WO2024185324A1 - Optical path setting device, optical path setting method, and optical path setting program - Google Patents

Abstract

[Problem] To efficiently set an optical signal path. [Solution] This path setting device executes the training of a machine learning model by using: as input, information of a first optical signal path, which includes characteristics of a plurality of first terminals through which, in an optical network, an optical signal passes from a start terminal of the optical signal to a destination terminal of the optical signal, and first log information of the plurality of first terminals; and, as output, the communication quality of the first optical signal path. The path setting device evaluates the communication quality of a plurality of second optical signal paths by inputting, into the machine learning model information of a second optical signal path, which includes: characteristics of a plurality of second terminals, through which the optical signal passes from one terminal to another terminal; and second log information of the plurality of second terminals.

Description

Optical path setting device, optical path setting method, and optical path setting program

The present invention relates to an optical path setting device, an optical path setting method, and an optical path setting program.

In conventional optical networks, multiple network elements (NEs) are interconnected and transmit optical signals via set optical signal paths.

For example, in conventional technology, the communication quality of an optical digital coherent signal path in operation is monitored, and if a fault occurs in the optical signal path in operation due to the effects of lightning or the like, causing a drop in communication quality, the path is switched to another optical signal path that has been set in advance. Furthermore, in conventional technology, when it is detected that the communication quality of the optical signal path before the switch has returned to its original state, the optical signal path in operation is switched from the destination optical signal path to the original optical signal path. For example, conventional technology switching methods include the intensity modulation with direct detection (IMDD) signal transmission method (transmitting signals by varying the strength of light) and the digital coherent signal transmission method (transmitting signals using the phase and polarization of light).

FIG. 11 is a diagram for explaining the conventional technology. The optical network shown in FIG. 11 includes NEs 10, 11, 12, 13, 14, 15, 16, 17, 18, and 19, and transmits optical signals via optical signal paths in operation.

For example, NE10 is mutually connected to NE11, 13, and 15. NE11 is mutually connected to NE10 and 12. NE12 is mutually connected to NE11 and 19. NE13 is mutually connected to NE10, 14, and 15. NE14 is mutually connected to NE13, 17, and 19. NE15 is mutually connected to NE10, 13, and 16. NE16 is mutually connected to NE15 and 17. NE17 is mutually connected to NE14, 16, and 18. NE18 is mutually connected to NE17 and 19. NE19 is mutually connected to NE12, 14, and 18.

The optical signal paths from NE10 to NE19 are designated as optical signal paths A, B, and C. Optical signal path A is an optical signal path that passes through NEs 10, 11, 12, and 19. Optical signal path B is an optical signal path that passes through NEs 10, 15, 13, 14, and 19. Optical signal path C is an optical signal path that passes through NEs 10, 15, 16, 17, 18, and 19.

For example, if lightning 1 occurs between NE12 and NE19 while path A is in operation, the communication quality of path A will deteriorate and the optical signal path in operation will be switched from path A to path B. After that, when lightning 1 subsides and the communication quality of path A returns to normal, the optical signal path in operation will be switched from path B back to the original path A.

JP 2007-96796 A Japanese Patent Application Publication No. 5-114899 JP 2011-145846 A JP 2019-153893 A

However, the above-mentioned conventional technology has the problem that it is not possible to set up optical signal paths efficiently.

FIGS. 12 and 13 are diagrams for explaining the problems with the conventional technology. The optical network shown in FIG. 12 includes NEs 10 to 19. The optical signal paths from NE 10 to NE 19 are designated as optical signal paths A, B, and C. The explanation of optical signal paths A to C is the same as the explanation of optical signal paths A to C explained in FIG. 11.

Here, we assume that as the linear precipitation band moves, the weather transitions as follows: At time T1, lightning 1 occurs between NE12 and NE19. At time T2, lightning 1 subsides, lightning 2a occurs between NE11 and NE12, and lightning 2b occurs between NE13 and NE14. At time T3, lightning 2a and 2b subside, and lightning 3 occurs between NE13 and NE15.

Figure 13 shows the relationship between communication quality and time for optical signal paths A to C. The vertical axis corresponds to communication quality, and the horizontal axis corresponds to time. Line segment 20a shows the relationship between communication quality and time for optical signal path A. Line segment 20b shows the relationship between communication quality and time for optical signal path B. Line segment 20c shows the relationship between communication quality and time for optical signal path C. Note that Figure 13 is schematic, as signal quality can be determined when optical/electrical (O/E) conversion is performed and the signal is regenerated.

For example, if the linear precipitation band moves as described above and lightning occurs at each time, the communication quality of multiple optical signal paths A and B will decrease in a chain reaction, and if such optical signal paths A and B are set as the switching destination and switching source, the optical signal paths cannot be set efficiently.

In the example shown in FIG. 13, optical signal path A is selected as the optical signal path to be operated until time T1 when lightning 1 occurs. Because the communication quality of optical signal path A deteriorates due to the effects of lightning 1, optical signal path B, which was set in advance, is selected as the optical signal path to be operated.

