Controller, method for reducing learning cost, and program

Abstract

To reduce the cost of retraining AI models after recovering from an abnormality.SOLUTION: A controller includes a collection unit that collects PM data representing the optical power level of each of optical signals of a plurality of optical transmission devices, a plurality of AI models that are provided for each of the plurality of optical transmission devices and use the PM data of the optical transmission devices as learning data to learn time-series fluctuations in the PM data of the optical transmission devices, a database that holds an average value and an allowable error of the PM data of the optical transmission devices for each of the plurality of optical transmission devices, an abnormality detection unit that uses the plurality of AI models to detect abnormalities in an optical transmission network in which the plurality of optical transmission devices are provided, and a re-learning determination unit that determines that re-learning of the AI model of an optical transmission device is necessary when the PM data of the optical transmission device after recovery from an abnormality falls outside a threshold range derived from the average value and the allowable error.SELECTED DRAWING: Figure 10

Description

Technical Field

[0001] The present disclosure relates to a controller, a method for reducing learning cost, and a program.

Background Art

[0002] A CSP (Communication Service Provider) such as a communication carrier performs operation and maintenance services for a network to be managed. However, as the network becomes larger and more complex, network changes due to construction and failures occur frequently. Therefore, it has been difficult to automate the operation and maintenance of the network, and it has been manually performed until now.

[0003] However, when performing the operation and maintenance of the network manually, services that span multiple departments managing the network, such as determining the degree of impact on services at the time of a failure and identifying the location of the cause of the failure, occur on demand. This is one of the reasons for the increase in OPEX (Operating Expense).

[0004] On the other hand, in recent years, the main business of CSPs has been shifting from communication business to non-communication business, and it is expected that the suppression of capital investment in the communication business will continue in the future. Therefore, reduction of CAPEX (Capital Expenditure) and OPEX in the communication business is required. In addition, with the declining population, a shortage of specialized personnel is also expected. Therefore, there is also a demand for automatically performing operations that have been manually performed until now.

[0005] Against the background as described above, automation of the operation and maintenance services of the network is required. The operation and maintenance of the network is mainly performed by a controller that manages the network and devices provided in the network. The controller manages each device information (configuration information, failure information, PM (Performance Monitoring) information, etc.) and network information (topology information, logical path information, service information, etc.). In addition, the controller is manually operated by an operator or some operations are automated.

[0006] There are mainly two approaches to the automation of network operation and maintenance. One approach is rule-based automation, and the other approach is automation utilizing AI (Artificial Intelligence).

[0007] This disclosure focuses on automation utilizing AI applicable to various networks. Further, this disclosure assumes that the target network is an optical transmission network and the target device is an optical transmission device. Here, as a technique for automating the operation and maintenance of an optical transmission network by utilizing AI, for example, the technique disclosed in Patent Document 1 can be cited.

Prior Art Document

Patent Document

[0008]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0009] By the way, in an optical transmission network, when an abnormality occurs in the physical port of an optical transmission device or in the optical fiber between optical transmission devices, the PM data representing the optical power level (OP: Optical Power) of the optical signal of each optical transmission device fluctuates.

[0010] Therefore, in advance, the AI model uses the PM data as learning data to learn (machine learning) the temporal fluctuations of the PM data. When an abnormality occurs, the controller uses the AI model to detect the abnormality by detecting the fluctuations of the PM data associated with the occurrence of the abnormality.

[0011] However, after the abnormality recovery, depending on the way of recovery, the state of the optical transmission device or the optical fiber may change slightly, and the PM data may fluctuate. In this case, the PM data after the abnormality recovery may have a different pattern from the PM data used as learning data in the AI model before the occurrence of the abnormality, so the AI model may need to be retrained.

[0012] Furthermore, if the influence of the abnormality and recovery is large, the subsequent optical transmission device at the location where the abnormality occurred may also be affected by the influence (fluctuation of PM data), and the AI model of the affected optical transmission device may also need to be retrained.

