Improvement device, improvement method, and planning system

Abstract

The weight coefficients of the objective function are determined automatically. [Solution] The improvement device includes an input unit that accepts the specification of the priority of each term for an objective function that includes multiple terms, and an adjustment unit that calculates the amount of change for each term between the result of applying a predetermined solution to the objective function and the result of applying a comparison solution obtained by rearranging the solution to the objective function, and adjusts the objective function so that if the amount of change of the first priority term is less than the sum of the amounts of change of the terms with a lower priority than the first priority, the amount of change of the first priority term is greater than the sum of the amounts of change of the terms with a lower priority than the first priority.

Description

[Technical Field]

[0001] This disclosure relates to an improvement device, an improvement method, and a planning system. [Background technology]

[0002] For example, technologies are being considered for creating work plans that improve the efficiency of various tasks, such as picking operations in warehouses, delivery operations, or production operations in manufacturing. Patent Document 1 discloses a configuration in which the business objective function is defined as a weighted linear sum of a weight coefficient wj and an evaluation value Vji of a group of evaluation values, and the group of evaluation values ​​that gives the business objective function the maximum value is recognized as the group of evaluation values ​​with the highest evaluation. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-8955 [Overview of the project] [Problems that the invention aims to solve]

[0004] However, Patent Document 1 requires skilled engineers to manually adjust the weight coefficients of the objective function through trial and error, making it difficult to use.

[0005] Therefore, this disclosure aims to provide a technology that can automatically determine the weight coefficients of the objective function. [Means for solving the problem]

[0006] One aspect of the present disclosure provides an improvement device comprising: an input unit that accepts the specification of the priority of each term for an objective function including a plurality of terms; and an adjustment unit that calculates the amount of change for each term between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function, and adjusts the objective function such that if the amount of change of the term with first priority is less than the sum of the amounts of change of the terms with priority lower than the first priority, the amount of change of the term with first priority is greater than the sum of the amounts of change of the terms with priority lower than the first priority.

[0007] One aspect of this disclosure provides an improvement method that accepts the designation of the priority of each term in an objective function including multiple terms, calculates the change in each term between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function, and adjusts the objective function such that if the change in the term with first priority is less than the sum of the changes in the terms with lower priority than the first priority, the change in the term with first priority is greater than the sum of the changes in the terms with lower priority than the first priority.

[0008] One aspect of the present disclosure provides a planning device comprising: an input unit that accepts the specification of the priority of each term for an objective function including a plurality of terms; an adjustment unit that calculates the amount of change for each term between the result of applying a predetermined solution to the objective function and the result of applying a comparison solution obtained by rearranging the solution to the objective function, and adjusts the objective function such that if the amount of change of the term with first priority is less than the sum of the amounts of change of the terms with priority lower than the first priority, the amount of change of the term with first priority is greater than the sum of the amounts of change of the terms with priority lower than the first priority; an improvement determination unit that updates the solution to the comparison solution if the result of applying the comparison solution to the objective function is an improvement over the result of applying the solution to the objective function; and a planning generation unit that generates a plan based on the solution obtained after repeating the processing by the improvement determination unit and the processing by the adjustment unit multiple times.

[0009] These comprehensive or specific embodiments may be implemented as systems, devices, methods, integrated circuits, computer programs, or recording media, or as any combination of systems, devices, methods, integrated circuits, computer programs, and recording media. [Effects of the Invention]

[0010] According to this disclosure, the weight coefficients of the objective function can be automatically determined. [Brief explanation of the drawing]

[0011] [Figure 1] Block diagram showing a first example of a planning device according to Embodiment 1. [Figure 2] Flowchart showing the processing in the first example of the planning device according to Embodiment 1. [Figure 3] This figure illustrates a specific example of the first iteration of the planning loop in Calculation Example 1 according to Embodiment 1. Figure 3(a) is a graph showing the change before coefficient adjustment, and Figure 3(b) is a graph showing the change after coefficient adjustment. [Figure 4] This figure illustrates a specific example of the second iteration of the planning loop in Calculation Example 1 according to Embodiment 1. Figure 4(a) is a graph showing the change before coefficient adjustment, and Figure 4(b) is a graph showing the change after coefficient adjustment. [Figure 5] Figure 5(a) is a graph showing the change before coefficient adjustment, and Figure 5(b) is a graph showing the change after coefficient adjustment. [Figure 6] Block diagram showing a second example of the planning device according to Embodiment 1. [Figure 7] Flowchart showing the processing in a second example of the planning device according to Embodiment 1. [Figure 8] Figure 8(a) is a graph showing the change before normalization, Figure 8(b) is a graph showing the change after normalization and before coefficient adjustment, and Figure 8(c) is a graph showing the change after normalization and after coefficient adjustment. [Figure 9]A diagram illustrating an example of the configuration of solution x according to Embodiment 1. [Figure 10] A diagram showing an example of the work plan generation result according to Embodiment 1. [Figure 11] A graph showing the changes in each term when the planning loop is repeated with automatic coefficient adjustment according to Embodiment 1. [Figure 12] A graph showing the changes in each term when a skilled technician manually adjusts the coefficients and repeats the planning loop. [Figure 13] A diagram showing a first example of the input screen according to Embodiment 1. [Figure 14] This figure shows a second example of the input screen according to Embodiment 1. [Figure 15] This figure shows a third example of the input screen according to Embodiment 1. [Figure 16] This figure shows a fourth example of the input screen according to Embodiment 1. [Figure 17] This figure shows a fifth example of the input screen according to Embodiment 1. [Figure 18] Block diagram showing the hardware configuration of a computer that implements the functions of the planning device described herein using a computer program. [Modes for carrying out the invention]

[0012] Embodiments of the present disclosure will be described in detail below, with appropriate reference to the drawings. However, descriptions that are unnecessarily detailed may be omitted. For example, detailed descriptions of already well-known matters and redundant descriptions of substantially identical configurations may be omitted. This is to avoid the following description becoming unnecessarily verbose and to facilitate understanding for those skilled in the art. The accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure and are not intended to limit the subject matter of the claims. The functions of one configuration shown in this embodiment may be realized by two or more physical configurations, or the functions of two or more configurations may be realized by, for example, one physical configuration.

[0013] (Embodiment 1) <First Example of Planning Device> FIG. 1 is a block diagram showing a first example of a planning device according to Embodiment 1.

[0014] The planning device 10 is a device that formulates a work plan, and includes, as functions, an initial plan unit 11, a plan rearrangement unit 12, an improvement determination unit 13, and a plan generation unit 18. These functions may be realized by a processor 1001 (see FIG. 18) included in the planning device 10 cooperating with a memory 1002 (see FIG. 18) and the like to execute a program.

[0015] The initial plan unit 11 acquires a work list and creates an initial best solution x best from the work list.

[0016] The plan rearrangement unit 12 exchanges the assignments between, for example, two randomly selected workers or randomly rearranges the order within one worker for the initial best solution x best or the best solution x best improved by the improvement determination unit 13, and creates a comparison solution x best based on the best solution x tmp by, for example, randomly selecting two workers and exchanging their assignments or randomly rearranging the order within one worker.

[0017] The improvement determination unit 13 determines whether the comparison solution x tmp created by the plan rearrangement unit 12 improves the objective function. Also, in the iterative process, the improvement determination unit 13 inputs the best solution x best to the plan rearrangement unit 12. Further, the improvement determination unit 13 outputs the final best solution x best after the iterative process. Note that the improvement determination unit 13 may be configured as an improvement device.

[0018] The plan generation unit 18 generates a work plan from the best solution x best output from the improvement determination unit 13. Note that the plan generation unit 18 may display this work plan on a display device 1005 (see FIG. 18).