If lightning 2b occurs at time T2 after optical signal path B is selected, the communication quality of optical signal path B will deteriorate due to the effects of lightning 2b, and since the communication quality of optical signal path A has been restored just before time T2, a switchback of optical signal path A will occur. However, lightning 2a occurs at time T2, causing the communication quality of optical signal path A to deteriorate, and the switchback will fail.

While the switch back to optical signal path A fails, the communication quality of optical signal path B is restored, and optical signal transmission is resumed via optical signal path B. At time T3, lightning 3 occurs, the communication quality of optical signal path B decreases, and optical signal path A is selected as the switching destination, but at time T3, the communication quality of optical signal path A has not been restored, and ultimately optical signal path C is selected and optical signal transmission begins.

Of the two switching methods, intensity modulation direct detection (IMDD) signal transmission and digital coherent signal transmission, it is difficult to monitor signal quality with the latter digital coherent method by monitoring only optical power; it is necessary to monitor the state of polarization and phase shift, and there is no way to monitor the SOP as light (without photoelectric conversion) for individual wavelengths multiplexed in the transmission path. In the case of IMDD, signal quality can be monitored by conventional optical power monitoring, but if signal degradation occurs for reasons that cannot be detected by optical power measurement alone (fluctuations in the polarization state), it is not possible to identify the location where the signal quality degradation is occurring, and appropriate switching cannot be performed.

As described above, in conventional technology, the location where signal quality degradation is occurring cannot be identified, resulting in unnecessary switching between optical signal path A and optical signal path B, and there is a risk of intermittent communication failures. Therefore, it is necessary to efficiently set the optimal optical signal path. For example, if optical signal path C, which has stable communication quality, can be selected at time T1, unnecessary switching can be prevented.

In one aspect, the present invention aims to provide an optical path setting device, an optical path setting method, and an optical path setting program that can efficiently set optical signal paths.

In the first proposal, the optical path setting device has a control unit that executes the following process. The control unit receives as input information on a first optical signal path, including characteristics of multiple first terminals that are passed through from a start terminal of an optical signal included in an optical network to a destination terminal of the optical signal, and first log information of the multiple first terminals, and executes machine learning of a machine learning model with the communication quality of the first optical signal path as output. The control unit inputs information on multiple second optical signal paths, including characteristics of multiple second terminals that are passed through from one terminal to another terminal, and second log information of the multiple second terminals, into the machine learning model, and evaluates the communication quality of the multiple second optical signal paths.

Optical signal paths can be set up efficiently.

FIG. 1 is a diagram illustrating an example of a system according to the present embodiment. FIG. 2 is a diagram for explaining the process of the learning phase. FIG. 3 is a diagram illustrating an example of a data structure of span information. FIG. 4 is a diagram (1) for explaining the processing in the inference phase. FIG. 5 is a diagram (2) for explaining the processing in the inference phase. FIG. 6 is a functional block diagram showing the configuration of a path setting device according to the present embodiment. FIG. 7 is a flowchart showing the processing steps of the path setting device in the learning phase. FIG. 8 is a flowchart showing the processing steps of the path setting device in the inference phase. FIG. 9 is a diagram showing an example of a teacher data table in which the signal load ratio is set. FIG. 10 is a diagram showing an example of a hardware configuration of a computer that realizes the same functions as the path setting device of this embodiment. FIG. 11 is a diagram for explaining the prior art. FIG. 12 is a diagram (1) for explaining the problem of the conventional technology. FIG. 13 is a diagram (2) for explaining the problem of the conventional technology.

Below, examples of the optical path setting device, optical path setting method, and optical path setting program disclosed in the present application are described in detail with reference to the drawings. Note that the present invention is not limited to these examples.

An example of a system according to this embodiment will now be described. FIG. 1 is a diagram showing an example of a system according to this embodiment. As shown in FIG. 1, this system has NEs 10, 11, 12, 13, 14, 15, 16, 17, 18, and 19, and a path setting device 100. The path setting device 100 is an example of an optical path setting device.

NEs 10 to 19 and the path setting device 100 are connected to each other via an operation monitoring network 50. Although NEs 10 to 19 are shown in FIG. 1, other NEs may also be included.

NEs 10 to 19 are devices that transmit optical signals via optical signal paths set by path setting device 100. For example, when optical signal path A described in FIG. 11 is set by path setting device 100, an optical signal is transmitted from NE 10 to NE 19 via optical signal path A via NEs 11 and 12.

For example, NE10 is mutually connected to NE11, 13, and 15. NE11 is mutually connected to NE10 and 12. NE12 is mutually connected to NE11 and 19. NE13 is mutually connected to NE10, 14, and 15. NE14 is mutually connected to NE13, 17, and 19. NE15 is mutually connected to NE10, 13, and 16. NE16 is mutually connected to NE15 and 17. NE17 is mutually connected to NE14, 16, and 18. NE18 is mutually connected to NE17 and 19. NE19 is mutually connected to NE12, 14, and 18.

NEs 10 to 19 save log information each time they execute various processes, and transmit the log information to the path setting device 100 in response to a request from the path setting device 100.

