Distribution result determination device and distribution result determination method

Abstract

A distribution result determination device is configured to have a change cost calculation unit (3) that acquires a first distribution result determined at a first time point and a second distribution result determined at a second time point that is a time point subsequent to the first time point, calculates a change cost as a distribution result indicating a distribution order for a plurality of objects to be distributed, and calculates the change cost as a distribution result indicating a distribution order for the plurality of objects to be distributed. The changed cost is an increase in cost when the allocation result is changed from the first allocation result to the second allocation result. Furthermore, the allocation result determination device is provided with an allocation result selection unit (7) that selects the first allocation result or the second allocation result on the basis of the change cost calculated by the change cost calculation unit (3).

Description

Technical Field

[0001] The present invention relates to an allocation result determination device and an allocation result determination method. Background Art

[0002] As a device for determining the allocation order for a plurality of allocation objects, there is, for example, a landing order determination device for determining the landing order of a plurality of aircraft (for example, refer to Patent Document 1).

[0003] The landing order determination device includes a scheduler, which determines the landing order of the plurality of aircraft based on the estimated arrival time of each aircraft at the runway and the body size of each aircraft. After determining the landing order of the plurality of aircraft, the scheduler re-determines the landing order of the plurality of aircraft if, for example, the estimated arrival time of any aircraft changes.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application No. 2006-523874 Summary of the invention

[0007] Problems to be solved by the invention

[0008] After the landing order of a plurality of aircraft is determined, if the estimated arrival time of any aircraft is changed, the flight cost of changing the landing order may be lower than maintaining the determined landing order, or the flight cost of maintaining the landing order may be lower than changing the landing order. As the flight cost, for example, in addition to the fuel cost of the aircraft, there is also a burden cost related to the physical burden or mental burden of the pilot.

[0009] The landing order determination device disclosed in Patent Document 1 has a problem that after the scheduler determines the landing order of a plurality of aircraft, if the estimated arrival time of any aircraft changes, the landing order of the plurality of aircraft may be changed, thereby sometimes increasing the flight cost.

[0010] The present invention is completed to solve the above-mentioned problems, and its purpose is to obtain an allocation result determination device and an allocation result determination method, which can select the first allocation result or the second allocation result according to the cost when the second allocation result is determined after the first allocation result is determined as an allocation result representing the allocation order for multiple allocation objects.

[0011] Means for solving problems

[0012] The allocation result determination device of the present invention has a change cost calculation unit, which obtains a first allocation result determined at a first moment and a second allocation result determined at a second moment, which is a moment later than the first moment, and calculates a change cost as an allocation result indicating an allocation order for a plurality of allocation objects, wherein the change cost is an increase in cost when the allocation result is changed from the first allocation result to the second allocation result. In addition, the allocation result determination device has an allocation result selection unit, which selects the first allocation result or the second allocation result based on the change cost calculated by the change cost calculation unit.

[0013] Effects of the Invention

[0014] According to the present invention, when the second allocation result is determined as the allocation result indicating the allocation order for a plurality of allocation objects after the first allocation result is determined, the first allocation result or the second allocation result can be selected according to the cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] [ Figure 1 ] is a structural diagram showing the allocation result determination device of implementation mode 1.

[0016] [ Figure 2 ] is a hardware structure diagram showing the hardware of the allocation result determination device of implementation mode 1.

[0017] [ Figure 3 ] is a hardware structure diagram of a computer when the allocation result determination device is implemented by software or firmware.

[0018] [ Figure 4 ] is a structural diagram showing the differential prediction processing unit 6 of the allocation result determination device of implementation mode 1.

[0019] [ Figure 5 ] is an explanatory diagram showing an example of the allocation result representing the landing allocation order of three aircraft.

[0020] [ Figure 6 ] is shown Figure 1 The illustrated flowchart is a processing procedure of the allocation result determination device, that is, an allocation result determination method.

[0021] [ Figure 7 ] Figure 7 A is the scheduling information S a When the first allocation result acquisition unit 1 is provided with the first allocation result X, the first allocation result X acquired by the first allocation result acquisition unit 1 a An illustration of an example of Figure 7 B is the schedule information S bWhen the second allocation result acquisition unit 2 is provided, the second allocation result X acquired by the second allocation result acquisition unit 2 b An illustration of an example of .

[0022] [ Figure 8 ] is an explanatory diagram showing an example of a change cost table.

[0023] [ Fig. 9 ] is an explanatory diagram showing the attenuation function g(j).

[0024] [ Fig.10 ] Fig.10 A is the difference information d indicating that the allocation order of aircraft j4 is changed from the fourth from the beginning to the last. ab , Fig.10 B indicates that the scheduling information S is not included in the a Aircraft j8 in the dispatch information S b The difference information in the case of ab 's illustration.

[0025] [ Fig.11 ] is a structural diagram showing the allocation result determination device of implementation mode 2.

[0026] [ Fig.12 ] is a hardware structure diagram showing the hardware of the allocation result determination device of implementation mode 2.

[0027] [ Fig.13 ] is a structural diagram showing the reward value difference calculation unit 8 of the distribution result determination device according to embodiment 2.

[0028] [ Fig.14 ] is a structural diagram showing the differential prediction processing unit 10 of the allocation result determination device of implementation mode 2.

[0029] [ Fig.15 ] is a structural diagram showing the allocation result determination device of implementation mode 3.

[0030] [ Fig.16 ] is a hardware structure diagram showing the hardware of the allocation result determination device of implementation mode 3.

[0031] [ Fig.17 ] Fig.17 A is an explanatory diagram showing the time when allocation can be made and the time when allocation cannot be made. Fig.17 B is an explanatory diagram showing a penalty table. DETAILED DESCRIPTION

[0032] Hereinafter, in order to explain the present invention in more detail, specific embodiments will be described with reference to the accompanying drawings.

[0033] Implementation method 1.

[0034] Figure 1 This is a diagram showing the structure of the allocation result determination device according to the first embodiment.

[0035] Figure 2 This is a hardware configuration diagram showing the hardware of the allocation result determination device according to the first embodiment.

[0036] Figure 1 The distribution result determination device shown includes a first distribution result acquisition unit 1 , a second distribution result acquisition unit 2 , a change cost calculation unit 3 , a reward value difference prediction unit 4 , and a distribution result selection unit 7 .

[0037] set up Figure 1 The allocation result determination device shown in the figure determines, for example, an allocation result indicating the allocation order of takeoff and landing of a plurality of aircraft as an allocation result indicating the allocation order for a plurality of allocation objects. However, the allocation objects are not limited to aircraft, and may be, for example, cargo or taxis. If the allocation object is, for example, a taxi, then Figure 1 The allocation result determination means shown determines an allocation result indicating the order of dispatching taxi vehicles.

[0038] The first allocation result acquisition unit 1 performs, for example, Figure 2 The first allocation result acquisition circuit 21 shown is implemented.

[0039] The first allocation result acquisition unit 1 obtains the scheduling information S of the plurality of aircrafts that are allocation targets at the first time. a Provided to the first learning model 1a, the first allocation result X is obtained from the first learning model 1a a .

[0040] The first allocation result acquisition unit 1 obtains the first allocation result X a The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0041] Scheduling Information a For example, it includes information indicating the estimated landing time or take-off time of each aircraft and the size of each aircraft. a It is the allocation result determined at the first moment.

[0042] During learning, the first learning model 1 a is provided with scheduling information S of a plurality of aircraft as input data and an allocation result X indicating the allocation order of takeoffs and landings of the plurality of aircraft as training data, and learns the allocation result X.