[0013] However, if all the AI models of the optical transmission devices affected by the abnormality and recovery are retrained, it will cause a problem of an increase in the retraining cost of the AI model.

[0014] Therefore, an object of the present disclosure is to provide a controller, a learning cost reduction method, and a program capable of reducing the retraining cost of the AI model after the abnormality recovery in view of the above-described problems.

Means for Solving the Problems

[0015] A controller according to one aspect is a collection unit that collects PM (Performance Monitoring) data representing the optical power level of each optical signal of a plurality of optical transmission devices provided in an optical transmission network; a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, and learning the time-series variation of the PM data of the optical transmission device using the PM data of the optical transmission device as learning data; a database that holds the average value and the allowable error of the PM data of the optical transmission device for each of the plurality of optical transmission devices; an abnormality detection unit that detects an abnormality in the optical transmission network using the plurality of AI models; When there is an optical transmission device whose PM data after abnormal recovery is outside the threshold range derived from the average value and the tolerance error, a relearning determination unit that determines that the AI model of the corresponding optical transmission device needs to be relearned, is provided.

[0016] A learning cost reduction method according to one aspect is a learning cost reduction method executed by a controller, collecting PM (Performance Monitoring) data representing the optical power level of the optical signal of each of a plurality of optical transmission devices provided in an optical transmission network; using the PM data of the optical transmission device as learning data by a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, and learning the time-series variation of the PM data of the optical transmission device; for each of the plurality of optical transmission devices, holding the average value and the tolerance error of the PM data of the optical transmission device in a database; detecting an abnormality in the optical transmission network using the plurality of AI models; when there is an optical transmission device whose PM data after abnormal recovery is outside the threshold range derived from the average value and the tolerance error, determining that the AI model of the corresponding optical transmission device needs to be relearned; including.

[0017] A program according to one aspect is for a computer, a procedure for collecting PM (Performance Monitoring) data representing the optical power level of the optical signal of each of a plurality of optical transmission devices provided in an optical transmission network; a procedure for using the PM data of the optical transmission device as learning data by a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, and learning the time-series variation of the PM data of the optical transmission device; For each of the plurality of optical transmission devices, a procedure for storing in a database the average value and tolerance of the PM data of the optical transmission device, and a procedure for detecting an abnormality in the optical transmission network using the plurality of AI models, and if there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the tolerance, a procedure for determining that the AI model of the corresponding optical transmission device needs to be retrained, and cause to execute.

Advantages of the Invention

[0018] According to the above aspect, an effect can be obtained in that a controller, a learning cost reduction method, and a program capable of reducing the retraining cost of an AI model after abnormality recovery can be provided.

Brief Description of the Drawings

[0019]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Mode for Carrying Out the Invention

[0020] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. Note that the following description and drawings are appropriately omitted and simplified for clarity of explanation. In addition, in each of the following drawings, the same elements are denoted by the same reference numerals, and redundant descriptions are omitted as necessary. Also, the specific numerical values shown below are merely examples for facilitating the understanding of the present disclosure and are not limited thereto.

[0021] <Embodiment 1> First, with reference to FIG. 1, a configuration example of the network system 1 will be described. As shown in FIG. 1, the network system 1 includes an OSS (Operation Support System) / BSS (Business Support System) orchestrator 10, controllers 20X, 20Y, 20Z, optical transmission networks 30X, 30Y, 30Z, and optical transmission devices (for example, optical transmission devices 40A, 40B, 40C, 40D, 40E provided in the optical transmission networks 30X, 30Y, 30Z), respectively.

[0022] In the drawings described in FIG. 1 and later, the optical transmission devices 40A, 40B, 40C, 40D, 40E are appropriately shown as devices A, B, C, D, E.

[0023] In the following description, when it is not specified which of the controllers 20X, 20Y, and 20Z is being referred to, they are collectively referred to as controller 20. Similarly, the optical transmission networks 30X, 30Y, and 30Z are collectively referred to as optical transmission network 30, and the optical transmission devices 40A, 40B, 40C, 40D, and 40E are collectively referred to as optical transmission device 40.