The path setting device 100 uses a trained machine learning model to select an optical signal path that can maintain stable communication quality from among multiple optical signal paths from the NE that is the source of the optical signal to the NE that is the destination of the optical signal. The path setting device 100 sets the selected optical signal path for NEs 10 to 19.

For example, the path setting device 100 executes a learning phase process to train a machine learning model, and an estimation phase process to use the trained machine learning model to estimate an optical signal path capable of maintaining stable communication quality. The learning phase process and the inference phase process of the path setting device 100 will be described below in order.

First, an example of the learning phase processing executed by the path setting device 100 will be described. FIG. 2 is a diagram for explaining the learning phase processing. For example, the path setting device 100 executes machine learning of the machine learning model M1 using the teacher data table 141. The machine learning model M1 is a Neural Network (NN) or the like.

For example, the teacher data table 141 associates path identification information, input data, and labels. The path identification information is information that identifies an optical signal path. A plurality of pieces of span information are set in the input data. A piece of span information indicates information about a pair of adjacent NEs among a plurality of NEs included in a certain optical signal path.

For example, span information A-1 corresponding to optical signal path A and the first span is information about NEs 10 and 11 out of NEs 10, 11, 12, and 19 included in optical signal path A. Span information A-2 corresponding to optical signal path A and the second span is information about NEs 11 and 12. Span information A-3 corresponding to optical signal path A and the third span is information about NEs 12 and 19.

Here, we will explain the span information. Figure 3 is a diagram showing an example of the data structure of span information. As shown in Figure 3, transmission path information is set for each piece of span information. For example, the transmission path information includes source NE, source NE type, destination NE, destination NE type, fiber type, distance between NEs, span loss, log, and lightning strike flag. The transmission path information may also include other information.

The source NE indicates the NE that is the source of the span information among the set of NEs. The source NE type is set to ROADM (reconfigurable optical add/drop multiplexer) or ILA (In Line Amp) as the type of the source NE. The destination NE indicates the NE that is the destination of the span information among the set of NEs. The destination NE type is set to ROADM or ILA as the type of the destination NE.

The fiber type is set to the type of optical fiber used between the NEs of the span information. The distance between NEs is set to the distance between the NEs of the span information. The span loss indicates the amount of attenuation of the optical signal when it is transmitted between the NEs of the span information. The log information is set to a value obtained by converting the log information of the NE that sent the span information and the log information of the NE that sent the span information into a vector.

The lightning strike flag is information that indicates whether or not a lightning strike occurred between the NEs in the span information during the time period in which the log information was collected. If a lightning strike occurred, the lightning strike flag is "on." On the other hand, if no lightning strike occurred, the lightning strike flag is "off."

Returning to the explanation of the teacher data table 141 in Figure 2, the signal type included in the input data is the type of signal used in the corresponding optical signal path. The label is a value (correct label) indicating the communication quality of the corresponding optical signal path. A larger communication quality value indicates better communication quality.

As explained in FIG. 3, even for the same optical signal path (path identification information), the log information collected from each NE and the lightning strike flag during the time period in which the log information was collected may change. For this reason, multiple input data and labels may be set for the same optical signal path (path identification information) in the teacher data table 141.

The path setting device 100 performs machine learning on the machine learning model M1 based on the backpropagation method so that the difference between the output result of the machine learning model M1 and the label when input data is input to the machine learning model M1 is reduced (trains the machine learning model M1).

For example, the path setting device 100 inputs input data in1 (span information A-1, A-2, A-3, ..., signal type) to the machine learning model M1, and updates the parameters of the machine learning model M1 so that the difference between the output of the machine learning model M1 and the label l1 becomes smaller.

The path setting device 100 inputs the input data in2 (span information B-1, B-2, B-3, ..., signal type) to the machine learning model M1, and updates the parameters of the machine learning model M1 so that the difference between the output of the machine learning model M1 and the label l2 becomes smaller.

The path setting device 100 inputs the input data in3 (span information C-1, C-2, C-3, ..., signal type) to the machine learning model M1, and updates the parameters of the machine learning model M1 so that the difference between the output of the machine learning model M1 and the label l3 becomes smaller.

The path setting device 100 inputs the input data in4 (span information D-1, D-2, D-3, ..., signal type) to the machine learning model M1, and updates the parameters of the machine learning model M1 so that the difference between the output of the machine learning model M1 and the label l4 becomes smaller.

The path setting device 100 updates the parameters of the machine learning model M1 by repeatedly executing the above process for other input data and label sets stored in the teacher data table 141.

Next, an example of the inference phase processing executed by the path setting device 100 will be described. Figures 4 and 5 are diagrams for explaining the inference phase processing. First, Figure 4 will be described. When executing the inference phase processing, the path setting device 100 refers to the network information 142. In the network information 142, the optical signal path from the NE that is the start point of the transmission of the optical signal to the NE that is the end point, and information on the NEs included in the optical signal path are set.