[0043] The first learning model 1a is provided with the scheduling information S of a plurality of aircrafts during the inference. a When the output is consistent with the scheduling information S aThe corresponding first allocation result X a .

[0044] Here, the first learning model 1a is learned by supervised learning, but this is just an example, and the first learning model 1a may be learned by, for example, unsupervised learning, reinforcement learning, or a mathematical optimization method.

[0045] The second allocation result acquisition unit 2 performs, for example, Figure 2 The second allocation result acquisition circuit 22 shown is implemented.

[0046] The second allocation result acquisition unit 2 obtains the scheduling information S of the plurality of aircrafts that are allocation targets at the second time. b Provided to the second learning model 2a, the second allocation result X is obtained from the second learning model 2a b , the second moment is a moment later than the first moment.

[0047] Scheduling Information b For example, it includes information indicating the estimated landing time or take-off time of each aircraft and the size of each aircraft. b It is the allocation result determined at the second moment.

[0048] The second allocation result acquisition unit 2 obtains the second allocation result X b The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0049] The second learning model 2a is provided with the scheduling information S of a plurality of aircraft as input data and the allocation result X indicating the allocation order of takeoffs and landings of the plurality of aircraft as training data during learning, and learns the allocation result X.

[0050] The second learning model 2a is provided with the scheduling information S of a plurality of aircrafts during the inference. b When the output is consistent with the scheduling information S b The corresponding second allocation result X b .

[0051] Here, the second learning model 2a is learned by supervised learning, but this is just an example, and the second learning model 2a may be learned by unsupervised learning, reinforcement learning, or mathematical optimization method, for example.

[0052] The change cost calculation unit 3 performs, for example, Figure 2 The change cost calculation circuit 23 shown is implemented.

[0053] The change cost calculation unit 3 obtains the first allocation result X from the first allocation result acquisition unit 1. a, obtain the second allocation result X from the second allocation result acquisition unit 2 b .

[0054] The change cost calculation unit 3 calculates the change cost C ab , change cost C ab Is to change the distribution result from the first distribution result X a Change to the second allocation result X b If the allocation object is an aircraft, the cost calculated by the change cost calculation unit 3 is the flight cost. As the flight cost, for example, in addition to the fuel cost of the aircraft, there is also a burden cost related to the pilot's physical burden or the pilot's mental burden.

[0055] The change cost calculation unit 3 calculates the change cost C ab Output to the allocation result selection unit 7.

[0056] The reward value difference prediction unit 4 uses, for example, Figure 2 The reward value difference prediction circuit 24 shown is expressed.

[0057] The reward value difference prediction unit 4 includes a distribution result difference detection unit 5 and a difference prediction processing unit 6 .

[0058] The reward value difference prediction unit 4 allocates 1 to the result X a and the second allocation result X b Provided to Figure 4 The reward value prediction learning model 6c shown in FIG. 1 is used to obtain the first distribution result X from the learning model 6c. a The first reward value R of the degree of goodness or badness preda The sum represents the second allocation result X b The second reward value R of the degree of goodness or badness predb .

[0059] The reward value difference prediction unit 4 uses the second reward value R predb Subtract the first reward value R preda , and thus predict the first reward value R preda and the second reward value R predb The difference in reward value ΔR pred .

[0060] The reward value difference prediction unit 4 calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0061] The allocation result difference detection unit 5 detects the scheduling information S at the first time. a The scheduling information S at the second moment b The difference between abOutput to the differential prediction processing unit 6.

[0062] If the difference information d output from the difference prediction processing unit 6 ab indicates that there is a difference, the difference prediction processing unit 6 assigns 1 to the result X a and the second allocation result X b The first reward value R is obtained from the learning model 6c. preda and the second reward value R predb .

[0063] The difference prediction processing unit 6 calculates the second reward value R predb Subtract the first reward value R preda , and thus predict the first reward value R preda and the second reward value R predb The difference in reward value ΔR pred .

[0064] The difference prediction processing unit 6 calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0065] The allocation result selection unit 7 performs, for example, Figure 2 The distribution result selection circuit 27 shown is implemented.

[0066] The allocation result selection unit 7 selects the change cost C calculated by the change cost calculation unit 3. ab , select the first allocation result X a Or the second distribution result X b .

[0067] Specifically, if the reward value difference ΔR predicted by the reward value difference prediction unit 4 pred Greater than 0 and the change cost is C ab is less than the cost threshold Thc, the allocation result selection unit 7 selects the second allocation result X b .

[0068] If the reward value difference ΔR pred Is below 0 or the change cost C ab is greater than the cost threshold Thc, the allocation result selection unit 7 selects the first allocation result X a .

[0069] The cost threshold value Thc may be stored in the internal memory of the allocation result selection unit 7, or may be provided from outside the allocation result determination device.

[0070] exist Figure 1In the above description, it is assumed that the first allocation result acquisition unit 1, the second allocation result acquisition unit 2, the change cost calculation unit 3, the reward value difference prediction unit 4 and the allocation result selection unit 7, which are the structural elements of the allocation result determination device, are respectively Figure 2 That is, it is assumed that the allocation result determination device is implemented by the first allocation result acquisition circuit 21, the second allocation result acquisition circuit 22, the change cost calculation circuit 23, the reward value difference prediction circuit 24 and the allocation result selection circuit 27.

[0071] The first allocation result acquisition circuit 21, the second allocation result acquisition circuit 22, the change cost calculation circuit 23, the reward value difference prediction circuit 24 and the allocation result selection circuit 27 are, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array) or a circuit formed by combining them.

[0072] The structural elements of the allocation result determination device are not limited to being implemented by dedicated hardware, and the allocation result determination device may also be implemented by software, firmware, or a combination of software and firmware.

[0073] Software or firmware is stored as a program in a computer's memory. A computer refers to hardware that executes programs, such as a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a processor, or a DSP (Digital Signal Processor).

[0074] Figure 3 This is a hardware configuration diagram of a computer when the allocation result determination device is implemented by software, firmware, etc.

[0075] When the allocation result determination device is implemented by software or firmware, a program for causing a computer to execute the respective processing orders of the first allocation result acquisition unit 1, the second allocation result acquisition unit 2, the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7 is stored in the memory 41. Then, the processor 42 of the computer executes the program stored in the memory 41.

[0076] In addition, Figure 2 In FIG. 1 , an example is shown in which the components of the allocation result determination device are respectively implemented by dedicated hardware. Figure 3, an example is shown in which the allocation result determination device is implemented by software or firmware, etc. However, this is just an example, and some of the structural elements in the allocation result determination device may be implemented by dedicated hardware, and the remaining structural elements may be implemented by software or firmware, etc.

[0077] Figure 4 1 is a diagram showing the structure of the difference prediction processing unit 6 of the allocation result determination device according to the first embodiment.

[0078] Figure 4 The difference prediction processing unit 6 shown includes a first prediction processing unit 6a, a second prediction processing unit 6b, a learning model 6c for reward value prediction, and a difference calculation processing unit 6d.

[0079] If the difference information d outputted from the allocation result difference detection unit 5 is ab indicates that there is a difference, the first prediction processing unit 6a will obtain the first allocation result X output from the first allocation result acquisition unit 1 a The reward value R is provided to the learning model 6c for reward value prediction, and the first reward value R is obtained from the learning model 6c. preda .

[0080] The first prediction processing unit 6a calculates the first reward value R preda Output to the difference calculation processing unit 6d.