[0019] The improvement determination unit 13 includes an input unit 14, an objective function unit 15, and an automatic coefficient adjustment unit 16.

[0020] The input unit 14 is an objective function f that includes multiple terms. all For each term f (x) i The priority of (x) is specified. The input unit 14 may display an input screen (see Figures 13 to 17) for specifying this priority on the display device 1005. In this embodiment, the objective function is defined as follows: i is an integer of 1 or more. Also, w i This is the weighting coefficient. f all (x) = Σ i w i f i (x)

[0021] The objective function section 15 consists of coefficients w that have been automatically adjusted by the coefficient automatic adjustment section 16. i Using the objective function f all The calculations for (x) are performed. Furthermore, the objective function 15 is defined as the change in each term |Δw|. i f i Calculate (x)|.

[0022] The automatic coefficient adjustment unit 16 adjusts the weight coefficient w i The automatic adjustment of the weight coefficients will be described later. The automatic coefficient adjustment unit 16 also adjusts the objective function f all If there is a bias in how easily the value changes among the terms of (x) when the solution x changes, the change in each term is also normalized. If there is a bias in how easily the value changes, the weight coefficient w of the terms that change easily is used. i Adjustments focusing on this become easier, and the weight coefficients w of other terms i Adjustments focusing on this become more difficult. However, since the solution x is applied to multiple terms that make up the objective function, maintaining a balance of the changes in multiple terms makes it easier to obtain the optimal solution x. Therefore, the coefficient automatic adjustment unit 16 maintains this balance by performing normalization, while adjusting the weight coefficient w iAdjust the values. An example of a case where a term is more prone to value changes than other terms is when a term with different units is included. Also, even if the units of the terms are the same, if it is recognized that some terms are more prone to value changes, the user may be allowed to specify which terms to normalize.

[0023] Figure 2 is a flowchart showing the processing in a first example of the planning device 10 according to Embodiment 1.

[0024] The coefficient automatic adjustment unit 16 adjusts the coefficient w i Initialize (S101). Note that the coefficient w i The value of is ultimately automatically adjusted, so coefficient w i The initial value can be any value.

[0025] The input unit 14 is for each term f i The system accepts input for the priority and stores the input priority in memory 1002 (see Figure 18) (S102). For example, the input unit 14 displays an input screen as shown in Figures 13 to 17 on the display device 1005 and accepts priority input from the user.

[0026] The initial planning unit 11 is the initial best solution x best Create (S103).

[0027] The planning device 10 repeats the following planning loop (S104~S113) a predetermined number of times (S104).

[0028] The reorganization unit 12 determines the best solution x best Rearrange the elements to get the comparison solution x tmp Create (S105).

[0029] The planning device 10 repeats the coefficient loop (S106~S109) (S106) for (number of priorities - 1) times, in order from the highest priority term to the lowest priority term.

[0030] The objective function part 15 is the best solution x best Change |Δw based on if i Calculate (x)| (S107). This change is the comparison solution x tmp The value of each term in the best solution x best This value indicates how much the value of each corresponding term has changed, and is expressed as an absolute value. That is, the best solution x best Compared solution x tmp Between these two points, we do not consider whether the value of each term has increased or decreased, but only the magnitude of the change (amount of change). By using this amount of change, we can find the best solution x best Compared solution x tmp In the change between these two points, we can evaluate which terms had the greatest impact.

[0031] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S107. i f i Determine whether (x)| satisfies the satisfaction criterion (S108). The details of the satisfaction criterion will be described later.

[0032] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S107. i f i If (x)| does not satisfy the satisfaction criterion (S108: NO), the coefficient update formula is applied to update the coefficient (S109), and the process proceeds to step S110. Details of the coefficient update formula will be described later.

[0033] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S107. i f i If (x)| satisfies the satisfaction criterion (S108: YES), proceed to step S110.

[0034] The planning device 10 repeats the coefficient loop (S106~S109) (S110) a number of times equal to (number of priorities - 1), and then proceeds to the next step S111.

[0035] The improvement determination unit 13 determines the best solution x best Update conditions "f all (x best )>f all(x tmp Determine whether the condition ")" is met (S111).

[0036] Best solution x best If the update conditions are met (S111:YES), the improvement determination unit 13 determines x best to x tmp The data is updated (S112), and the process proceeds to step S113.

[0037] Best solution x best If the update conditions are not met (S111: NO), the improvement determination unit 13 proceeds to step S113.

[0038] The planning device 10 repeats the planning loop (S104~S113) a set number of times (S113), and then proceeds to the next step S114.

[0039] The improvement determination unit 13 determines the best solution x best The output is then generated, and the plan generation unit 18 outputs the best solution x best A work plan is generated from this (S114). Then, this process ends.

[0040] <Calculation example 1> As Calculation Example 1, we will explain a specific example of the process shown in Figure 2. Note that Calculation Example 1 explains a specific example where there are no terms with equal priority and no normalization is performed on the change amounts of each term.

[0041] Figure 3 is a diagram illustrating a specific example of the first iteration of the planning loop in Calculation Example 1 according to Embodiment 1. Figure 3(a) is a graph showing the change before coefficient adjustment, and Figure 3(b) is a graph showing the change after coefficient adjustment.

[0042] First, let's explain each variable in Calculation Example 1.

[0043] x is the solution to the work plan, and includes, for example, who to assign which task to, at what time, and the order of the tasks. best This shows the best solution, x tmpThis shows the comparative solution.

[0044] w1, w2, and w3 are weight coefficients. w1, w2, and w3 may also be simply called coefficients. In step S101, the initial weight coefficients are set to w1=1.0, w2=1.0, and w3=1.0.

[0045] f1(x), f2(x), and f3(x) are evaluation values ​​or penalty values. Note that f1(x), f2(x), and f3(x) may also be called terms.

[0046] The objective function is f all (x) = w1f1(x) + w2f2(x) + w3f3(x). This objective function f all Based on (x), the revised plan (comparison solution x) tmp The quality of the objective function f is evaluated. all The smaller the value of (x), the better the plan (the more efficient the plan), and the objective function f all A larger value of (x) indicates a worse plan (inefficient plan). However, the content of this disclosure is limited to the objective function f all The larger the value of (x), the better the plan (the more efficient the plan), and the objective function f all The smaller the value of (x), the more applicable this can be, even for bad (inefficient) plans.

[0047] In Calculation Example 1, the priority order is set as f1(x) term > f2(x) term > f3(x) term. In this embodiment, this priority order is expressed as [1,2,3]. In other words, in Calculation Example 1, f1(x) is minimized with the highest priority.

[0048] c is a constant, and in this embodiment, c = 0.9.

[0049] <Calculation Example 1: Satisfaction Check Before Coefficient Update (First Loop)> In the first iteration of the planning loop, the objective function part 15, with the coefficients before the update, represents the current best solution x best Changes in each term |Δw i f i Calculate (x)|.

[0050] As shown in the graph in Figure 3(a), the changes in each term at this point in time are as follows: Δw1f1(x)=+1 Δw2f2(x)=-4 Δw3f3(x)=-6

[0051] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1 |Δw2f2(x)|=4 |Δw3f3(x)|=6

[0052] The satisfaction criterion is defined as "change in high-priority terms > sum of changes in low-priority terms". If this satisfaction criterion is satisfied, the objective function f all In (x), the influence of the high-priority terms is greater than the sum of the influences of the low-priority terms. Alternatively, the satisfaction criterion can be defined as "the change in a high-priority term is greater than the change in any of the low-priority terms." However, in this case, even if the change in each low-priority term is smaller than the change in a high-priority term, the sum of the changes in the low-priority terms may exceed the change in the high-priority terms. In such a situation, the objective function f all In (x), the influence of the high-priority term is smaller than the influence of all the other low-priority terms combined, so it cannot be said that the high-priority term has a greater influence than the low-priority term (set of terms). Therefore, in this embodiment, the satisfaction criterion used is "change in high-priority term > sum of changes in low-priority term".