For example, in the network information 142, information on optical signal paths A, B, and C is set as optical signal paths from NE10, which is the starting point, to NE19, which is the end point. The information on optical signal paths A, B, and C includes identification information of the NEs included in each optical signal path, the NE type, the fiber type, the distance between NEs, span loss, etc. Furthermore, the NE information may include information identifying the area in which the NE is installed.

In FIG. 4, a case is described in which the path setting device 100 selects an optical signal path that can maintain stable communication quality from among optical signal paths A, B, and C that lead to NE10 to NE19.

The path setting device 100 generates input data 30A corresponding to optical signal path A by executing the following process. The input data 30A includes span information A'-1, span information A'-2, span information A'-3, and optical signal type.

Span information A'-1 is information about NEs 10 and 11. Span information A'-2 is information about NEs 11 and 12. Span information A'-3 is information about NEs 12 and 19.

This sets the process when the path setting device 100 sets transmission path information in span information A'-1. The path setting device 100 acquires and sets the sending NE, source NE type, destination NE, destination NE type, fiber type, NE distance, and span loss for NE10 and 11 based on the network information 142. For example, the sending NE is "NE10", the source NE type is "ROADM", the destination NE is "NE11", the destination NE type is "ILA", the fiber type is "SMF", the NE distance is "5.8 km", and the span loss is "2.9 dB".

The path setting device 100 communicates with NEs 10 and 11 associated with span information A'-1, and acquires log information and signal type from NEs 10 and 11. The path setting device 100 calculates a vector of the log information based on the log vector dictionary 143, and sets the vector of the log information to span information A'-1. The path setting device 100 sets the signal type to input data 30A.

Here, the log vector dictionary 143 is a dictionary that associates multiple morphemes that may be included in the log information acquired from the NE with the vectors of the relevant morphemes. For example, when the path setting device 100 acquires log information from NEs 10 and 11, it performs morphological analysis on the acquired log information to decompose it into multiple morphemes. The path setting device 100 compares each decomposed morpheme with the log vector dictionary 143 to identify the vector of each morpheme, and calculates the vector of the log information by accumulating the vectors of each identified morpheme.

The path setting device 100 uses the transmission line environment information 144 to determine whether or not lightning is occurring within a specified area based on NEs 10 and 11 in the span information A'-1 during the time period in which the log information was acquired. The transmission line environment information 144 includes information about the weather for each region and each hour, such as information about whether or not lightning is occurring in a certain region at a certain time. If lightning is occurring within the specified area based on NEs 10 and 11, the path setting device 100 sets the lightning flag in the span information A'-1 to "on." On the other hand, if lightning is not occurring within the specified area based on NEs 10 and 11, the path setting device 100 sets the lightning flag to "off."

By performing the above process, the path setting device 100 sets transmission path information in span information A'-1. The path setting device 100 also sets transmission path information for span information A'-2 and span information A'-3 by performing the same process as for span information A'-1. As a result, input data 30A is generated.

The path setting device 100 generates input data 30B corresponding to optical signal path B. The input data 30B includes span information B'-1, span information B'-2, span information B'-3, and optical signal type. The process by which the path setting device 100 sets transmission line information corresponding to span information B'-1 to B'-3 is similar to the process by which transmission line information is set in span information A'-1.

The path setting device 100 generates input data 30C corresponding to the optical signal path C. The input data 30C includes span information C'-1, span information C'-2, span information C'-3, and an optical signal type. The process by which the path setting device 100 sets the transmission line information corresponding to the span information C'-1 to C'-3 is similar to the process by which the transmission line information is set in the span information C'-1.

Now, let us move on to the explanation of FIG. 5. The path setting device 100 calculates a communication quality score 40A by inputting input data 30A into the machine learning model M1. The path setting device 100 calculates a communication quality score 40B by inputting input data 30B into the machine learning model M1. The path setting device 100 calculates a communication quality score 40C by inputting input data 30C into the machine learning model M1.

The path setting device 100 selects the optical signal path corresponding to the maximum score among the communication quality scores 40A to 40C. For example, if the communication quality score 40C is the maximum score among the communication quality scores 40A to 40C, the path setting device 100 selects the optical signal path C corresponding to the communication quality score 40C.

The path setting device 100 controls the network's NEs 10 to 19 so that the optical signal is transmitted via the selected optical signal path C.

As described above, the path setting device 100 according to this embodiment executes machine learning of the machine learning model M1 based on the relationship between the input data of the optical signal paths stored in the teacher data table 141 and the communication quality labels. The path setting device 100 inputs input data of a plurality of optical signal paths to be evaluated to the trained machine learning model M1, calculates the communication quality score of each optical signal path, selects the optical signal path with the maximum communication quality score, and sets it in the optical network. The input data of the optical signal path includes vectors of NE log information and meteorological information, so that it is possible to prevent the selection of an optical signal path whose communication quality becomes intermittently unstable due to weather, and to set the optical signal path efficiently.

Next, an example of the configuration of a path setting device 100 that executes the process shown in FIG. 5 will be described. FIG. 6 is a functional block diagram showing the configuration of a path setting device according to this embodiment. As shown in FIG. 6, this path setting device 100 has a communication unit 110, an input unit 120, a display unit 130, a storage unit 140, and a control unit 150.