[0081] If the difference information d outputted from the allocation result difference detection unit 5 is ab indicates that there is a difference, the second prediction processing unit 6b will obtain the second allocation result X output from the second allocation result acquisition unit 2 b The reward value R is provided to the learning model 6c for reward value prediction, and the second reward value R is obtained from the learning model 6c. predb .

[0082] The second prediction processing unit 6b calculates the second reward value R predb Output to the difference calculation processing unit 6d.

[0083] The reward value prediction learning model 6c is provided with the distribution result X as input data and the reward value R as input data during learning. pred As training data, the reward value R pred For example, if the cost of selecting the allocation result X is high, the reward value R pred is a smaller value. If the cost of selecting the allocation result X is lower, the reward value R pred is a larger value.

[0084] The learning model 6c is provided with the first distribution result X during inference. a Or the second distribution result X b When the output is the same as the first distribution result Xa The corresponding first reward value R preda or with the second distribution result X b The corresponding second reward value R predb .

[0085] Here, the learning model 6c is learned by supervised learning, but this is just an example, and the learning model 6c may be learned by unsupervised learning, reinforcement learning, or mathematical optimization method, for example.

[0086] The difference calculation processing unit 6d calculates the second reward value R predb Subtract the first reward value R preda , thus, the first reward value R is calculated preda and the second reward value R predb The difference in reward value ΔR pred .

[0087] The difference calculation processing unit 6d calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0088] Figure 5 This is an explanatory diagram showing an example of the allocation result indicating the landing allocation order of three aircraft.

[0089] exist Figure 5 In the example, the three aircraft are small aircraft, medium aircraft, or large aircraft.

[0090] exist Figure 5 In the example, the landing prohibition time after a small aircraft lands is 60 [sec], the landing prohibition time after a medium aircraft lands is 180 [sec], and the landing prohibition time after a large aircraft lands is 240 [sec].

[0091] If landing is allowed in the order of medium-sized aircraft, large aircraft, and small aircraft, Figure 5 As shown, the shortest time until all three aircraft land is 420 (= 180 + 240) [sec].

[0092] If landing is allowed in the order of medium-sized aircraft, small aircraft, and large aircraft, Figure 5 As shown, the shortest time until all three aircraft land is 240 (= 180 + 60) [sec].

[0093] Therefore, when landing is permitted in the order of medium-sized aircraft, small aircraft, and large aircraft, the shortest time until all aircraft land is shortened by 180 (= 420-240) [sec] compared to when landing is permitted in the order of medium-sized aircraft, large aircraft, and small aircraft.

[0094] Next, Figure 1 The operation of the allocation result determination device shown will be described.

[0095] Figure 6 It is shown Figure 1 The illustrated flowchart is a processing procedure of the allocation result determination device, that is, an allocation result determination method.

[0096] The first allocation result acquisition unit 1 acquires the scheduling information S of a plurality of aircraft at a first time. a .

[0097] Scheduling Information a For example, it includes information indicating the estimated landing time or the estimated take-off time of each aircraft and the size of each aircraft.

[0098] The first allocation result acquisition unit 1 receives the scheduling information S a Provided to the first learning model 1a, the first allocation result X is obtained from the first learning model 1a a ( Figure 6 Step ST1).

[0099] The first allocation result acquisition unit 1 obtains the first allocation result X a The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0100] Figure 7 A is the scheduling information S a When the first allocation result acquisition unit 1 is provided with the first allocation result X, the first allocation result X acquired by the first allocation result acquisition unit 1 a An illustration of an example of .

[0101] exist Figure 7 In A, t1, t2, ..., t8 are time points, and j1, j2, ..., j5 are IDs (IDentification) for identifying aircraft.

[0102] "0" means that the aircraft takeoff and landing cannot be allocated, and "1" means that the aircraft takeoff and landing can be allocated.

[0103] exist Figure 7 In the example of A, the first allocation result X is obtained in which takeoff and landing are allowed in the order of aircraft j3, aircraft j5, aircraft j1, aircraft j2, and aircraft j4. a .

[0104] The second allocation result acquisition unit 2 acquires the dispatch information S of a plurality of aircraft at a second time which is a time later than the first time. b .

[0105] Scheduling Information bFor example, it includes information indicating the estimated landing time or the estimated take-off time of each aircraft and the size of each aircraft.

[0106] The second allocation result acquisition unit 2 receives the scheduling information S b Provided to the second learning model 2a, the second allocation result X is obtained from the second learning model 2a b ( Figure 6 Step ST2).

[0107] The second allocation result acquisition unit 2 obtains the second allocation result X b The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0108] Figure 7 B is the schedule information S b When the second allocation result acquisition unit 2 is provided, the second allocation result X acquired by the second allocation result acquisition unit 2 b An illustration of an example of .

[0109] exist Figure 7 In B, t1, t2, ..., t8 are time moments, and j1, j2, ..., j5 are IDs for identifying aircraft.

[0110] "0" means that the aircraft takeoff and landing cannot be allocated, and "1" means that the aircraft takeoff and landing can be allocated.

[0111] exist Figure 7 In the example of B, the second allocation result X is obtained in which takeoff and landing are allowed in the order of aircraft j3, aircraft j1, aircraft j5, aircraft j2, and aircraft j4. b .

[0112] The change cost calculation unit 3 obtains the first allocation result X from the first allocation result acquisition unit 1. a , obtain the second allocation result X from the second allocation result acquisition unit 2 b .

[0113] The change cost calculation unit 3 refers to, for example, Figure 8 The change cost table shown in the figure calculates the change cost C ab , change cost C ab Is to change the distribution result from the first distribution result X a Change to the second allocation result X b The increase in cost under the condition of Figure 6 Step ST3).

[0114] The change cost calculation unit 3 calculates the change cost C ab Output to the allocation result selection unit 7.

[0115] Figure 8 This is an explanatory diagram showing an example of a change cost table.

[0116] exist Figure 8 In the table, j1, j2, ..., j5 are identification symbols of aircraft. The numbers in the table represent the change costs.

[0117] For example, in the first allocation the result is X a =[j3, j5, j1, j2, j4], the second allocation result is X b = [j3, j1, j5, j2, j4], the order of aircraft j5 and aircraft j1 is swapped. Therefore, the change cost C ab is "100".

[0118] For example, in the first allocation the result is X a =[j3, j5, j1, j2, j4], the second allocation result is X b =[j3, j2, j5, j1, j4], the order of aircraft j5 and aircraft j2 is swapped, and then the order of aircraft j5 and aircraft j1 is swapped. Therefore, the change cost C ab is "180" (=80+100).

[0119] exist Figure 1 In the allocation result determination device shown in FIG. 1 , the change cost calculation unit 3 refers to Figure 8 The change cost table shown in the figure calculates the change cost C ab However, this is just an example, and the change cost calculation unit 3 may also calculate the change cost C as follows: ab .

[0120] First, as shown in the following formula (1), the change cost calculation unit 3 calculates the second allocation result X b 'Subtract the first allocation result X a , and the distribution difference ΔX is calculated. b 'Make the second distribution result X b The moment and the first distribution result X a For example, if the first allocation result X a The time is t1, t2, ..., t8, the second allocation result X b The time is t3, t4, ..., t 10 , then let the second allocation result X b The time t3 is t1, the time t4 is t2, and the time t 10 is t8.

[0121] ΔX=X b '-X a (1)

[0122] Next, the change cost calculation unit 3 substitutes the allocation difference ΔX into the following equation (2) to calculate the change cost C0 associated with the change in the order.

