flights are
The state assigned to 1 (FIG. 4(a)) is generated as the initial state.

In this way, it can be seen that the trunks T2 and T3, which have no job assignments, have dummy flight assignments from office B to office B and office office C to office C, respectively.

Therefore, if we take the loop swap shown in the shaded area in the figure as a swap candidate between Tl and 12, the current evaluation value is 50 x 12-200 for T1.
400 for T2, and -200 for T2, and if this swap is performed (in Figure 4), then 0 for T1.
, 200 for T2, it is determined that an improvement can be obtained according to the above-mentioned determination condition (■), and the exchange is performed to obtain the result shown in FIG.
Obtain the state b).

If we search for swap candidates between T2 and 13, where the interval between evaluation values is large, and pick up the tail swap in the shaded area in the state shown in Figure 4(b), we will execute the exchange of this swap candidate. The evaluation value of the result (Figure 4 (C)) is T
Since both T2 and T3 are O, it is judged that an improvement can be obtained according to the judgment condition
Condition C) and Condition B are obtained. Since this state satisfies all the constraints, it becomes a scheduling solution.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, in computer processing of a scheduling problem in which jobs are assigned to machines so as to satisfy required constraint conditions, necessary constraint conditions, etc. are used as input data, and appropriate job assignment is performed. Since the initial state is successively improved to obtain the solution state, the number of states that need to be tested is limited, making it possible to obtain a reasonable scheduling solution in a relatively short practical processing time, and also to solve the problem of constraint conditions. It has a remarkable industrial effect in that it is easy to adapt to changes.

[Brief explanation of the drawing]

FIG. 1 is a flowchart of processing showing the configuration of the present invention, FIG. 2 is an explanatory diagram of an example of a scheduling problem, FIG. 3 is an explanatory diagram of swap candidates, and FIG. 4 is an explanatory diagram of an example of scheduling processing. In the figure, kuguchi 〈口〈
Mouth and texture @ @
castle

Claims (1)

[Claims] In a process of inputting a condition representing a machine, a condition representing a job, and a constraint condition, and generating a state in which the job is assigned to the machine so as to satisfy the constraint condition, an evaluation of the state is performed. A value and a judgment condition are set, a state that does not necessarily satisfy all of the constraints is generated as an initial state (1), and when the state satisfies all of the constraints, the state is used as a scheduling solution. (2, 6), and if the state does not satisfy all of the constraints, the state is part of the job's assignment for the state and is interchangeable between different machines. The part is searched as a swap candidate (2, 3), and the judgment condition is that the evaluation value as a result of executing the swap of the swap candidate is improved from the evaluation value before executing the swap. If it is determined based on this, the exchange is executed to generate a new state (4, 5
), a scheduling processing method comprising: repeating the search for the swap candidate and the execution of the swap until the new state becomes the scheduling solution.