[0053] In calculation example 1, the sufficiency criterion is: |Σ i Δw i f i (x)|>Σ j |Δw j f j (x) (i,j)=([1],[2,3]),([2],[3]) It is defined as follows. In other words, if the inequality holds when i=[1] and j=[2,3] are substituted, and when i=[2] and j=[3] are substituted, then the criterion for satisfaction is determined to be satisfied; otherwise, the criterion for satisfaction is determined to be unsatisfied.

[0054] Here, when i=[1] and j=[2,3], the left side represents the change in the f1(x) term, and the right side represents the sum of the changes in the f2(x) term and the f3(x) term. In other words, the satisfaction criterion in this case is defined as "change in the first priority term > sum of the changes in the terms with lower priority than the first priority". Note that if there is one term with lower priority than the first priority, the satisfaction criterion may be defined as "change in the first priority term > second priority term with lower priority than the first priority".

[0055] Note that the right-hand side of the satisfaction criterion is defined as "the sum of the changes in the low-priority terms," ​​assuming there are multiple low-priority terms. However, if there is only one low-priority term, the change in that low-priority term is treated as "the sum of the changes in the low-priority terms."

[0056] Furthermore, when i=[2] and j=[3], the left side represents the change in the f2(x) term, and the right side represents the change in the f3(x) term. In other words, the satisfaction criterion in this case means that "the change in the second-highest priority term > the change in the third-highest priority term or lower."

[0057] Therefore, if the satisfaction criterion is satisfied, the relationship "change in the high-priority term > sum of the changes in the low-priority term" holds for the f1(x), f2(x), and f3(x) terms.

[0058] According to the graph in Figure 3(a), the change in the high-priority term |Δw1f1(x)| is 1, and the sum of the changes in the low-priority terms |Δw2f2(x)|+|Δw3f3(x)| is 10. Therefore, |Δw1f1(x)|<|Δw2f2(x)|+|Δw3f3(x)| (i.e., 1<10), and thus the satisfaction criterion is not satisfied.

[0059] Therefore, the automatic coefficient adjustment unit 16 determines step S108 to be NO, applies the coefficient update formula in step S109, and performs the coefficient update (coefficient adjustment) described below.

[0060] Furthermore, when we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=1-4-6<0 As a result, even though the high-priority f1(x) term worsened (i.e., increased), the objective function f all (x) is improving (i.e., decreasing). This means that the higher-priority terms are not having an appropriate impact on the objective function.

[0061] <Calculation Example 1: Coefficient Update (1st Cycle)> In step S109, the automatic coefficient adjustment unit 16 automatically adjusts the coefficients so that "the change in high priority > the sum of the changes in low priority".

[0062] In the first iteration of the coefficient loop, the change in the high-priority term is |Δw1f1(x)|, and the sum of the changes in the low-priority terms is (|Δw2f2(x)|+|Δw3f3(x)|).

[0063] And, as mentioned above, Since |Δw1f1(x)|≦(|Δw2f2(x)|+|Δw3f3(x)|) (i.e., 1<(4+6)), the satisfaction criterion is not met.

[0064] moreover, |Δw1f1(x)|≠0 and |Δw2f2(x)|+|Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficients w2 and w3 in the following calculation. w2←w2×(|Δw1f1(x)| / (|Δw2f2(x)|+|Δw3f3(x)|))×c=1×(1 / (4+6))×0.9=0.09 w3←w3×(|Δw1f1(x)| / (|Δw2f2(x)|+|Δw3f3(x)|))×c=1×(1 / (4+6))×0.9=0.09

[0065] In the second iteration of the coefficient loop, the change in the high-priority term is |Δw2f2(x)|, and the sum of the changes in the low-priority terms is |Δw3f3(x)|.

[0066] Then, in the second iteration of the coefficient loop, applying the w2=0.09 and w3=0.09 calculated above, Since |Δw2f2(x)|≦|Δw3f3(x)| (i.e., (0.09×4)<(0.09×6)), the satisfaction criterion is not met.

[0067] moreover, |Δw2f2(x)|≠0 and |Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficient w3 in the following calculation. w3←w3×(|Δw2f2(x)| / |Δw3f3(x)|)×c=0.09×(0.36 / 0.54)×0.9=0.054

[0068] Therefore, the coefficient is, w1=1.0, w2=0.09, w3=0.054 It will be updated.

[0069] <Calculation Example 1: Satisfaction Check After Coefficient Update (First Loop)> The objective function part 15 is the adjusted coefficient, representing the best solution x at this point. best Changes in each term |Δw i f i Calculate (x)|.

[0070] As shown in the graph in Figure 3(b), the changes in each term when using the adjusted coefficients w1=1.0, w2=0.09, and w3=0.054 are as follows. Δw1f1(x)=+1.0 Δw2f2(x)=-0.36 Δw3f3(x) = -0.324

[0071] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1.0 |Δw2f2(x)|=0.36 |Δw3f3(x)|=0.324

[0072] This is a satisfiesment criterion. |Δw1f1(x)|>(|Δw2f2(x)|+|Δw3f3(x)|) This satisfies the condition 1.0 > (0.36 + 0.324).

[0073] When we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=1.0-0.36-0.324=0.316>0 As a result, the high-priority f1(x) term worsened (i.e., increased), and the objective function f all This has also worsened (i.e., increased). In other words, the high-priority terms are having an appropriate influence on the objective function. Therefore, we can conclude that we were able to adjust the weight coefficients of the objective function so that we can judge the improvement of the solution according to its priority.

[0074] Furthermore, since the purpose is to reflect priority in the weight coefficients, the weight coefficients are updated regardless of whether the objective function improves or worsens. Also, in the above case, since the objective function worsened, the best solution is not updated. In other words, step S111 is NO, and the search for the best solution continues with the objective function after the coefficient update.

[0075] Figure 4 is a diagram illustrating a specific example of the second iteration of the planning loop in Calculation Example 1 according to Embodiment 1. Figure 4(a) is a graph showing the change before coefficient adjustment, and Figure 4(b) is a graph showing the change after coefficient adjustment.

[0076] <Calculation Example 1: Satisfaction Check Before Coefficient Update (2nd Round)> In the second iteration of the planned loop, the current coefficients are w1=1.0, w2=0.09, and w3=0.054, as calculated above.

[0077] As shown in the graph in Figure 4(a), the changes in each term at this point in time are as follows: Δw1f1(x)=-1 Δw2f2(x)=-0.36 Δw3f3(x)=+1.5

[0078] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1 |Δw2f2(x)|=0.36 |Δw3f3(x)|=1.5

[0079] Furthermore, from the above formula, f2(x) = -0.36 / Δw2 = -0.36 / 0.02 = 4 f3(x)=+1.5 / Δw3=+1.5 / 0.054≒27.8 This is the result.

[0080] If the relationship "change in high-priority terms > sum of changes in low-priority terms" is satisfied, then the evaluation of the solution improvement will take priority into account. However, with the current coefficients, |Δw1f1(x)|>(|Δw2f2(x)|+|Δw3f3(x)|) (i.e., 1>(0.36+1.5)), so the satisfaction criterion is not satisfied.

[0081] When we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=-1-0.36+1.5>0 Thus, even though the high-priority f1(x) term was improved (i.e., reduced), the objective function f all (x) is worsening (i.e., increasing). This means that the high-priority terms are not having an appropriate impact on the objective function.