[0123] Furthermore, the change cost calculation unit 3 calculates the change cost Ct associated with the change in time by substituting the allocation difference ΔX into the following equation (3).

[0124] C o =γ o [g(j)⊙(1-d ab )]·|ΔX| (2)

[0125] C t =γ t [g(j)☉(1-d ab )]·ΔX 2 (3)

[0126] In equations (2) and (3), ⊙ is a mathematical symbol representing an inner product.

[0127] g(j) is Fig. 9 The decay function shown, for example, g(j) = e (-j / T) . j is the ID of the aircraft, and T is the time constant.

[0128] d ab is the difference information d outputted from the allocation result difference detection unit 5 to the change cost calculation unit 3. ab .exist Figure 1 In the example, the arrow from the allocation result difference detection unit 5 to the change cost calculation unit 3 is omitted. a With scheduling information b The difference is d ab = 0, if there is scheduling information S a With scheduling information b The difference is d ab =1.

[0129] γ0, γ t are the coefficients respectively.

[0130] For example, as shown in the following formula (4), the change cost calculation unit 3 performs weighted addition of the change cost C0 associated with the change in sequence and the change cost Ct associated with the change in time to calculate the change cost C ab .

[0131] C ab =C0+w·C t (4)

[0132] In formula (4), w is the weight coefficient.

[0133] The reward value difference prediction unit 4 predicts the reward value difference ΔR pred ( Figure 6 Step ST4).

[0134] The following specifically describes how the reward value difference prediction unit 4 calculates the reward value difference ΔR. pred prediction processing.

[0135] The allocation result difference detection unit 5 of the reward value difference prediction unit 4 obtains the scheduling information S at the first time. a and the scheduling information S at the second moment b .

[0136] like Fig.10 As shown, the allocation result difference detection unit 5 detects the scheduling information S a With scheduling information b The difference between ab The change cost C is output to the difference prediction processing unit 6. The change cost calculation unit 3 calculates the change cost C by using the formula (4): ab In the case of ab It is also output to the change cost calculation unit 3 .

[0137] Fig.10 A is the difference information d indicating that the allocation order of aircraft j4 is changed from the fourth from the beginning to the last. ab 's illustration.

[0138] Fig.10 B indicates that the scheduling information S is not included in the a Aircraft j8 in the dispatch information S b The difference information in the case of ab 's illustration.

[0139] exist Fig.10 A and Fig.10 In B, the numbers in the circle are the IDs for identifying the aircraft. However, the symbol j is omitted.

[0140] If there is no scheduling information S a With scheduling information b The difference between ab = 0, if there is scheduling information S a With scheduling information b The difference between ab =1.

[0141] The first prediction processing unit 6a of the difference prediction processing unit 6 obtains the first allocation result X from the first allocation result obtaining unit 1 a , obtain the difference information d from the allocation result difference detection unit 5 ab .

[0142] If the difference information d ab is "1", the first prediction processing unit 6a converts the first allocation result X a The reward value R is provided to the learning model 6c for reward value prediction, and the first reward value R is obtained from the learning model 6c. preda .

[0143] The first prediction processing unit 6a calculates the first reward value R preda Output to the difference calculation processing unit 6d.

[0144] The second prediction processing unit 6b of the difference prediction processing unit 6 obtains the second allocation result X from the second allocation result obtaining unit 2 b , obtain the difference information d from the allocation result difference detection unit 5 ab .

[0145] If the difference information d ab is "1", the second prediction processing unit 6b sets the second allocation result X b The reward value R is provided to the learning model 6c for reward value prediction, and the second reward value R is obtained from the learning model 6c. predb .

[0146] The second prediction processing unit 6b calculates the second reward value R predb Output to the difference calculation processing unit 6d.

[0147] The difference calculation processing unit 6d obtains the first reward value R from the first prediction processing unit 6a. preda , obtain the second reward value R from the second prediction processing unit 6b predb .

[0148] As shown in the following formula (5), the difference calculation processing unit 6d calculates the second reward value R predb Subtract the first reward value R preda , thus, the first reward value R is calculated preda and the second reward value R predb The difference in reward value ΔR pred . In the reward value difference ΔR pred If it is a negative value, the second allocation result X is selected. b The cost ratio of the case is to choose the first allocation result X a The cost is high in the case of pred If it is a positive value, the second allocation result X is selected. b The cost ratio of the case is to choose the first allocation result X a The cost is low in the case.

[0149] ΔR pred =R predb -Rpreda (5)

[0150] The difference calculation processing unit 6d calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0151] The allocation result selection unit 7 obtains the first allocation result X from the first allocation result acquisition unit 1. a , obtain the second allocation result X from the second allocation result acquisition unit 2 b .

[0152] The allocation result selection unit 7 selects the change cost C calculated by the change cost calculation unit 3. ab and the reward value difference ΔR predicted by the reward value difference prediction unit 4 pred , select the first allocation result X a Or the second distribution result X b ( Figure 6 Step ST5).

[0153] That is, if the reward value difference ΔR predicted by the reward value difference prediction unit 4 pred Greater than 0 and the change cost is C ab is less than the cost threshold Thc, the allocation result selection unit 7 selects the second allocation result X b .

[0154] If the reward value difference ΔR pred Is below 0 or the change cost C ab is greater than the cost threshold Thc, the allocation result selection unit 7 selects the first allocation result X a .

[0155] exist Figure 1 In the allocation result determination device shown in FIG. 1 , the allocation result selection unit 7 selects the allocation result according to the change cost C ab and the difference in reward value ΔR pred , select the first allocation result X a Or the second distribution result X b However, this is just an example, and the allocation result selection unit 7 may also select only the change cost C ab , select the first allocation result X a Or the second distribution result X b In the allocation result selection unit 7, only the change cost C ab Select the first allocation result X a Or the second distribution result X b In this case, the allocation result determination device does not need to include the reward value difference prediction unit 4.

[0156] Alternatively, the allocation result selection unit 7 may select the reward value difference ΔR only. pred, select the first allocation result X a Or the second distribution result X b In the allocation result selection unit 7, only the reward value difference ΔR pred Select the first allocation result X a Or the second distribution result X b In this case, the allocation result determination device does not need to have a change cost calculation unit 3.

[0157] In the above-mentioned embodiment 1, the allocation result determination device is configured to include a change cost calculation unit 3, which obtains the first allocation result determined at the first moment and the second allocation result determined at the second moment which is a moment later than the first moment as the allocation result indicating the allocation order for a plurality of allocation objects, and calculates the change cost, which is the increase in cost when the allocation result is changed from the first allocation result to the second allocation result. In addition, the allocation result determination device includes an allocation result selection unit 7, which selects the first allocation result or the second allocation result based on the change cost calculated by the change cost calculation unit 3. Therefore, in the case where the second allocation result is determined as the allocation result indicating the allocation order for a plurality of allocation objects after the first allocation result is determined, the allocation result determination device can select the first allocation result or the second allocation result based on the cost.

[0158] Implementation method 2.

[0159] In the second embodiment, a distribution result determination device including the reward value difference prediction unit 9 which updates the learning model 10c will be described.

[0160] Fig.11 2 is a block diagram showing an allocation result determination device according to Embodiment 2. Fig.11 In, with Figure 1 The same reference numerals denote the same or corresponding parts, and thus description thereof will be omitted.

[0161] Fig.12 2 is a hardware configuration diagram showing the hardware of the allocation result determination device according to the second embodiment. Fig.12 In, with Figure 2 The same reference numerals denote the same or corresponding parts, and thus description thereof will be omitted.