[0024] Also, in FIG. 1, three controllers 20 and three optical transmission networks 30 are shown, respectively. However, this number is merely an example and is not limited thereto. Further, in FIG. 1 and the drawings described hereinafter, five optical transmission devices 40A, 40B, 40C, 40D, and 40E are shown as the optical transmission devices 40 provided in the optical transmission network 30X. However, this number is merely an example and is not limited thereto.

[0025] The controller 20 is connected to the optical transmission network 30. The controller 20 manages the network information (such as topology information, logical path information, and service information) of the connected optical transmission network 30, and also manages the device information (such as configuration information, fault information, and PM information) of each optical transmission device 40 provided in the connected optical transmission network 30. Further, some operations of the controller 20 are automated, and the remaining operations are performed manually by an operator, or all operations are automated. In the present disclosure, at least the operations of the autonomous control function unit 25 described hereinafter are automated.

[0026] Hereinafter, a configuration example of the controller 20 and the optical transmission device 40 will be described. First, with reference to FIG. 2, a configuration example of the controller 20 will be described. As shown in FIG. 2, the controller 20 includes an NBI (Northbound Interface) 21, an SBI (Southbound Interface) 22, a device management function unit 23, an NW (Network) management function unit 24, an autonomous control function unit 25, and a database 26.

[0027] The device management function unit 23 includes a configuration management function unit 231, a fault management function unit 232, and a PM management function unit 233. The device management function unit 23 is realized by, for example, an EMS (Element Management System).

[0028] The configuration management function unit 231 collects the configuration information of each optical transmission device 40 provided in the connected optical transmission network 30 and manages the collected configuration information. The fault management function unit 232 collects the fault information of each optical transmission device 40 provided in the connected optical transmission network 30 and manages the collected fault information. The PM management function unit 233 collects the PM data of each optical transmission device 40 provided in the connected optical transmission network 30 and manages the collected PM data as PM information. As described above, the PM data is data representing the optical power level of an optical signal.

[0029] The NW (Network) management function unit 24 includes a topology management function unit 241, a logical path management function unit 242, and a service management function unit 243. The NW management function unit 24 is realized by, for example, an NMS (Network Management System).

[0030] The topology management function unit 241 collects the topology information of the connected optical transmission network 30 and manages the collected topology information. The logical path management function unit 242 collects the logical path information of the connected optical transmission network 30 and manages the collected logical path information. The service management function unit 243 collects the service information of the connected optical transmission network 30 and manages the collected service information.

[0031] Note that the controller 20 includes both the device management function unit 23 and the NW management function unit 24, but is not limited thereto. The controller 20 may include only the device management function unit 23, and the NW management function unit 24 may be provided in another device.

[0032] The self-control function unit 25 includes a relearning determination function unit 251, an AI (machine learning) function unit 252, and a closed loop control unit 253.

[0033] The AI function unit 252 uses the PM data as learning data and performs machine learning on the time-series fluctuations of the PM data. In the present disclosure, the purpose of machine learning is to perform prognostic detection to detect abnormal behavior (abnormalities in the physical ports of the optical transmission device 40 or abnormalities in the optical fiber) before a failure of the optical transmission device 40 or a failure of the optical transmission device 40 or the optical fiber occurs within the optical transmission network 30.

[0034] In addition, the AI model of machine learning can be created in various units such as the optical transmission device 40 unit, the path unit, and the optical transmission network 30 unit. Here, a path is a path constructed to pass an optical signal within the optical transmission network 30, and is a logical path that spans a plurality of optical transmission devices 40. The path between the optical transmission devices 40 is physically constructed to pass through the optical fiber connecting the optical transmission devices 40. Also, in the path between the optical transmission devices 40, optical signals of a plurality of wavelengths are transmitted through a single optical fiber by an optical wavelength multiplexing method typified by WDM (Wavelength Division Multiplexing).