[0082] <Calculation Example 1: Coefficient Update (2nd Iteration)> In step S109, the coefficient automatic adjustment unit 16 automatically adjusts the coefficients so that "the amount of change in the high-priority term > the sum of the amounts of change in the low-priority term".

[0083] At present, the coefficient w1 = 1.0.

[0084] In the first iteration of the coefficient loop, Since |Δw1f1(x)|≦(|Δw2f2(x)|+|Δw3f3(x)|) (i.e., 1<(0.36+1.5)), the satisfaction criterion is not met.

[0085] moreover, |Δw1f1(x)|≠0 and |Δw2f2(x)|+|Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficients w2 and w3 in the following calculation. w2←w2×(|Δw1f1(x)| / (|Δw2f2(x)|+|Δw3f3(x)|))×c=0.09×(1 / (0.36+1.5))×0.9=0.044 w3←w3×(|Δw1f1(x)| / (|Δw2f2(x)|+|Δw3f3(x)|))×c=0.054×(1 / (0.36+1.5))×0.9=0.026

[0086] In the second iteration of the coefficient loop, the change in the high-priority term is |Δw2f2(x)|, and the sum of the changes in the low-priority terms is |Δw3f3(x)|.

[0087] Then, in the second iteration of the coefficient loop, applying the w2=0.044 and w3=0.026 calculated above, Since |Δw2f2(x)|≦|Δw3f3(x)| (i.e., (0.044×4)<(0.026×27.8)), the satisfaction criterion is not met.

[0088] moreover, |Δw2f2(x)|≠0 and |Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficient w3 in the following calculation. w3←w3×(|Δw2f2(x)| / |Δw3f3(x)|)×c=0.026×(0.176 / 0.72)×0.9=0.0057

[0089] Therefore, the coefficients are w1 = 1.0, w2 = 0.044, w3 = 0.0057 updated to.

[0090] <Calculation Example 1: Satisfaction Judgment Formula after Coefficient Update (Second Round)> The objective function part 15 calculates the change amount |Δw best f i f i (x)| of each term at the current best solution x

[0091] As shown in Figure 4(b), assuming that the changes in each term when using the adjusted coefficients w1 = 1.0, w2 = 0.044, and w3 = 0.0057 are as follows. Δw1f1(x) = -1 Δw2f2(x) = -0.176 Δw3f3(x) = +0.158

[0092] Therefore, the change amount (absolute value) of each term is as follows. |Δw1f1(x)| = 1 |Δw2f2(x)| = 0.176 |Δw3f3(x)| = 0.158

[0093] This satisfies the satisfaction judgment formula |Δw1f1(x)| > |Δw2f2(x)| + |Δw3f3(x)| That is, it satisfies 1.0 > (0.176 + 0.158).

[0094] When examining the objective function at this time, Δf all (x) = Δw1f1(x) + Δw2f2(x) + Δw3f3(x) = -1 - 0.176 + 0.158 = -1.018 < 0 As a result, due to the improvement (i.e., decrease) of the higher-priority f1(x) and f2(x) terms, the objective function Δf all (x) has also improved (i.e., decreased). That is, the higher-priority terms have an appropriate influence on the objective function.

[0095] Therefore, the weight coefficient of the objective function can be adjusted so that the improvement determination of the solution can be made according to the priority. Thus, upon receiving the improvement of the solution, the improvement determination unit 13 determines YES in step S111, and in step S112, the best solution x best is updated with x tmp .

[0096] <Calculation Example 2> As Calculation Example 2, a specific example of the process shown in FIG. 2 will be described. Note that in Calculation Example 2, a case where there are equal priorities in the priorities of each term and normalization is not performed on the change amount of each term will be described.

[0097] FIG. 5 is a diagram for explaining a specific example in Calculation Example 2 according to Embodiment 1. FIG. 5(a) is a graph showing the change amount before coefficient adjustment, and FIG. 5(b) is a graph showing the change amount after coefficient adjustment.

[0098] First, each variable in Calculation Example 2 will be described. However, descriptions of variables similar to those in Calculation Example 1 will be omitted.

[0099] In Calculation Example 2, the initial values of the weight coefficients are set to w1 = 1.0, w2 = 1.0, and w3 = 1.0.

[0100] In Calculation Example 2, the priorities are set such that the f1(x) term = the f2(x) term > the f3(x) term. In this embodiment, this priority is expressed as [[1, 2], 3]. That is, it is assumed that the priorities of the f1(x) term and the f2(x) term are equal. In this case, w1f1(x) + w2f2(x) is regarded as one term and is minimized with the highest priority.

[0101] Note that other variables are the same as those in Calculation Example 1.

[0102] <Calculation Example 2: Satisfaction Judgment before Coefficient Update> The objective function part calculates the change amount |Δw best of each term at the current best solution x i f i (x)| with the coefficients before update.

[0103] As shown in the graph in Figure 5(a), the changes in each term at this point in time are as follows: Δw1f1(x)=-1 Δw2f2(x)=+4 Δw3f3(x)=-6

[0104] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1 |Δw2f2(x)|=4 |Δw3f3(x)|=6

[0105] The satisfaction criterion is the same as in formula 1: "Change in high-priority terms > Sum of changes in low-priority terms". Therefore, the satisfaction criterion is: |Σ i Δw i f i (x)|>Σ j |Δw j f j (x) (i,j)=([1,2],[3]) This is defined as follows. Here, the change in the high-priority term is the sum of equal-priority terms (|Δw1f1(x)+Δw2f2(x)|).

[0106] In other words, when there are multiple terms with the same priority level, the satisfaction criterion can be applied by treating those terms with the same priority level as a single term. Note that in calculation example 2, there are only two levels of priority, so a comparison of the changes between terms with priority levels 2 and below, such as (i,j)=([2],[3]) in calculation example 1, is not performed.

[0107] Using the values ​​shown in the graph of Figure 4(a), the change in the high-priority term (|Δw1f1(x)+Δw2f2(x)|) is 3 (=|-1+4|), and the sum of the changes in the low-priority terms |Δw3f3(x)| is 6. Therefore, (|Δw1f1(x)+Δw2f2(x)|)<|Δw3f3(x)| (i.e., 3<6), and the satisfaction criterion is not satisfied.

[0108] Therefore, the automatic coefficient adjustment unit 16 determines step S108 to be NO, applies the coefficient update formula in step S109, and performs the coefficient update (coefficient adjustment) described below.

[0109] Furthermore, when we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=(-1+4)-6<0 As a result, even though the high-priority (w1f1(x)+w2f2(x)) worsened (increased), the objective function f all This indicates an improvement (i.e., a decrease). In other words, the high-priority terms are not having an appropriate impact on the objective function.

[0110] <Calculation Example 2: Coefficient Update> In step S109, the coefficient automatic adjustment unit 16 automatically adjusts the coefficients so that "the amount of change in the high-priority term > the sum of the amounts of change in the low-priority term".

[0111] The change in the high-priority terms is |Δw1f1(x)+Δw2f2(x)|, and the sum of the changes in the low-priority terms is |Δw3f3(x)|.

[0112] And, as mentioned above, Since |Δw1f1(x)+Δw2f2(x)|<|Δw3f3(x)| (i.e., (-1+4)<6), the satisfaction criterion is not met.

[0113] moreover, |Δw1f1(x)+Δw2f2(x)|≠0 and |Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficient w3 in the following calculation. w3←w3×((|Δw1f1(x)+Δw2f2(x)|) / |Δw3f3(x)|)×c=1×(3 / 6)×0.9=0.45

[0114] Therefore, the coefficient is, w1=1.0, w2=1.0, w3=0.45 It will be updated.