[0162] Fig.11 The distribution result determination device shown includes a first distribution result acquisition unit 1 , a second distribution result acquisition unit 2 , a change cost calculation unit 3 , a reward value difference prediction unit 9 , a distribution result selection unit 7 , and a reward value difference calculation unit 8 .

[0163] The reward value difference calculation unit 8 performs, for example, Fig.12 The reward value difference calculation circuit 28 shown is implemented.

[0164] The reward value difference calculation unit 8 calculates the first distribution result X a Provided to the reward function to calculate the first reward value R a , the second distribution result X b Provided to the reward function to calculate the second reward value R b .

[0165] The reward value difference calculation unit 8 calculates the second reward value R b Subtract the first reward value R a , thus, the first reward value R is calculated a and the second reward value R b The reward value difference ΔR.

[0166] The reward value difference calculation unit 8 outputs the reward value difference ΔR to the reward value difference prediction unit 9 .

[0167] For example, the reward value difference prediction unit 9 Fig.12 The reward value difference prediction circuit 29 shown is implemented.

[0168] The reward value difference prediction unit 9 includes a distribution result difference detection unit 5 and a difference prediction processing unit 10 .

[0169] The reward value difference prediction unit 9 allocates the result X a and the second allocation result X b Provided to Fig.14 The reward value prediction learning model 10c shown in FIG. 1 is used to obtain the first distribution result X from the learning model 10c. a The first reward value R of the degree of goodness or badness preda The sum represents the second allocation result X b The second reward value R of the degree of goodness or badness predb .

[0170] The reward value difference prediction unit 9 calculates the second reward value R predb Subtract the first reward value R preda , thus, predicting the first reward value R preda and the second reward value R predb The difference in reward value ΔR pred .

[0171] The reward value difference prediction unit 9 calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0172] Furthermore, the reward value difference prediction unit 9 updates the learning model 10c so that the predicted reward value difference ΔR pred The difference from the reward value difference ΔR calculated by the reward value difference calculation unit 8 becomes smaller.

[0173] exist Fig.11 In the above, it is assumed that the first allocation result acquisition unit 1, the second allocation result acquisition unit 2, the change cost calculation unit 3, the reward value difference prediction unit 9, the allocation result selection unit 7 and the reward value difference calculation unit 8, which are the structural elements of the allocation result determination device, are respectively Fig.12 That is, it is assumed that the allocation result determination device is implemented by the first allocation result acquisition circuit 21, the second allocation result acquisition circuit 22, the change cost calculation circuit 23, the reward value difference prediction circuit 29, the allocation result selection circuit 27 and the reward value difference calculation circuit 28.

[0174] The first allocation result acquisition circuit 21, the second allocation result acquisition circuit 22, the change cost calculation circuit 23, the reward value difference prediction circuit 29, the allocation result selection circuit 27 and the reward value difference calculation circuit 28 are, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA or a circuit formed by combining them.

[0175] The structural elements of the allocation result determination device are not limited to being implemented by dedicated hardware, and the allocation result determination device may also be implemented by software, firmware, or a combination of software and firmware.

[0176] When the allocation result determination device is implemented by software or firmware, a program for causing a computer to execute the respective processing orders of the first allocation result acquisition unit 1, the second allocation result acquisition unit 2, the change cost calculation unit 3, the reward value difference prediction unit 9, the allocation result selection unit 7, and the reward value difference calculation unit 8 is stored in Figure 3 The memory 41 shown. Then, Figure 3 The illustrated processor 42 executes programs stored in the memory 41 .

[0177] In addition, Fig.12 In FIG. 1 , an example is shown in which the components of the allocation result determination device are respectively implemented by dedicated hardware. Figure 3 , an example is shown in which the allocation result determination device is implemented by software or firmware, etc. However, this is just an example, and some of the structural elements in the allocation result determination device may be implemented by dedicated hardware, and the remaining structural elements may be implemented by software or firmware, etc.

[0178] Fig.13 1 is a diagram showing the configuration of the reward value difference calculation unit 8 of the distribution result determination device according to the second embodiment.

[0179] Fig.13 The reward value difference calculation unit 8 shown includes a first reward value calculation unit 8a, a second reward value calculation unit 8b, and a difference calculation processing unit 8c.

[0180] The first reward value calculation unit 8a obtains the first distribution result X from the first distribution result acquisition unit 1 a .

[0181] The first reward value calculation unit 8a calculates the first distribution result X a Provided to the reward function to calculate the first reward value R a , the first reward value R a Output to the difference calculation processing unit 8c.

[0182] The second reward value calculation unit 8b obtains the second distribution result X from the second distribution result acquisition unit 2 b .

[0183] The second reward value calculation unit 8b calculates the second distribution result X b Provided to the reward function to calculate the second reward value R b , the second reward value R b Output to the difference calculation processing unit 8c.

[0184] The difference calculation processing unit 8c obtains the first reward value R from the first reward value calculation unit 8a. a , obtain the second reward value R from the second distribution result acquisition unit 2 b .

[0185] The difference calculation processing unit 8c calculates the second reward value R b Subtract the first reward value R a , thus, the first reward value R is calculated a and the second reward value R b The reward value difference ΔR.

[0186] The difference calculation processing unit 8 c outputs the reward value difference ΔR to the reward value difference prediction unit 9 .

[0187] Fig.14 It is a diagram showing the structure of the difference prediction processing unit 10 of the allocation result determination device according to the second embodiment.

[0188] Fig.14 The difference prediction processing unit 10 shown includes a first prediction processing unit 10a, a second prediction processing unit 10b, a learning model 10c for reward value prediction, and a difference calculation processing unit 10d.

[0189] If the difference information d outputted from the allocation result difference detection unit 5 is ab indicates that there is a difference, the first prediction processing unit 10a will obtain the first allocation result X output from the first allocation result acquisition unit 1 a The reward value R is provided to the learning model 10c for reward value prediction, and the first reward value R is obtained from the learning model 10c. preda .

[0190] The first prediction processing unit 10a calculates the first reward value R preda Output to the difference calculation processing unit 10d.

[0191] If the difference information d outputted from the allocation result difference detection unit 5 is ab indicates that there is a difference, the second prediction processing unit 10b will output the second allocation result X output from the second allocation result acquisition unit 2 b The reward value R is provided to the learning model 10c for reward value prediction, and the second reward value R is obtained from the learning model 10c. predb .

[0192] The second prediction processing unit 10b calculates the second reward value R predb Output to the difference calculation processing unit 10d.

[0193] The reward value prediction learning model 10c is provided with the distribution result X as input data and the reward value R as input data during learning. pred As training data, the reward value R pred To study.

[0194] The learning model 10c is provided with the first distribution result X during inference. a Or the second distribution result X b When the output is the same as the first distribution result X a The corresponding first reward value R preda or with the second distribution result X b The corresponding second reward value R predb .

[0195] Here, the learning model 10c is learned by supervised learning, but this is just an example, and the learning model 10c may be learned by, for example, unsupervised learning, reinforcement learning, or mathematical optimization method.

[0196] The difference calculation processing unit 10d calculates the second reward value R predb Subtract the first reward value R preda , thus, the first reward value R is calculated preda and the second reward value R predb The difference in reward value ΔR pred .

[0197] Next, Fig.11 The operation of the allocation result determination device shown in FIG. 1 is described below. However, except for the reward value difference calculation unit 8 and the reward value difference prediction unit 9, Figure 1 Therefore, only the operations of the reward value difference calculation unit 8 and the reward value difference prediction unit 9 will be described here.