[0035] However, if a single AI model is created for the entire plurality of optical transmission networks 30 or an AI model is created in units of the optical transmission network 30, even if there is a slight change in the optical transmission network 30, the AI model needs to be relearned, so frequent relearning is required and the AI model is unstable.

[0036] Also, if an AI model is created in units of paths, there is a concern that the computing resources required for machine learning will increase because the number of AI models also increases linearly with the linear increase in paths.

[0037] Therefore, in the present disclosure, it is premised that an AI model is created in units of the optical transmission device 40. More specifically, in the present disclosure, it is assumed that an AI model is created for each of the OPT (Optical Power Transceiver) side and the OPR (Optical Power Receiver) side of the optical transmission device 40 in units of the optical transmission device 40.

[0038] Therefore, the PM management function unit 233 periodically collects the PM data of the OPT side and the OPR side of each optical transmission device 40 from each optical transmission device 40 provided in the connected optical transmission network 30.

[0039] In addition, the AI function unit 252 creates an AI model that has learned the temporal variation of the corresponding PM data for each optical transmission device 40 provided in the connected optical transmission network 30 and for each of the OPT side and the OPR side.

[0040] Here, when any abnormality (such as slight contamination of a physical port or stress on an optical fiber) occurs in the physical port of the optical transmission device 40 through which the path constructed in the optical transmission network 30 passes or in the optical fiber, the PM data after the abnormality recovery may be different from the PM data that has been used as learning data in the AI model until then. In that case, basically, retraining of the AI model is required. However, if all AI models with different PM data before and after the abnormality recovery are retrained, a problem of increased retraining cost of the AI model will occur.

[0041] Therefore, in the present disclosure, the retraining determination function unit 251 determines whether retraining of the AI model is necessary for each optical transmission device 40 and for each of the OPT side and the OPR side after the abnormality recovery. For the AI model determined by the retraining determination function unit 251 not to require retraining, retraining is not performed. As a result, since the number of AI models to be retrained is reduced, the retraining cost of the AI model is reduced. The details of the operation of the retraining determination function unit 251 will be described later.

[0042] The closed-loop control unit 253 controls the connected optical transmission network 30 and each optical transmission device 40 provided in the connected optical transmission network 30 by using the AI model created by the AI function unit 252 and the determination result determined by the re-learning determination function unit 251.

[0043] The database 26 holds the configuration information, fault information, and PM information of each optical transmission device 40 provided in the connected optical transmission network 30, which are managed by the configuration management function unit 231, the fault management function unit 232, and the PM management function unit 233.

[0044] In addition, the database 26 holds the topology information, logical path information, and service information of the connected optical transmission network 30, which are managed by the topology management function unit 241, the logical path management function unit 242, and the service management function unit 243.

[0045] Here, as for the PM information, the database 26 holds the average value Av and the tolerance R of the corresponding PM data for each optical transmission device 40 provided in the connected optical transmission network 30 and for each of the OPT side and the OPR side.

[0046] For example, as shown in FIG. 3, the database 26 in the controller 20X holds the average value Av and the tolerance R of the corresponding PM data for each of the optical transmission devices 40A, 40B, 40C, 40D, 40E provided in the optical transmission network 30X and for each of the OPT side and the OPR side. In the example of FIG. 3, for the sake of simplicity of explanation, the values of the average value Av and the tolerance R for all the OPT sides and OPR sides of the optical transmission devices 40A, 40B, 40C, 40D, 40E are all set to the same value. However, this value is just an example and is not limited thereto, and they may be different from each other.

[0047] After the abnormality recovery, the relearning determination function unit 251 determines, for each optical transmission device 40 and for each of the OPT side and the OPR side, whether relearning of the AI model is necessary using the value Value of the PM data after the abnormality recovery, the average value Av of the PM data held in the database 26, and the allowable error R according to the following mathematical formula 1.

Number

[0048] For example, in the example of FIG. 3, the average value Av of the PM data on the OPR side of the optical transmission device 40C is 0.5, and the allowable error R is 0.3. Therefore, the relearning determination function unit 251 of the controller 20X determines that relearning of the AI model on the OPR side of the optical transmission device 40C is necessary when the value Value of the PM data after the abnormality recovery on the OPR side of the optical transmission device 40C is greater than 0.8 (= 0.5 + 0.3) or less than 0.2 (= 0.5 - 0.3).