[0115] <Calculation Example 2: Satisfaction Check After Coefficient Update> The objective function part 15 is the adjusted coefficient, representing the best solution x at this point. best Changes in each term |Δw i f i Calculate (x)|.

[0116] As shown in the graph in Figure 5(b), the changes in each term when using the adjusted coefficients w1=1.0, w2=1.0, and w3=0.45 are as follows. Δw1f1(x)=-1 Δw2f2(x)=+4 Δw3f3(x) = -2.7

[0117] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1 |Δw2f2(x)|=4 |Δw3f3(x)|=2.7

[0118] This satisfies the criterion (|Δw1f1(x)+Δw2f2(x)|>|Δw3f3(x)|). In other words, (-1+4)>2.7.

[0119] When we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=-1+4-2.7=0.3>0 As a result, the high-priority (Δw1f1(x)+Δw2f2(x)) term deteriorates (i.e., increases), and the objective function Δf all (x) has also worsened (i.e., increased). This means that the higher-priority terms are having an appropriate effect on the objective function. Therefore, we can conclude that we were able to adjust the weight coefficients of the objective function so that we can judge the improvement of the solution according to its priority.

[0120] <Second example of a planning device> The second example of the planning device 10 describes the case where normalization is performed on terms with different units.

[0121] Figure 6 is a block diagram showing a second example of the planning device 10 according to Embodiment 1.

[0122] The second example of the planning device 10 includes a normalization unit 17 in addition to the functions of the first example of the planning device 10 shown in Figure 1, in addition to the coefficient automatic adjustment unit 16.

[0123] The normalization unit 17 normalizes at least one of the multiple terms included in the objective function by adjusting its coefficient. For example, if the multiple terms included in the objective function include terms with different units, the normalization unit 17 normalizes that term.

[0124] Figure 7 is a flowchart showing the processing in a second example of the planning device 10 according to Embodiment 1.

[0125] The coefficient automatic adjustment unit 16 adjusts the coefficient w i Initialize (S201).

[0126] The input unit 14 is for each term f i The system accepts input for the priority level and stores the input priority level in memory 1002 (see Figure 18) (S202). For example, the input unit 14 displays an input screen as shown in Figures 13 to 17 on the display device 1005 and accepts priority level input from the user.

[0127] The input unit 14 receives the input of a normalization flag and stores the input normalization flag in the memory 1002 (S203). The normalization flag is a flag that indicates which term should be normalized. For example, the input unit 14 displays an input screen on the display device 1005 as shown in Figures 14, 16, and 17 and receives the input of a normalization flag from the user.

[0128] The initial planning unit 11 is the initial best solution x best Create (S204).

[0129] The planning device 10 repeats the following planning loop (S205~S215) a predetermined number of times (S205).

[0130] The reorganization unit 12 determines the best solution x best Rearrange the elements to get the comparison solution x tmp Create (S206).

[0131] The planning device 10 repeats the coefficient loop (S207~S212) (S207) for (number of priorities - 1) times, in order from the highest priority term to the lowest priority term.

[0132] The objective function part 15 is the best solution x best Change |Δw based on i f i Calculate (x)| (S208).

[0133] If the amount of change calculated in step S208 is the largest ever, the normalization unit 17 normalizes and updates the coefficient using that largest amount of change (S209). If the amount of change calculated in step S208 is not the largest ever, the normalization unit 17 does not need to normalize the coefficient.

[0134] The objective function part 15 uses the coefficients normalized in step S209 to find the best solution x best Change |Δw based on i f i (x)| is calculated (S210). If the coefficients were not normalized in step S209, the objective function 15 may omit step S210 and use the change calculated in step S208.

[0135] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S210. i f i Determine whether (x)| satisfies the satisfaction criterion (S211).

[0136] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S210. i f iIf (x)| does not satisfy the satisfaction criterion (S211: NO), the coefficient update formula is applied to update the coefficient (S212), and the process proceeds to step S212.

[0137] The coefficient automatic adjustment unit 16 adjusts the change amount |Δw| calculated in step S210. i f i If (x)| satisfies the satisfaction criterion (S211: YES), proceed to step S212.

[0138] The planning device 10 repeats the coefficient loop (S207~S212) (S212) a number of times equal to (number of priorities - 1), and then proceeds to step S213.

[0139] The improvement determination unit 13 determines the best solution x best Update conditions "f all (x best )>f all (x tmp Determine whether the condition ")" is met (S213).

[0140] Best solution x best If the update conditions are met (S213:YES), the improvement determination unit 13 determines x best to x tmp The data is updated (S214), and the process proceeds to step S215.

[0141] Best solution x best If the update conditions are not met (S213: NO), the improvement determination unit 13 proceeds to step S215.

[0142] The planning device 10 repeats the planning loop (S205~S215) a set number of times (S215), and then proceeds to the next step S216.

[0143] The improvement determination unit 13 determines the best solution x best The output is then generated, and the plan generation unit 18 outputs the best solution x best A work plan is generated from this (S216). Then, this process ends.

[0144] <Calculation example 3> As calculation example 3, we will explain a specific example of the process shown in Figure 7. In calculation example 3, we will explain the case where each term has equal priority and normalization is performed for different units of each term.

[0145] Figure 8 is a diagram illustrating a specific example in Calculation Example 3 according to Embodiment 1. Figure 8(a) is a graph showing the change before normalization, Figure 8(b) is a graph showing the change after normalization and before coefficient adjustment, and Figure 8(c) is a graph showing the change after normalization and after coefficient adjustment.

[0146] First, let's explain each variable in Calculation Example 3. However, we will omit explanations for variables that are the same as those in Calculation Example 1.

[0147] In calculation example 3, the initial values ​​of the weight coefficients are set to w1=0.5, w2=1.0, and w3=1.0.

[0148] In calculation example 3, the priority is set as f1(x) term = f2(x) term > f3(x) term. In this embodiment, this priority is expressed as [[1,2],3]. That is, the priority of the f1(x) term and the f2(x) term is assumed to be equal. In this case, w1f1(x) + w2f2(x) is treated as a single term and minimized with the highest priority.

[0149] Note that the other variables are the same as in Calculation Example 1.

[0150] <Calculation example 3: Normalization> The objective function part 15, in step S208, uses the coefficients before the update to obtain the best solution x at this point. best Changes in each term |Δw i f i Calculate (x)|.

[0151] As shown in the graph in Figure 8(a), the changes in each term at this point in time are as follows: Δw1f1(x)=-1.2 Δw2f2(x)=+4 Δw3f3(x)=-6

[0152] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1.2 |Δw2f2(x)|=4 |Δw3f3(x)|=6

[0153] The normalization unit 17 represents the largest change in history at the present time |Δw i f i (x) max lol i The effect of normalization is included in the coefficient by dividing by . For example, the maximum change so far |Δw1f1(x)| max =2 |Δw2f2(x)| max =1 Let's assume that was the case.

[0154] The amount of change this time is |Δw1f1(x)|=1.2 |Δw2f2(x)|=4 Therefore, with respect to |Δw2f2(x)|, the maximum change has been updated from 1 to 4, so we normalize w2 by dividing it by the maximum value and update it as follows. w2←w2 / |Δw2f2(x)| max w2 ← 1 / 4 = 0.25

[0155] Furthermore, the update of the coefficients for normalization in step S209 is performed only when the amount of change calculated in step S208 is the largest ever recorded. If the amount of change calculated in step S208 is less than or equal to the largest ever recorded amount, it does not need to be performed.

[0156] <Calculation Example 3: Satisfaction Check Before Coefficient Update> The objective function part 15 is the coefficient after normalization, representing the best solution x at this point. best Changes in each term |Δw i f i Calculate (x)|.