[0198] The first reward value calculation unit 8a of the reward value difference calculation unit 8 obtains the first distribution result X from the first distribution result acquisition unit 1 a .

[0199] The first reward value calculation unit 8a calculates the first distribution result X a The first reward value R is calculated by applying the reward function shown in the following formula (6): a .

[0200] R a =R assigna +α·R separationa (6)

[0201] In formula (6), R assigna It is an evaluation value used to evaluate whether the allocation time of each aircraft is appropriate. assigna According to the first distribution result X a The determined value becomes larger as the allocation time of each aircraft is earlier within the range of allocatable time.

[0202] R separationa It is an evaluation value related to the allocation interval of multiple aircraft. separationa According to the first distribution result X a A determined value. If the allocation interval is larger than the minimum interval that can be allocated, the smaller the allocation interval, the larger the value.

[0203] α is the weight coefficient.

[0204] The first reward value calculation unit 8a calculates the first reward value R a Output to the difference calculation processing unit 8c.

[0205] The second reward value calculation unit 8b obtains the second distribution result X from the second distribution result acquisition unit 2 b .

[0206] The second reward value calculation unit 8b calculates the second distribution result X b The second reward value R is calculated by applying the reward function shown in the following formula (7): b .

[0207] R b =R assignb +β·R separationb (7)

[0208] In formula (7), R assignb It is an evaluation value used to evaluate whether the allocation time of each aircraft is appropriate. assignb According to the second distribution result X b The determined value becomes larger as the allocation time of each aircraft is earlier within the range of allocatable time.

[0209] R separationb It is an evaluation value related to the allocation interval of multiple aircraft. separationb According to the second distribution result X b A determined value. If the allocation interval is larger than the minimum interval that can be allocated, the smaller the allocation interval, the larger the value.

[0210] β is the weight coefficient.

[0211] The second reward value calculation unit 8b calculates the second reward value R b Output to the difference calculation processing unit 8c.

[0212] The difference calculation processing unit 8c obtains the first reward value R from the first reward value calculation unit 8a. a , obtain the second reward value R from the second distribution result acquisition unit 2 b .

[0213] As shown in the following formula (8), the difference calculation processing unit 8c calculates the second reward value R b Subtract the first reward value R a , thus, the first reward value R is calculated a and the second reward value R b The reward value difference ΔR.

[0214] ΔR pred =R b -R a (8)

[0215] The difference calculation processing unit 8 c outputs the reward value difference ΔR to the difference prediction processing unit 10 of the reward value difference prediction unit 9 .

[0216] The first prediction processing unit 10a of the difference prediction processing unit 10 obtains the first allocation result X from the first allocation result obtaining unit 1 a , obtain the difference information d from the allocation result difference detection unit 5 ab .

[0217] If the difference information d ab is "1", the first prediction processing unit 10a will a The reward value R is provided to the learning model 10c for reward value prediction, and the first reward value R is obtained from the learning model 10c. preda .

[0218] The first prediction processing unit 10a calculates the first reward value R preda Output to the difference calculation processing unit 10d.

[0219] The second prediction processing unit 10b obtains the second allocation result X from the second allocation result acquisition unit 2 b, obtain the difference information d from the allocation result difference detection unit 5 ab .

[0220] If the difference information d ab is "1", the second prediction processing unit 10b sets the second allocation result X b The reward value R is provided to the learning model 10c for reward value prediction, and the second reward value R is obtained from the learning model 10c. predb .

[0221] The second prediction processing unit 10b calculates the second reward value R predb Output to the difference calculation processing unit 10d.

[0222] The difference calculation processing unit 10d obtains the first reward value R from the first prediction processing unit 10a. preda , obtain the second reward value R from the second prediction processing unit 10b predb .

[0223] As shown in the above formula (5), the difference calculation processing unit 10d calculates the second reward value R from predb Subtract the first reward value R preda , thus, the first reward value R is calculated preda and the second reward value R predb The difference in reward value ΔR pred .

[0224] The difference calculation processing unit 10d calculates the reward value difference ΔR pred Output to the allocation result selection unit 7.

[0225] The first prediction processing unit 10a and the second prediction processing unit 10b update the learning model 10c respectively so that the reward value difference ΔR calculated by the difference calculation processing unit 10d is pred The difference from the reward value difference ΔR calculated by the difference calculation processing unit 8 c of the reward value difference calculation unit 8 becomes smaller.

[0226] Specifically, the first prediction processing unit 10a and the second prediction processing unit 10b each update the weight of the learning model 10c so that (ΔR - ΔR pred ) 2 Minimum.

[0227] In the above embodiment 2, Fig.11 The distribution result determination device shown is configured to include a reward value difference calculation unit 8, which provides the first distribution result to the reward function to calculate the first reward value, provides the second distribution result to the reward function to calculate the second reward value, and subtracts the first reward value from the second reward value, thereby calculating the reward value difference between the first reward value and the second reward value. In addition, Fig.11The distribution result determination device shown in FIG. 1 updates the learning model 10c so that the difference between the reward value difference predicted by the reward value difference prediction unit 9 and the reward value difference calculated by the reward value difference calculation unit 8 becomes smaller. Figure 1 The distribution results shown determine the device compared to Fig.11 The allocation result determination device shown can improve the selection accuracy of the allocation result.

[0228] Implementation method 3.

[0229] In the third embodiment, an allocation result determination device including the penalty value calculation unit 11 will be described.

[0230] Fig.15 2 is a block diagram showing an allocation result determination device according to Embodiment 3. Fig.15 In, with Figure 1 The same reference numerals denote the same or corresponding parts, and thus description thereof will be omitted.

[0231] Fig.16 2 is a hardware configuration diagram showing the hardware of the allocation result determination device according to Embodiment 3. Fig.16 In, with Figure 2 The same reference numerals denote the same or corresponding parts, and thus description thereof will be omitted.

[0232] Fig.15 The distribution result determination device shown includes a first distribution result acquisition unit 1 , a second distribution result acquisition unit 15 , a change cost calculation unit 3 , a reward value difference prediction unit 4 , a distribution result selection unit 7 , and a penalty value calculation unit 11 .

[0233] The penalty value calculation unit 11 is, for example, Fig.16 The penalty value calculation circuit 31 shown is implemented.

[0234] The penalty value calculation unit 11 includes a penalty value calculation processing unit 12 , an objective function value calculation unit 13 , and a function value addition unit 14 .

[0235] If there is an allocation violation in the allocation result selected by the allocation result selection unit 7 , the penalty value calculation unit 11 calculates a penalty value for the allocation violation.

[0236] The penalty value calculation unit 11 outputs the penalty value to the second allocation result acquisition unit 15 .

[0237] If there is an allocation violation in the allocation result selected by the allocation result selection unit 7 , the penalty value calculation processing unit 12 calculates a penalty value for the allocation violation.

[0238] The penalty value calculation processing unit 12 outputs the penalty value to the function value adding unit 14 .

[0239] The objective function value calculation unit 13 provides the allocation result selected by the allocation result selection unit 7 to the objective function to calculate the value of the objective function, that is, the objective function value.

[0240] The objective function value calculation unit 13 outputs the objective function value to the function value addition unit 14 .

[0241] The function value adding unit 14 adds the objective function value calculated by the objective function value calculating unit 13 to the penalty value calculated by the penalty value calculation processing unit 12 .