The communication unit 110 performs data communication (packet communication) with each NE on the operation monitoring network 50. The communication unit 110 is a NIC (Network Interface Card) or the like.

The input unit 120 is an input device that inputs various information to the control unit 150 of the path setting device 100. For example, the input unit 120 corresponds to a keyboard, a mouse, a touch panel, etc.

The display unit 130 is a display device that displays information output from the control unit 150.

The storage unit 140 has a teacher data table 141, network information 142, a log vector dictionary 143, transmission path environment information 144, and a machine learning model M1. The storage unit 140 is a storage device such as a memory.

The teacher data table 141 holds information used when performing machine learning on the machine learning model M1. As described in FIG. 2, the teacher data table 141 associates path identification information, input data, and labels (communication quality). Other explanations regarding the teacher data table 141 are the same as those described in FIG. 2.

The network information 142 includes information about the optical signal path from the NE that is the start point of the transmission of the optical signal to the NE that is the end point, and the NEs included in the optical signal path. The explanation of the network information 142 is the same as the explanation of the network information 142 explained in FIG. 4.

The log vector dictionary 143 is a dictionary that associates multiple morphemes that may be included in the log information obtained from the NE with the vectors of the morphemes. The explanation of the log vector dictionary 143 is the same as the explanation of the log vector dictionary 143 explained in FIG. 4.

The transmission path environment information 144 includes information about the weather for each region and each hour, for example, information about whether or not lightning is occurring. The explanation of the transmission path environment information 144 is the same as the explanation of the transmission path environment information 144 explained in FIG. 4.

The machine learning model M1 is a model that outputs a communication quality score of an optical signal path when input data (information on each span included in the optical signal path) is input. The machine learning model is, for example, NN.

We now move on to the explanation of the control unit 150 in FIG. 6. The control unit 150 has an acquisition unit 151, a learning processing unit 152, an estimation processing unit 153, and a setting unit 154. The control unit 150 is a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc.

The acquisition unit 151 communicates with external devices etc. via the communication unit 110, acquires the teacher data table 141, network information 142, log vector dictionary 143, and transmission path environment information 144, and registers them in the storage unit 140.

The acquisition unit 151 may acquire update information of the teacher data table 141, the network information 142, the log vector dictionary 143, and the transmission path environment information 144. When the acquisition unit 151 acquires update information, it updates the teacher data table 141, the network information 142, the log vector dictionary 143, and the transmission path environment information 144.

The learning processing unit 152 executes machine learning of the machine learning model M1 based on the teacher data table 141. For example, the learning processing unit 152 inputs the input data stored in the teacher data table 141 to the machine learning model M1 based on the backpropagation method, and updates the parameters of the machine learning model M1 so that the difference between the output result of the machine learning model M1 and the label is reduced.

The processing performed by the learning processing unit 152 is similar to the processing in the learning phase described in Figures 2 and 3.

The estimation processing unit 153 uses the trained machine learning model M1 to evaluate the communication quality of multiple optical signal paths from the NE that is the start point of the transmission of an optical signal to the NE that is the end point. For example, the estimation processing unit 153 generates input data corresponding to each of the multiple optical signal paths from the NE that is the start point to the NE that is the end point, and inputs each input data to the machine learning model M1 to calculate a communication quality score for each. The estimation processing unit 153 selects the optical signal path that has the maximum communication quality score from among the multiple optical signal paths.

For example, the estimation processing unit 153 executes the following process to generate input data. The estimation processing unit 153 receives designation of a start point NE and an end point NE from the input unit 120. The estimation processing unit 153 compares the pair of the start point NE and the end point NE with the network information 142, and acquires multiple optical signal paths from the start point NE to the end point NE, and information on the NEs included in these optical signal paths.

The estimation processing unit 153 performs data communication with each NE included in the optical signal path via the communication unit 110, and acquires log information from each NE included in the optical signal path. The estimation processing unit 153 performs morphological analysis on the log information to decompose it into a plurality of morphemes. The estimation processing unit 153 compares each decomposed morpheme with the log vector dictionary 143 to identify the vector of each morpheme, and calculates the vector of the log information by accumulating the vector of each identified morpheme.

The estimation processing unit 153 determines whether or not lightning is currently occurring between the relevant NEs based on the current time, the area in which each NE is installed, and the transmission line environment information. The estimation processing unit 153 may obtain time information from a server that manages time on the network. If lightning is occurring, the estimation processing unit 153 sets the lightning strike flag in the transmission line information to "on," and if lightning is not occurring, the estimation processing unit 153 sets the lightning strike flag in the transmission line information to "off."

By executing the above processing, the estimation processing unit 153 generates transmission path information corresponding to each optical signal path and sets it in the input data. The process by which the estimation processing unit 153 sets the transmission path information and sets it in the input data corresponds to the process described in FIG. 4. Furthermore, the process by which the estimation processing unit 153 inputs the input data to the machine learning model M1 and calculates the communication quality score corresponds to the process described in FIG. 5.