[0157] As shown in the graph in Figure 8(b), the changes in each term at this point in time are as follows: Δw1f1(x)=-1.2 Δw2f2(x)=+1 Δw3f3(x)=-6

[0158] Therefore, the change (absolute value) of each term is as follows: |Δw1f1(x)|=1.2 |Δw2f2(x)|=1 |Δw3f3(x)|=6

[0159] The satisfaction criterion is the same as in formula 1: "Change in high-priority terms > Sum of changes in low-priority terms". Therefore, the satisfaction criterion is: |ΣiΔwifi(x)|>Σj|Δwjfj(x)| (i,j)=([1,2],[3]) This is defined as follows. Here, the change in the high-priority term is the sum of the terms of equal priority (|Δw1f1(x)+Δw2f2(x)|).

[0160] Using the values ​​shown in the graph in Figure 8(b), the change in the high-priority term (|Δw1f1(x)+Δw2f2(x)|) is 0.2 (=|-1.2+1|), and the sum of the changes in the low-priority terms |Δw3f3(x)| is 6. Therefore, (|Δw1f1(x)+Δw2f2(x)|) < |Δw3f3(x)| (i.e., 0.2 < 6), and thus the satisfaction criterion is not satisfied.

[0161] Therefore, the automatic coefficient adjustment unit 16 determines step S211 to be NO, applies the coefficient update formula in step S212, and performs the coefficient update (coefficient adjustment) described below.

[0162] Furthermore, when we examine the change in the objective function at this time, Δf all (x)=Δw1f1(x)+Δw2f2(x)+Δw3f3(x)=0.2-6<0 As a result, even though the high-priority (w1f1(x)+w2f2(x)) worsened (increased), the objective function f allThis indicates an improvement (i.e., a decrease). In other words, the high-priority terms are not having an appropriate impact on the objective function.

[0163] <Calculation Example 3: Coefficient Update> In step S212, the coefficient automatic adjustment unit 16 automatically adjusts the coefficients so that "the change in the high-priority term > the sum of the changes in the low-priority term".

[0164] The change in the high-priority terms is |Δw1f1(x)+Δw2f2(x)|, and the sum of the changes in the low-priority terms is |Δw3f3(x)|. And, as mentioned above, Since |Δw1f1(x)+Δw2f2(x)|<|Δw3f3(x)| (i.e., 0.2<6), the criterion for satisfaction is not met.

[0165] moreover, |Δw1f1(x)+Δw2f2(x)|≠0 and |Δw3f3(x)|≠0 Therefore, the coefficient automatic adjustment unit 16 updates the coefficient w3 using the following formula. w3←w3×|w1f1(x)+Δw2f2(x)| / |Δw3f3(x)|×c=1×(0.2 / 6)×0.9=0.03

[0166] Therefore, the coefficient is, w1=1.0, w2=0.25, w3=0.03 It will be updated.

[0167] <Calculation Example 3: Priority Satisfaction Determination After Coefficient Update> The objective function part 15 is the adjusted coefficient, representing the best solution x at this point. best Changes in each term |Δw i f i Calculate (x)|.

[0168] As shown in Figure 8(c), the adjusted coefficients are w1=1.0, w2=0.25, and w3=0.03. The changes in each term when using are as follows: Δw1f1(x)=-1.2 Δw2f2(x)= +1 Δw3f3(x)= -0.18

[0169] Therefore, the change amount (absolute value) of each term is as follows. |Δw1f1(x)| = 1.2 |Δw2f2(x)| = 1 |Δw3f3(x)| = 0.18

[0170] This satisfies the satisfaction judgment formula (|Δw1f1(x) + Δw2f2(x)| > |Δw3f3(x)|). That is, it satisfies (|-1.2 + 1|) > 0.18.

[0171] When examining the change of the objective function at this time, Δf all (x) = Δw1f1(x) + Δw2f2(x) + Δw3f3(x) = -1.2 + 1 - 0.18 < 0 and the objective function Δf all (x) is also improved (i.e., decreased) by receiving the improvement (i.e., decrease) of the term with higher priority (w1f1(x) + w2f2(x)). That is, the term with higher priority has an appropriate influence on the objective function. Therefore, it can be determined that the weight coefficient of the objective function can be adjusted so that the improvement judgment of the solution according to the priority can be made.

[0172] <Application to the Work Plan Formulation of a Logistics Warehouse> The method described above will be explained for the case of applying it to the formulation of the work plan of a logistics warehouse.

[0173] FIG. 9 is a diagram for explaining a configuration example of the solution x according to Embodiment 1.

[0174] First, the work plan of the logistics warehouse is defined as the solution x. The solution x has a list structure as shown in FIG. 9 in the program. With the solution x having this list structure, it represents who, in what order, which work, and in which time period to perform the work.

[0175] In FIG. 9, numbers such as "5:", "6:", etc. indicate the operator numbers. Also, "A 13The uppercase letters such as ", B5, and C1 indicate the type of work, and the alphabetical subscript numbers indicate the part number. Also, for the same part number p, A p →C p B p →C p Assume there is an order constraint on A, and no other order constraints. For example, A 13 and A 14 The order of the elements can be rearranged.

[0176] And the objective function f is as follows: all We find the solution x that minimizes (x). This solution x that minimizes the objective function corresponds to the optimized work plan. f all (x)=w1f1(x)+w2f2(x)+w3f3(x)+w4f4(x)+w5f5(x)

[0177] f1(x) is the term for overdue time. f1(x) contains the total time spent working beyond the deadline. f2(x) is the term for the number of order violations. For the same part number, A p →C p B p →C p There is an order constraint, and the number of order violations f2(x) is the number obtained by reversing this order constraint. f3(x) is the term representing the work time. f3(x) contains the sum of the work time for each individual task. f4(x) is the term representing the transition time between tasks. The transition time f4(x) contains the sum of each transition time. f5(x) is the waiting time term. f5(x) represents the total time between the start time and the start time. By minimizing this waiting time, a plan can be created to process tasks that can be done early as early as possible.

[0178] The values ​​of each term in the objective function can be obtained by simulating the solution x using (1) the worker shift schedule and (2) information on each task (start time, deadline, work time, transition time).

[0179] In this embodiment, the priority is set as f1(x)=f2(x)>f3(x)=f4(x)>f5(x), that is, overdue time = number of sequence violations > work time = transition time > waiting time. With this priority, in order to obtain an executable plan, the constraint terms overdue time f1(x) and number of sequence violations f2(x), which should have a value of zero, are given the highest priority, while the waiting time f5(x), which is a term that does not hinder the plan, is given the lowest priority.

[0180] Figure 10 shows an example of the work plan generation result according to Embodiment 1. Figure 11 is a graph showing the changes in each term when the automatic coefficient adjustment is applied and the planning loop is repeated according to Embodiment 1. Figure 12 is a graph showing the changes in each term when a skilled technician manually adjusts the coefficients and the planning loop is repeated.

[0181] The planning device 10 uses the objective function f all The solution x that minimizes (x) = w1f1(x) + w2f2(x) + w3f3(x) + w4f4(x) + w5f5(x) is calculated, and a work plan like the one shown in Figure 10 is generated based on the calculated solution x. The work plan shown in Figure 10 indicates which worker, in what order, and at what time of day each worker with the worker number will perform the tasks.

[0182] In the graphs shown in Figures 11 and 12, the vertical axis represents the evaluation value or penalty value of the term (i.e., w i f i The graph shows the value of (x), and the horizontal axis shows the number of plan updates (i.e., the number of iterations in the planning loop). However, the value on the vertical axis is divided by the maximum value to be in the range of [0,1]. Each line in the graphs shown in Figures 11 and 12 corresponds to each term of the objective function.