[0242] The function value adding unit 14 outputs the penalty value obtained by adding the objective function value to the second allocation result acquiring unit 15 .

[0243] exist Fig.15 In the allocation result determination device shown, the penalty value calculation unit 11 includes a penalty value calculation processing unit 12, an objective function value calculation unit 13, and a function value addition unit 14. However, this is just an example, and for example, the penalty value calculation unit 11 may include only one of the penalty value calculation processing unit 12 and the objective function value calculation unit 13. In the case where the penalty value calculation unit 11 includes only the penalty value calculation processing unit 12, the penalty value calculated by the penalty value calculation processing unit 12 is output to the second allocation result acquisition unit 15. In the case where the penalty value calculation unit 11 includes only the objective function value calculation unit 13, the objective function value is output to the second allocation result acquisition unit 15 as the penalty value.

[0244] The second allocation result acquisition unit 15 performs, for example, Fig.16 The second allocation result acquisition circuit 35 shown is implemented.

[0245] The second allocation result acquisition unit 15 receives the second time schedule information S b Provided to the second learning model 15a, the second allocation result X is obtained from the second learning model 15a b .

[0246] The second allocation result acquisition unit 15 obtains the second allocation result X b The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0247] Furthermore, the second allocation result acquisition unit 15 updates the second learning model 15 a so that the penalty value calculated by the penalty value calculation unit 11 becomes smaller.

[0248] The second learning model 15 a is provided with the scheduling information S of a plurality of aircraft as input data and the allocation result X indicating the allocation order of takeoffs and landings of the plurality of aircraft as training data during learning, and learns the allocation result X.

[0249] The second learning model 15a is provided with the scheduling information S of a plurality of aircrafts during the inference. b When the output is consistent with the scheduling information S b The corresponding second allocation result X b .

[0250] Here, the second learning model 15a is learned by supervised learning, but this is just an example, and the second learning model 15a may be learned by, for example, unsupervised learning, reinforcement learning, or mathematical optimization method.

[0251] exist Fig.15 In the allocation result determination device shown in FIG. 1 , the second allocation result acquisition unit 15 and the penalty value calculation unit 11 are respectively applied to Figure 1 However, this is just an example, and the second allocation result acquisition unit 15 and the penalty value calculation unit 11 may be respectively applied to Fig.11 The allocation result shown determines the device.

[0252] exist Fig.15 In the allocation result determination device, the first allocation result acquisition unit 1, the second allocation result acquisition unit 15, the change cost calculation unit 3, the reward value difference prediction unit 4, the allocation result selection unit 7 and the penalty value calculation unit 11 are respectively configured by Fig.16 That is, it is assumed that the allocation result determination device is implemented by the first allocation result acquisition circuit 21, the second allocation result acquisition circuit 35, the change cost calculation circuit 23, the reward value difference prediction circuit 24, the allocation result selection circuit 27 and the penalty value calculation circuit 31.

[0253] The first allocation result acquisition circuit 21, the second allocation result acquisition circuit 35, the change cost calculation circuit 23, the reward value difference prediction circuit 24, the allocation result selection circuit 27 and the penalty value calculation circuit 31 are, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA or a circuit formed by combining them.

[0254] The structural elements of the allocation result determination device are not limited to being implemented by dedicated hardware, and the allocation result determination device may also be implemented by software, firmware, or a combination of software and firmware.

[0255] When the allocation result determination device is implemented by software or firmware, a program for causing a computer to execute the respective processing orders of the first allocation result acquisition unit 1, the second allocation result acquisition unit 15, the change cost calculation unit 3, the reward value difference prediction unit 4, the allocation result selection unit 7, and the penalty value calculation unit 11 is stored in Figure 3 The memory 41 shown. Then, Figure 3The illustrated processor 42 executes programs stored in the memory 41 .

[0256] In addition, Fig.16 In FIG. 1 , an example is shown in which the components of the allocation result determination device are respectively implemented by dedicated hardware. Figure 3 , an example is shown in which the allocation result determination device is implemented by software or firmware, etc. However, this is just an example, and some of the structural elements in the allocation result determination device may be implemented by dedicated hardware, and the remaining structural elements may be implemented by software or firmware, etc.

[0257] Next, Fig.15 The operation of the allocation result determination device shown in FIG. 1 is described below. However, except for the penalty value calculation unit 11 and the second allocation result acquisition unit 15, Figure 1 Therefore, only the operations of the penalty value calculation unit 11 and the second allocation result acquisition unit 15 will be described here.

[0258] The penalty value calculation processing unit 12 of the penalty value calculation unit 11 obtains the first allocation result X a Or the second distribution result X b , as the allocation result X selected by the allocation result selection unit 7 sel .

[0259] The penalty value calculation processing unit 12 determines the allocation result X sel Whether the allocation time of each aircraft shown is an allocatable time.

[0260] Fig.17 A is an explanatory diagram showing the time when allocation can be made and the time when allocation cannot be made.

[0261] exist Fig.17 In A, t1, t2, ..., t8 are time points, and j1, j2, ..., j5 are IDs for identifying aircraft.

[0262] "0" indicates a time that cannot be allocated, and "1" indicates a time that can be allocated.

[0263] If the distribution results in X sel If the allocation time of each aircraft is allocated to an allocatable time, there is no allocation violation in the allocation result. If it is allocated to an unallocatable time, there is an allocation violation in the allocation result.

[0264] Fig.17 B is an explanatory diagram showing a penalty table.

[0265] Fig.17 The penalty table shown in B shows the penalty value when the data is allocated to the time when the data can be allocated and the penalty value when the data is allocated to the time when the data cannot be allocated.

[0266] exist Fig.17 In the example of B, the penalty value when the data is allocated to the time that can be allocated is "0", and the penalty value when the data is allocated to the time that cannot be allocated is a negative value.

[0267] For example, regarding the penalty value in the case of allocation to a time earlier than the allocatable time, the absolute value becomes larger as the allocation to a time earlier than the allocatable time becomes.

[0268] If there is an allocation violation, the penalty value calculation processing unit 12 refers to Fig.17 The penalty table shown in B calculates the penalty value p.

[0269] For example, if there is an allocation violation in which the aircraft j2 is allocated to the time t2 and an allocation violation in which the aircraft j3 is allocated to the time t5, the penalty value p is -510 (= -500 - 10).

[0270] For example, if there is only an allocation violation in which aircraft j5 is allocated to time t6, the penalty value p is -5.

[0271] The penalty value calculation processing unit 12 outputs the penalty value p to the function value adding unit 14 .

[0272] Here, the penalty value calculation processing unit 12 refers to Fig.17 B, and calculates the penalty value. However, this is just an example, and for example, the penalty value calculation processing unit 12 may also calculate the penalty value of the distribution result X. sel Provided to the penalty function p(X) shown in the following formula (9) sel ), calculate the penalty function p(X sel ) is the penalty value p.

[0273]

[0274] In formula (9), the penalty function p(X sel ) is a decay function, which is 0 if there is no allocation violation.

[0275] γ j is the coefficient, j=j1,j2,…,J.

[0276] The objective function value calculation unit 13 obtains the first allocation result X a Or the second distribution result X b , as the allocation result X selected by the allocation result selection unit 7 sel .

[0277] The objective function value calculation unit 13 allocates the result X sel Provided to the objective function f(X) shown in the following formula (10) sel ) and calculate the objective function f(Xsel ) is the objective function value f.