[0049] In other words, the relearning determination function unit 251 determines that relearning of the AI model is necessary when the value Value of the PM data after the abnormality recovery is outside the threshold range of Av - R ≤ Value ≤ Av + R (in the example of FIG. 3, the threshold range of 0.2 ≤ Value ≤ 0.8).

[0050] Next, with reference to FIG. 4, a configuration example of the optical transmission device 40 will be described. As shown in FIG. 4, the optical transmission device 40 includes AMPs (Amplifiers) 41 and 42 and a control unit 43.

[0051] In the example of FIG. 4, it is assumed that a path for passing an optical signal is constructed from left to right in the figure. Therefore, the end point of the AMP 41 on the left side in the figure is the OPR, and the end point of the AMP 42 on the right side in the figure is the OPT.

[0052] Based on the optical power level of the optical signal on the OPR side (the received optical signal), the control unit 43 adjusts the optical power level of the optical signal on the OPT side (the transmitted optical signal) by a VOA (Variable Optical Attenuator) not shown in the figure. The VOA is an optical fiber component that adjusts the attenuation amount of the optical power level of the optical signal.

[0053] At this time, the control unit 43 adjusts the optical power level of the optical signal on the OPT side by the VOA so as to be the target OPT. However, an upper limit of the adjustment amount is set for the VOA, and the control unit 43 cannot perform an adjustment exceeding the upper limit of the adjustment amount of the VOA.

[0054] Also, the control unit 43 periodically measures the optical power level of the optical signal on the OPR side and the optical power level of the optical signal on the OPT side, and notifies the controller 20 of the PM data representing the optical power level on the OPR side and the PM data representing the optical power level on the OPT side.

[0055] Next, an operation example of the network system 1 will be described. Hereinafter, an operation example of the network system 1 will be described by taking the controller 20X, the optical transmission network 30X, and the optical transmission devices 40A, 40B, 40C, 40D, and 40E among the components of the network system 1 as examples.

[0056] First, with reference to FIG. 5, an operation example in the case of collecting PM data in the network system 1 will be described. In the example of FIG. 5, it is assumed that a path for passing an optical signal is constructed from the optical transmission device 40A toward the optical transmission device 40E.

[0057] As shown in FIG. 5, the optical transmission devices 40B, 40C, and 40D, which are the midpoints of the above paths, periodically notify the controller 20X of the PM data on the OPR side and the PM data on the OPT side. Also, the optical transmission device 40A, which is the starting point of the above path, periodically notifies the controller 20X of the PM data on the OPT side. Further, the optical transmission device 40E, which is the end point of the above path, periodically notifies the controller 20X of the PM data on the OPR side.

[0058] Here, examples of the timing at which the optical transmission devices 40A, 40B, 40C, 40D, and 40E periodically notify the PM data include timings such as every 15 minutes or every day. These timings are standardized in ITU-T (International Telecommunication Union - Telecommunication sector)'s Open ROADM (Reconfigurable Optical Add / Drop Multiplexers) and the like.

[0059] At this time, the optical transmission devices 40A, 40B, 40C, 40D, and 40E do not directly notify the controller 20X of the PM data for, say, 15 minutes. Instead, they summarize the PM data for 15 minutes and notify the controller 20X of the summarized PM data. Therefore, in the example of FIG. 5, for instance, the optical transmission device 40A notifies the summarized PM data of "0.5 ± 0.1" as the PM data on the OPT side. The same applies when the timing for notifying the PM data is every day.

[0060] Next, in the network system 1, operation examples in each of case 1 where an abnormality occurs such that a slight stress is applied to the optical fiber due to snowfall or the like, and case 2 where an abnormality occurs such that the optical fiber bends strongly, will be described.