[0183] The graph of automatic coefficient adjustment shown in Figure 11, similar to the graph of manual coefficient adjustment shown in Figure 12, initially shows a decrease to zero for the terms with the highest priority: deadline overdue time f1(x) and sequence violation count f2(x). Subsequently, the graph of automatic coefficient adjustment shown in Figure 11 shows a decrease for the terms with the second highest priority: work time f3(x) and switchover time f4(x). Finally, the term with the lowest priority, waiting time f5(x), shows an increasing trend as a result of prioritizing the decrease of the other terms.

[0184] Furthermore, as shown in the graph within the dotted line frame 301 in Figure 11, even terms with different units, such as the deadline overrun time f1(x) and the number of order violations f2(x), can be minimized while maintaining a balance of changes in each term by applying normalization.

[0185] Thus, when the automatic coefficient adjustment according to this embodiment is applied, it is possible to determine coefficients that can appropriately influence the objective function in order of priority, similar to when a skilled technician manually adjusts the coefficients. Therefore, according to this embodiment, a planning device 10 can be provided that allows even inexperienced technicians to easily create work plans.

[0186] <Input screen for priority and normalization flags> Next, we will describe the input screen for users to enter priority and normalization flags.

[0187] Figure 13 is a diagram showing a first example of an input screen according to Embodiment 1.

[0188] The input unit 14 displays an input screen 100A, as shown in Figure 13, on the display device 1005 (see Figure 18). The input screen 100A displays the priority input area 101.

[0189] As shown in FIG. 13, when the user inputs [[3,5],[1,2],[4]] into the priority input area 101, the priorities of f3(x) term = f5(x) term > f1(x) term = f2(x) term > f4(x) term are set. That is, the content of the priority input area 101 indicates that the order of numbers within the outer parentheses shows the priority rank, and the order of numbers within the inner parentheses indicates that the priorities are equal.

[0190] FIG. 14 is a diagram showing a second example of the input screen according to Embodiment 1.

[0191] The input unit 14 displays a priority input screen 100B as shown in FIG. 14 on the display device 1005 (see FIG. 18). The input screen 100B displays a priority input area 101 and a normalization flag input area 102.

[0192] The priority input area 101 is the same as that in FIG. 13.

[0193] As shown in FIG. 14, when the user inputs [[1,1],[0,0],[0]] into the normalization flag input area 102, among [[3,5],[1,2],[4]] input into the priority input area 101, the f3(x) term and f5(x) term corresponding to "1" in the normalization flag input area 102 are set as the normalization targets. That is, the content of the normalization flag input area 102 indicates that the terms in the priority input area 101 corresponding to "1" are the normalization targets.

[0194] FIG. 15 is a diagram showing a third example of the input screen according to Embodiment 1.

[0195] The input unit 14 displays a priority input screen 100C as shown in FIG. 15 on the display device 1005 (see FIG. 18). The input screen 100C displays a priority input area 103 corresponding to each term.

[0196] As shown in Figure 14, when the user enters 2, 2, 1, 3, and 1 in the priority input area 103, corresponding to f1, f2, f3, f4, and f5 respectively, the priority of the f3(x) term = f5(x) term > f1(x) term = f2(x) term > f4(x) term is set. In other words, the contents of the priority input area 103 indicate the priority value of each term.

[0197] Figure 16 shows a fourth example of the input screen according to Embodiment 1.

[0198] The input unit displays a priority input screen 100D, as shown in Figure 16, on the display device 1005 (see Figure 18). The input screen 100D displays a priority input area 103 and a normalization flag input area 104 corresponding to each item.

[0199] The priority input area 103 is the same as in Figure 15.

[0200] As shown in Figure 16, when the user enters 0, 0, 1, 1, and 0 in the normalization flag input area 104, corresponding to f1, f2, f3, f4, and f5 respectively, the f3(x) and f5(x) terms corresponding to "1" in the normalization flag input area 104 are set to be normalized. In other words, the contents of the normalization flag input area 104 indicate that the terms corresponding to "1" are subject to normalization.

[0201] Figure 17 shows a fifth example of the input screen according to Embodiment 1.

[0202] The input unit 14 displays a priority input screen 100E, as shown in Figure 17, on the display device 1005 (see Figure 18). The input screen 100E displays a priority input area 103 and a unit input area 105 corresponding to each item.

[0203] The priority input area 103 is the same as in Figure 15.

[0204] As shown in Figure 17, the user inputs units into the unit input area 105, corresponding to f1, f2, f3, f4, and f5, respectively. The normalization unit 17 sets terms with different units to be normalized. For example, in the input screen 100E shown in Figure 17, the unit "cnt" (number) corresponding to f5 is different from the unit "sec" (second) corresponding to the other f1, f2, f3, and f4, so the normalization unit 17 sets f5 to be normalized.

[0205] <Hardware Configuration> Figure 18 is a block diagram showing the hardware configuration of a computer that implements the functions of the planning device 10 related to this disclosure using a computer program.

[0206] The computer 1000 comprises a processor 1001, memory 1002, storage 1003, input device 1004, display device 1005, and communication device 1006.

[0207] The processor 1001 is a device that executes a computer program stored in the memory 1002 and realizes the functions of the planning device 10 described above. The processor 1001 may be read as a Central Processing Unit (CPU), controller, control unit, or control device. The processor 1001 may also include a Graphics Processing Unit (GPU) and / or a Neural Processing Unit (NPU).

[0208] Memory 1002 is a device that stores computer programs and data handled by computer 1000, and is composed of a volatile storage medium and / or a non-volatile storage medium.

[0209] Storage 1003 is a device composed of a non-volatile storage medium that stores computer programs and data handled by computer 1000. Examples of storage 1003 include a Hard Disk Drive (HDD), a Solid State Drive (SSD), or flash memory.

[0210] The input device 1004 is a device that receives data from the user to be input to the processor 1001. Examples of input devices 1004 include a keyboard, mouse, touchpad, and microphone. For example, the user inputs data through the input device 1004 to the input screens shown in Figures 13 to 17.

[0211] The display device 1005 is a device that displays data generated by the processor 1001. Examples of the display device 1005 include liquid crystal displays and organic EL displays. For example, the planning device 10 displays the input screens shown in Figures 13 to 17 on the display device 1005.

[0212] The communication device 1006 is a device that sends and receives data to and from other devices, such as a server device, via a communication network. Examples of communication networks include wired LAN, wireless LAN, the Internet, mobile communication networks, or Bluetooth®.

[0213] (Summary of Embodiment 1) Based on the description of Embodiment 1 above, the following technology is disclosed.

[0214] <Technology 1> An improvement device according to one embodiment (for example, an improvement determination unit 13) has multiple terms (for example, w i f i (x)) includes the objective function (e.g., f all An input unit (14) accepts the specification of the priority of each term for (x), and the objective function has a predetermined solution (for example, the best solution x best The result of applying ) and the comparative solution (x) obtained by rearranging the solution in the objective function tmp The change in each term (e.g., |Δw) between the result obtained by applying ) and the result obtained by applying ) i f iThe system includes an adjustment unit (e.g., an automatic coefficient adjustment unit 16) that calculates (x)|) and adjusts the objective function such that if the change in the first priority term is less than the sum of the changes in the terms with a lower priority than the first priority, the change in the first priority term is greater than the sum of the changes in the terms with a lower priority than the first priority. This allows the adjustment unit to automatically adjust the objective function.