The estimation processing unit 153 outputs the pair of the starting NE and the ending NE, and information on the selected optical signal path to the setting unit 154.

When the setting unit 154 receives information on a pair of a start point NE and an end point NE and a selected optical signal path, it executes communication data communication with each NE on the operation monitoring network 50 and sets a transmission route so that an optical signal is transmitted via the corresponding optical signal path.

Next, an example of the processing procedure of the path setting device according to this embodiment will be described. FIG. 7 is a flowchart showing the processing procedure of the path setting device in the learning phase. As shown in FIG. 7, the learning processing unit 152 of the path setting device 100 acquires a pair of input data and a label from the teacher data table 141 (step S101).

The learning processing unit 152 inputs the input data to the machine learning model M1 (step S102). The learning processing unit 152 calculates the difference between the output of the machine learning model M1 and the label (step S103).

The learning processing unit 152 updates the parameters of the machine learning model M1 so as to reduce the difference (step S104). If the learning processing unit 152 continues the machine learning (step S105, Yes), it proceeds to step S101. On the other hand, if the learning processing unit 152 does not continue the machine learning (step S105, No), it ends the process.

FIG. 8 is a flowchart showing the processing steps of the path setting device in the inference phase. As shown in FIG. 8, the estimation processing unit 153 of the path setting device 100 receives the specification of the NE that is the start point and the NE that is the end point of the optical signal (step S201).

The estimation processing unit 153 identifies multiple optical signal paths and information on the NEs included in the optical signal paths based on the starting NE and the ending NE and the network information 142 (step S202).

The estimation processing unit 153 selects an optical signal path (step S203). The estimation processing unit 153 acquires log information from each NE included in the optical path (step S204). The estimation processing unit 153 calculates a vector of the log information based on the log information and the log vector dictionary 143 (step S205).

The estimation processing unit 153 determines whether the lightning strike flag is "on" or not based on the current time, the area in which the NE is installed, and the transmission line environment information 144 (step S206). The estimation processing unit 153 sets the transmission line information corresponding to the input data of the optical signal path (step S207).

If the estimation processing unit 153 has not selected all optical signal paths (step S208, No), it proceeds to step S203. On the other hand, if the estimation processing unit 153 has selected all optical signal paths (step S208, Yes), it proceeds to step S209.

The estimation processing unit 153 inputs the input data corresponding to each optical signal path into the machine learning model M1 to calculate the communication quality score of each optical signal path (step S209).

The estimation processing unit 153 selects the optical signal path with the highest communication quality score from among the multiple optical signal paths (step S210). The setting unit 154 of the path setting device 100 sets the optical signal path selected by the estimation processing unit 153 on the optical network (step S211).

Next, the effects of the path setting device 100 according to this embodiment will be described. The path setting device 100 executes machine learning of the machine learning model M1 based on the relationship between the input data of the optical signal paths stored in the teacher data table 141 and the communication quality labels. The path setting device 100 inputs the input data of the multiple optical signal paths to be evaluated to the trained machine learning model M1, thereby calculating the communication quality score of each optical signal path and evaluating each optical signal path. By using such evaluation results, optical signal paths can be set efficiently.

The path setting device 100 selects the optical signal path with the highest communication quality score from among multiple optical signal paths, and sets it in the optical network. The input data for the optical signal path includes vectors of NE log information and weather information, so that it is possible to prevent the selection of an optical signal path whose communication quality becomes intermittently unstable due to weather such as lightning or earthquakes, and to set the optical signal path efficiently.

For example, as described in FIG. 12, in the conventional technology, unnecessary switching occurs between optical signal path A and optical signal path B. In response to this, the path setting device 100 inputs the input data of optical signal paths A, B, and C into the machine learning model M1, thereby enabling the path setting device 100 to preferentially select optical signal path C, which has a high communication evaluation score, and thus prevents unnecessary switching from occurring.

Incidentally, the path setting device 100 in the above embodiment sets multiple span information and signal types as input data corresponding to an optical signal path, but this is not limited to this. For example, the path setting device 100 may set a "signal load factor" in addition to multiple span information and signal types. For example, when a large-scale event occurs near an NE of a signal path, the signal load factor tends to increase, and the communication quality of the signal path may deteriorate.

FIG. 9 is a diagram showing an example of a teacher data table in which a signal load rate is set. As shown in FIG. 9, the input data of such teacher data table 241 has a signal load rate set in addition to the teacher data table 141 described in FIG. 2. By executing machine learning of the machine learning model M1 based on the teacher data table 241, the path setting device 100 can select an optical signal path with good communication quality, taking into account the impact of a large-scale event.

In addition, in the estimation phase, the path setting device 100 may set the signal load rate of the input data based on the event scale for each region and the signal load rate according to the event scale, and input the input data to the machine learning model M1 to calculate the communication quality score.

In addition, the communication quality of the optical signal path may be affected by earthquakes, etc. For this reason, in addition to the lightning strike flag set in the span information, a vibration flag may also be set. If an earthquake occurs in the area where the corresponding NE is installed, the vibration flag is set to "on," and if no earthquake has occurred, the vibration flag is set to "off."