[0278] f(X sel ) = f assign +ε·f separation (10)

[0279] In formula (10), f assign According to the distribution result X sel Determine the value. If the assignment results in X sel If the allocation time of each aircraft indicated is within the range of the allocation time, the earlier the allocation time is within the range of the allocation time, the higher the allocation time is. assign If the distribution result X sel The allocation time of each aircraft represented by f is the time that cannot be allocated. assign A smaller value such as -1000.

[0280] f separation According to the distribution result X sel If the allocation interval is larger than the minimum interval that can be allocated, the smaller the allocation interval, the smaller the value of f separation If the allocation interval is smaller than the minimum interval that can be allocated, then f separation becomes a smaller value such as -1000.

[0281] ε is the weight coefficient.

[0282] The objective function value calculation unit 13 outputs the objective function value f to the function value addition unit 14 .

[0283] The function value adding unit 14 obtains the penalty value p from the penalty value calculation processing unit 12 , and obtains the target function value f from the target function value calculating unit 13 .

[0284] As shown in the following formula (11), the function value adding unit 14 performs weighted addition of the penalty function p and the objective function value f.

[0285] p'=p+δ·f (11)

[0286] In formula (11), δ is a weight coefficient.

[0287] The function value adding unit 14 outputs the penalty value p′ to which the objective function value is added to the second allocation result acquiring unit 15 .

[0288] When the penalty value p′ is supplied from the penalty value calculation unit 11 , the second allocation result acquisition unit 15 updates the second learning model 15 a so that the penalty value p′ becomes smaller.

[0289] The second allocation result acquisition unit 15 receives the second time schedule information S bWhen the scheduling information S b Provided to the second learning model 15a, the second allocation result X is obtained from the second learning model 15a b .

[0290] The second allocation result acquisition unit 15 obtains the second allocation result X b The data are output to the change cost calculation unit 3, the reward value difference prediction unit 4, and the allocation result selection unit 7, respectively.

[0291] In the above embodiment 3, Fig.15 The allocation result determination device shown is configured to include a penalty value calculation unit 11, which calculates a penalty value for the allocation violation if there is an allocation violation in the allocation result selected by the allocation result selection unit 7. Fig.15 The second allocation result acquisition unit 15 of the allocation result determination device shown in FIG. 1 updates the second learning model 15a so that the penalty value calculated by the penalty value calculation unit 11 becomes smaller. Figure 1 The distribution results shown determine the device compared to Fig.15 The allocation result determination device shown can improve the selection accuracy of the allocation result.

[0292] Furthermore, the present invention can realize free combination of the various embodiments, modification of arbitrary components of the various embodiments, or omission of arbitrary components of the various embodiments.

[0293] Industrial Applicability

[0294] The present invention is suitable for an allocation result determination device and an allocation result determination method.

[0295] Description of symbols

[0296] 1: 1st allocation result acquisition unit; 1a: 1st learning model; 2: 2nd allocation result acquisition unit; 2a: 2nd learning model; 3: change cost calculation unit; 4: reward value difference prediction unit; 5: allocation result difference detection unit; 6: difference prediction processing unit; 6a: 1st prediction processing unit; 6b: 2nd prediction processing unit; 6c: learning model; 6d: difference calculation processing unit; 7: allocation result selection unit; 8: reward value difference calculation unit; 8a: 1st reward value calculation unit; 8b: 2nd reward value calculation unit; 8c: difference calculation processing unit; 9: reward value difference prediction unit; 10: difference prediction processing unit; 10a: 1st prediction processing unit; 10b: 2nd Prediction processing unit; 10c: learning model; 10d: difference calculation processing unit; 11: penalty value calculation unit; 12: penalty value calculation processing unit; 13: objective function value calculation unit; 14: function value addition unit; 15: second allocation result acquisition unit; 15a: second learning model; 21: first allocation result acquisition circuit; 22: second allocation result acquisition circuit; 23: change cost calculation circuit; 24: reward value difference prediction circuit; 27: allocation result selection circuit; 28: reward value difference calculation circuit; 29: reward value difference prediction circuit; 31: penalty value calculation circuit; 35: second allocation result acquisition circuit; 41: memory; 42: processor.

Claims

1. A distribution result determination device, wherein: The allocation result determination device comprises: a change cost calculation unit that obtains a first allocation result determined at a first moment and a second allocation result determined at a second moment that is a moment later than the first moment, and calculates a change cost as an allocation result indicating an allocation order for a plurality of allocation objects, wherein the change cost is an increase in cost when the allocation result is changed from the first allocation result to the second allocation result; as well as An allocation result selection unit selects the first allocation result or the second allocation result based on the change cost calculated by the change cost calculation unit.

2. The allocation result determination device according to claim 1, characterized in that: The allocation result determination device includes a reward value difference prediction unit, which provides the first allocation result and the second allocation result to a learning model for reward value prediction, obtains a first reward value indicating the degree of goodness or badness of the first allocation result and a second reward value indicating the degree of goodness or badness of the second allocation result from the learning model, subtracts the first reward value from the second reward value, and thereby predicts a reward value difference between the first reward value and the second reward value. The distribution result selection unit selects the first distribution result or the second distribution result based on the change cost calculated by the change cost calculation unit and the reward value difference predicted by the reward value difference prediction unit.

3. The allocation result determination device according to claim 2, characterized in that: The allocation result determination device includes the reward value difference prediction unit instead of the change cost calculation unit, The distribution result selection unit selects the first distribution result or the second distribution result based on the reward value difference predicted by the reward value difference prediction unit.

4. The allocation result determination device according to claim 2, characterized in that: The distribution result determination device includes a reward value difference calculation unit, which provides the first distribution result to a reward function to calculate the first reward value, provides the second distribution result to the reward function to calculate the second reward value, and subtracts the first reward value from the second reward value to calculate the reward value difference between the first reward value and the second reward value. The reward value difference prediction unit updates the learning model so that a difference between the predicted reward value difference and the reward value difference calculated by the reward value difference calculation unit becomes smaller.

5. The allocation result determination device according to claim 1, characterized in that: The allocation result determination device comprises: a first allocation result acquisition unit that provides the scheduling information of the plurality of allocation objects at the first time to a first learning model, acquires the first allocation result from the first learning model, and outputs the first allocation result to the change cost calculation unit; as well as A second allocation result acquisition unit provides the scheduling information of the plurality of allocation objects at the second time to a second learning model, acquires the second allocation result from the second learning model, and outputs the second allocation result to the change cost calculation unit.

6. The allocation result determination device according to claim 5, characterized in that: The allocation result determination device includes a penalty value calculation unit, which calculates a penalty value for the allocation violation if there is an allocation violation in the allocation result selected by the allocation result selection unit. The second allocation result acquisition unit updates the second learning model so that the penalty value calculated by the penalty value calculation unit becomes smaller.

7. The allocation result determination device according to claim 6, characterized in that: The penalty value calculation unit provides the allocation result selected by the allocation result selection unit to an objective function to calculate the value of the objective function, and adds the value of the objective function, that is, the objective function value, to the penalty value. The second allocation result acquisition unit updates the second learning model so that the penalty value obtained by adding the objective function value becomes smaller.

8. A method for determining an allocation result, wherein: The change cost calculation unit obtains a first allocation result determined at a first moment and a second allocation result determined at a second moment which is a moment later than the first moment, and calculates a change cost as an allocation result indicating an allocation order for a plurality of allocation objects, wherein the change cost is an increase in cost when the allocation result is changed from the first allocation result to the second allocation result. The allocation result selection unit selects the first allocation result or the second allocation result based on the change cost calculated by the change cost calculation unit.

Images