[0215] <Technology 2> In the improvement device described in Technology 1, if the terms with a priority lower than the first priority include the terms with the second priority and terms with a priority lower than the second priority, the adjustment unit adjusts the objective function so that the amount of change in the first priority term is greater than the sum of the amounts of change in the terms with a priority lower than the first priority, and then adjusts the objective function so that the amount of change in the second priority term is greater than the sum of the amounts of change in the terms with a priority lower than the second priority. This allows the adjustment unit to automatically adjust the objective function.

[0216] <Technology 3> In the improvement device described in Technology 1 or 2, the adjustment unit adjusts the objective function by treating the terms with the same priority as a single term when there are multiple terms with the same priority. This allows the adjustment unit to automatically adjust the objective function.

[0217] <Technology 4> The improvement device described in any one of the technologies 1 to 3 further comprises an improvement determination unit (13) that updates the solution to the comparison solution if the result of applying the comparison solution to the objective function is an improvement over the result of applying the solution to the objective function, and the adjustment unit further calculates the amount of change and adjusts the objective function using the updated solution. This allows the improvement device to automatically adjust the objective function while searching for a more optimal solution.

[0218] <Technology 5> In the improvement device described in Technical 4, the processing by the improvement determination unit and the processing by the adjustment unit are repeated multiple times, and the adjustment unit further normalizes each term of the objective function using the maximum amount of change calculated for each term in the repetition, recalculates the amount of change using the normalized objective function, and adjusts the objective function based on the recalculated amount of change. This allows the improvement device to normalize the terms in the objective function.

[0219] <Technology 6> In the improvement device described in Technical 5, the adjustment unit performs the normalization when the terms included in the objective function include terms with different units. This allows the adjustment unit to normalize terms with different units included in the objective function.

[0220] <Technology 7> In the improvement device described in any one of the technologies 1 to 6, the objective function is a weighted sum of functions using the solutions corresponding to each term, and the adjustment unit adjusts the objective function by adjusting the weights of each term. This allows the adjustment unit to automatically adjust the objective function.

[0221] <Technology 8> In the improvement device described in any one of the technologies 1 to 7, the input unit displays an input screen that accepts the specification of the priority of each of the above items. This allows users to specify the priority of each item through the input screen.

[0222] <Technology 9> In the improvement device described in any one of the technologies 1 to 7, the input unit displays an input screen that accepts the specification of the priority of each item and the specification of the item to be normalized. This allows users to specify the priority of each item and the items to be normalized through the input screen.

[0223] <Technology 10> An improvement method according to one embodiment accepts the designation of the priority of each term in an objective function that includes multiple terms, calculates the change in each term between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function, and adjusts the objective function such that if the change in the term with the first priority is less than the sum of the changes in the terms with a lower priority than the first priority, the change in the term with the first priority is greater than the sum of the changes in the terms with a lower priority than the first priority. This allows the objective function to be automatically adjusted.

[0224] <Technology 11> A planning device according to one embodiment includes: an input unit that accepts the specification of the priority of each term in an objective function that includes multiple terms; an adjustment unit that calculates the amount of change for each term between the result of applying a predetermined solution to the objective function and the result of applying a comparison solution obtained by rearranging the solution to the objective function, and adjusts the objective function so that the amount of change of the first priority term is greater than the sum of the amounts of change of the terms with a priority lower than the first priority if the amount of change of the first priority term is less than the sum of the amounts of change of the terms with a priority lower than the first priority; an improvement determination unit that updates the solution to the comparison solution if the result of applying the comparison solution to the objective function is an improvement over the result of applying the solution to the objective function; and a planning generation unit that generates a plan based on the solution obtained after repeating the processing by the improvement determination unit and the processing by the adjustment unit multiple times. This allows the planning system to create an improved plan based on the explored solutions, with the objective function automatically adjusted.

[0225] While embodiments have been described above with reference to the attached drawings, this disclosure is not limited to such examples. It is clear to those skilled in the art that various modifications, alterations, substitutions, additions, deletions, and equivalents can be conceived within the scope of the claims, and these are also understood to fall within the technical scope of this disclosure. Furthermore, the components of the embodiments described above can be combined in any way without departing from the spirit of the invention. [Industrial applicability]

[0226] The technology disclosed herein is useful for developing improved plans using objective functions. [Explanation of Symbols]

[0227] 10. Planning device 11. Initial Planning Department 12. Planning and Reorganization Department 13 Improvement Judgment Department 14 Input section 15. Objective Function Section 16. Automatic coefficient adjustment unit 17 Normalization section 18. Planning Generation Unit 100A, 100B, 100C, 100D, 100E input screen 101,103 Priority input area 102,104 Normalization flag input area 105 Unit Input Area 1000 computers 1001 Processor 1002 memory 1003 Storage 1004 Input device 1005 Display device 1006 Communication device

Claims

1. An input section that accepts the specification of the priority of each term in an objective function containing multiple terms, The amount of change for each term is calculated between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function. If the change in the first priority term is less than the sum of the changes in the terms with a lower priority than the first priority, the adjustment unit adjusts the objective function so that the change in the first priority term is greater than the sum of the changes in the terms with a lower priority than the first priority. An improvement device equipped with the following features.

2. The adjustment unit is, If the terms with a lower priority than the first priority include terms with a second priority and terms with a lower priority than the second priority, After adjusting the objective function so that the change in the first priority term is greater than the sum of the changes in the terms with lower priority than the first priority, the objective function is then adjusted so that the change in the second priority term is greater than the sum of the changes in the terms with lower priority than the second priority. The improvement device according to claim 1.

3. The adjustment unit is, If there are multiple terms with the same priority level, the terms with the same priority level are treated as a single term, and the objective function is adjusted accordingly. The improvement device according to claim 1.

4. The system further includes an improvement determination unit that updates the solution to the comparison solution if the result of applying the comparison solution to the objective function is better than the result of applying the solution to the objective function, The adjustment unit further calculates the amount of change and adjusts the objective function using the updated solution. The improvement device according to any one of claims 1 to 3.

5. The process by the improvement determination unit and the process by the adjustment unit are repeated multiple times. The adjustment unit further normalizes each term of the objective function using the maximum change calculated for each term in the iteration, recalculates the change using the normalized objective function, and adjusts the objective function based on the recalculated change. The improvement device according to claim 4.

6. The adjustment unit performs the normalization when the terms included in the objective function include terms with different units. The improvement device according to claim 5.

7. The objective function is a weighted sum of functions using the solutions corresponding to each term. The adjustment unit adjusts the objective function by adjusting the weights of each term. The improvement device according to claim 1.

8. The input unit displays an input screen that accepts the specification of the priority of each of the above items. The improvement device according to claim 1.

9. The input unit displays an input screen that accepts the specification of the priority of each item and the specification of the items to be normalized. The improvement device according to claim 5.

10. For objective functions containing multiple terms, we accept the specification of the priority of each term. The amount of change for each term is calculated between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function. If the change in the first priority term is less than the sum of the changes in the terms with a lower priority than the first priority, the objective function is adjusted so that the change in the first priority term is greater than the sum of the changes in the terms with a lower priority than the first priority. How to improve.

11. An input section that accepts the specification of the priority of each term in an objective function containing multiple terms, The amount of change for each term is calculated between the result of applying a predetermined solution to the objective function and the result of applying a comparative solution obtained by rearranging the solution to the objective function. If the change in the first priority term is less than the sum of the changes in the terms with a lower priority than the first priority, the adjustment unit adjusts the objective function so that the change in the first priority term is greater than the sum of the changes in the terms with a lower priority than the first priority. If the result of applying the comparison solution to the objective function is improved compared to the result of applying the solution to the objective function, the improvement determination unit updates the solution to the comparison solution. The system includes a planning generation unit that generates a plan based on the solution after the processing by the improvement determination unit and the processing by the adjustment unit have been repeated multiple times. Planning device.

Images