Next, an example of the hardware configuration of a computer that realizes the same functions as the path setting device 100 described above will be described. Figure 10 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the path setting device of this embodiment.

As shown in FIG. 10, computer 300 has a CPU 301 that executes various types of arithmetic processing, an input device 302 that accepts data input from the user, and a display 303. Computer 300 also has a communication device 304 that transmits and receives data to and from external devices, etc., via a wired or wireless network, and an interface device 305. Computer 300 also has a RAM 306 that temporarily stores various types of information, and a hard disk device 307. Each of devices 301 to 307 is connected to a bus 308.

The hard disk device 307 has an acquisition program 307a, a learning processing program 307b, an estimation processing program 307c, and a setting program 307d. The CPU 301 reads each of the programs 307a to 307d and loads them into the RAM 306.

The acquisition program 307a functions as an acquisition process 306a. The learning process program 307b functions as a learning process 306b. The inference process program 307c functions as an inference process 306c. The setting program 307d functions as a setting process 306d.

The processing of the acquisition process 306a corresponds to the processing of the acquisition unit 151. The processing of the learning processing process 306b corresponds to the processing of the learning processing unit 152. The processing of the estimation processing process 306c corresponds to the processing of the estimation processing unit 153. The processing of the setting process 306d corresponds to the processing of the setting unit 154.

Note that each of the programs 307a to 307d does not necessarily have to be stored in the hard disk device 307 from the beginning. For example, each program may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 300. Then, the computer 300 may read and execute each of the programs 307a to 307d.

The following notes are further provided with respect to the embodiments including the above examples.

(Supplementary Note 1) A machine learning model is executed using information on a first optical signal path, including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the plurality of second optical signal paths pass from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting device having a control unit for executing processing.

(Appendix 2) The optical path setting device described in appendix 1, characterized in that the information on the first optical signal path further includes first weather information for the time period during which the first log information was acquired, and the information on the multiple second optical signal paths further includes second weather information for the time period during which the second log information was acquired.

(Appendix 3) The optical path setting device described in appendix 1 or 2, characterized in that the evaluation process generates information on multiple second optical signal paths with different routes from the one terminal to another terminal, and inputs each of the multiple second optical signal paths into the machine learning model to evaluate the communication quality of each of the multiple second optical signal paths.

(Appendix 4) The optical path setting device described in Appendix 3, further comprising a process of selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process, and setting the selected second optical signal path in the optical network.

(Appendix 5) The optical path setting device described in appendix 1 or 2 further executes a process of calculating vectors of the first log information and the second log information, the process of executing the machine learning sets the vector of the first log information to the information of the first optical signal path, and the process of evaluating sets the vector of the second log information to the information of the second optical signal path.

(Appendix 6) The optical path setting device described in appendix 2 is characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information.

(Appendix 7) The optical path setting device described in appendix 1 or 2, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path.

(Supplementary Note 8) A machine learning model is executed using information of a first optical signal path, including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the plurality of second optical signal paths pass from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting method, the processing of which is executed by a computer.

(Appendix 9) The optical path setting method described in appendix 8, characterized in that the information on the first optical signal path further includes first weather information for the time period during which the first log information was acquired, and the information on the multiple second optical signal paths further includes second weather information for the time period during which the second log information was acquired.

(Appendix 10) The optical path setting method described in appendix 8 or 9, characterized in that the evaluation process generates information on multiple second optical signal paths with different routes from the one terminal to another terminal, and inputs each of the multiple second optical signal paths into the machine learning model to evaluate the communication quality of each of the multiple second optical signal paths.

(Appendix 11) The optical path setting method described in appendix 10, further comprising the steps of: selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process; and setting the selected second optical signal path in the optical network.

(Appendix 12) The optical path setting method described in appendix 8 or 9, further comprising the step of calculating vectors of the first log information and the second log information, the step of executing the machine learning setting the vector of the first log information to the information of the first optical signal path, and the step of evaluating setting the vector of the second log information to the information of the second optical signal path.

(Appendix 13) The optical path setting method described in appendix 9, characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information.

(Appendix 14) The optical path setting method described in appendix 8 or 9, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path.

(Supplementary Note 15) A machine learning model is executed using information of a first optical signal path including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the optical signal path passes from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting program that causes a computer to execute a process.

(Appendix 16) The optical path setting program described in appendix 15, characterized in that the information on the first optical signal path further includes first weather information for the time period during which the first log information was acquired, and the information on the multiple second optical signal paths further includes second weather information for the time period during which the second log information was acquired.

(Appendix 17) The optical path setting program described in appendix 15 or 16, characterized in that the evaluation process generates information on multiple second optical signal paths with different routes from the one terminal to another terminal, and inputs each of the multiple second optical signal paths into the machine learning model to evaluate the communication quality of each of the multiple second optical signal paths.

(Appendix 18) The optical path setting program described in appendix 17, further comprising a process of selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process, and setting the selected second optical signal path in the optical network.

(Appendix 19) The optical path setting program described in appendix 15 or 16, further comprising a process of calculating vectors of the first log information and the second log information, the process of executing the machine learning setting the vector of the first log information to the information of the first optical signal path, and the process of evaluating setting the vector of the second log information to the information of the second optical signal path.

(Appendix 20) The optical path setting program described in appendix 16, characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information.

(Appendix 21) The optical path setting program described in appendix 15 or 16, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path.

REFERENCE SIGNS LIST 100 Path setting device 110 Communication unit 120 Input unit 130 Display unit 140 Storage unit 141 Teacher data table 142 Network information 143 Log vector dictionary 144 Transmission path environment information 150 Control unit 151 Acquisition unit 152 Learning processing unit 153 Estimation processing unit 154 Setting unit

Claims (21)

execute machine learning of a machine learning model using information of a first optical signal path, the information including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the plurality of second optical signal paths pass from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting device having a control unit for executing processing. The optical path setting device according to claim 1, characterized in that the information of the first optical signal path further includes first weather information for the time period during which the first log information was acquired, and the information of the plurality of second optical signal paths further includes second weather information for the time period during which the second log information was acquired. The optical path setting device according to claim 1 or 2, characterized in that the evaluation process generates information on a plurality of second optical signal paths having different routes from the one terminal to another terminal, and inputs each of the plurality of second optical signal paths into the machine learning model, thereby evaluating the communication quality of each of the plurality of second optical signal paths. The optical path setting device according to claim 3, further comprising a process of selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process, and setting the selected second optical signal path in the optical network. The optical path setting device according to claim 1 or 2, further comprising a process for calculating vectors of the first log information and the second log information, the process for executing the machine learning sets the vector of the first log information to the information of the first optical signal path, and the process for evaluating sets the vector of the second log information to the information of the second optical signal path. The optical path setting device according to claim 2, characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information. The optical path setting device according to claim 1 or 2, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path. execute machine learning of a machine learning model using information of a first optical signal path, the information including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the plurality of second optical signal paths pass from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting method, the processing of which is executed by a computer. The optical path setting method according to claim 8, characterized in that the information of the first optical signal path further includes first weather information for a time period during which the first log information is acquired, and the information of the second optical signal paths further includes second weather information for a time period during which the second log information is acquired. The optical path setting method according to claim 8 or 9, characterized in that the evaluation process generates information on a plurality of second optical signal paths having different routes from the one terminal to another terminal, and inputs each of the plurality of second optical signal paths into the machine learning model, thereby evaluating the communication quality of each of the plurality of second optical signal paths. The optical path setting method according to claim 10, further comprising the steps of: selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process; and setting the selected second optical signal path in the optical network. The optical path setting method according to claim 8 or 9, further comprising the step of calculating vectors of the first log information and the second log information, the step of executing the machine learning setting the vector of the first log information to the information of the first optical signal path, and the step of evaluating setting the vector of the second log information to the information of the second optical signal path. The optical path setting method according to claim 9, characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information. The optical path setting method according to claim 8 or 9, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path. execute machine learning of a machine learning model using information of a first optical signal path, the information including characteristics of a plurality of first terminals through which an optical signal included in an optical network passes from a start terminal of the optical signal to a destination terminal of the optical signal and first log information of the plurality of first terminals as an input, and using communication quality of the first optical signal path as an output;
inputting information on a plurality of second optical signal paths, including characteristics of a plurality of second terminals through which the optical signal path passes from a certain terminal to another terminal, and second log information of the plurality of second terminals, into the machine learning model, and evaluating communication quality of the plurality of second optical signal paths;
An optical path setting program that causes a computer to execute a process. The optical path setting program according to claim 15, characterized in that the information on the first optical signal path further includes first weather information for the time period during which the first log information was acquired, and the information on the second optical signal paths further includes second weather information for the time period during which the second log information was acquired. The optical path setting program according to claim 15 or 16, characterized in that the evaluation process generates information on a plurality of second optical signal paths having different routes from the one terminal to another terminal, and inputs each of the plurality of second optical signal paths into the machine learning model, thereby evaluating the communication quality of each of the plurality of second optical signal paths. The optical path setting program according to claim 17, further comprising a process of selecting one second optical signal path from the plurality of second optical signal paths based on the evaluation result of the evaluation process, and setting the selected second optical signal path in the optical network. The optical path setting program according to claim 15 or 16, further comprising a process of calculating vectors of the first log information and the second log information, the process of executing the machine learning setting the vector of the first log information to the information of the first optical signal path, and the process of evaluating setting the vector of the second log information to the information of the second optical signal path. The optical path setting program according to claim 16, characterized in that the process of executing the machine learning sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the first weather information, and the process of evaluating sets information on whether or not lightning has occurred or whether or not an earthquake has occurred as the second weather information. The optical path setting program according to claim 15 or 16, characterized in that the process of executing the machine learning further sets information on a signal load rate to the information on the first optical signal path, and the process of evaluating further sets information on a signal load rate to the information on the second optical signal path.