[0061] Hereinafter, for the sake of simplicity of explanation, it is assumed that the normal range of the PM data on the OPT side and the OPR side of all the optical transmission devices 40A, 40B, 40C, 40D, and 40E is "-2 to +2". Therefore, when any of the PM data on the OPT side and the OPR side of the optical transmission devices 40A, 40B, 40C, 40D, and 40E deviates from the above normal range, the closed-loop control unit 253 of the controller 20X determines that a failure has occurred in the corresponding optical transmission device 40, or a failure has occurred in the corresponding optical transmission device 40 or the optical fiber connected to the corresponding optical transmission device 40.

[0062] Here, the abnormalities in Cases 1 and 2 are detected as omens before the above-mentioned failures or obstacles occur. Therefore, it is assumed that the abnormalities in Cases 1 and 2 occur when the PM data is within the above normal range. Further, it is assumed that the abnormalities in Cases 1 and 2 are detected by the closed-loop control unit 253 of the controller 20X using an AI model.

[0063] Also, it is assumed that all the optical transmission devices 40A, 40B, 40C, 40D, and 40E have an OPT target of "0.5" and the upper limit of the adjustment amount of the VOA is "+0.5". Further, it is assumed that the database 26 of the controller 20X holds the information shown in FIG. 3 as the average value Av and the allowable error R of the PM data on the OPT side and the OPR side of the optical transmission devices 40A, 40B, 40C, 40D, and 40E. Also, in the optical transmission network 30X, it is assumed that a path for passing an optical signal is constructed from the optical transmission device 40A to the optical transmission device 40E.

[0064] (1) Case 1 First, Case 1 in which an abnormality occurs in that a slight stress is applied to the optical fiber will be described.

[0065] First, with reference to FIG. 6, an example of PM data when the abnormality in Case 1 occurs in the network system 1 will be described. As shown in FIG. 6, during normal operation, the optical transmission devices 40A, 40B, 40C, 40D, and 40E adjust the optical power level on the OPT side to the target OPT by using a VOA.

[0066] Here, it is assumed that an abnormality has occurred where a slight stress is applied to the optical fiber between the optical transmission devices 40B and 40C. During the occurrence of the above abnormality, the optical power level on the OPR side of the optical transmission device 40C drops significantly and becomes "-0.5 ± 0.1". Then, the optical transmission device 40B detects the drop in the optical power level on the OPR side of the optical transmission device 40C and raises the optical power level on the OPT side to the upper limit of the adjustment amount of the VOA to make it "0.9 ± 0.1".

[0067] Also, the subsequent optical transmission devices 40D and 40E are also affected by the above abnormality, and the optical power levels on the OPR sides of the optical transmission devices 40D and 40E have also dropped. Therefore, the optical transmission devices 40C and 40D also raise the optical power level on the OPT side to the upper limit of the adjustment amount of the VOA.

[0068] Subsequently, it is assumed that the above abnormality has been recovered. At this time, the optical power level after the recovery of the abnormality on the OPR side of the optical transmission device 40C is "0.2 ± 0.1", which has changed from the normal power level of "0.4 ± 0.1". On the other hand, the optical power levels after the recovery of the abnormality on the OPR sides of the other optical transmission devices 40B, 40D, and 40E have not changed from the normal power levels.

[0069] Next, in the network system 1, an example of the operation after the recovery of the abnormality when the abnormality of Case 1 occurs will be described.

[0070] After the recovery of the abnormality of Case 1, the controller 20X determines whether re-learning of the AI model is necessary for each of the optical transmission devices 40A, 40B, 40C, 40D, and 40E, and for each of the OPT side and the OPR side.

[0071] Hereinafter, with reference to FIG. 7, among the above operations performed by the controller 20X after the recovery from the abnormality in Case 1, an operation of determining whether retraining of the AI models on the OPR side of the optical transmission devices 40B, 40C, 40D, and 40E is necessary will be described.

[0072] As shown in FIG. 7, the PM management function unit 233 of the controller 20X collects PM data on the OPR side of the optical transmission devices 40B, 40C, 40D, and 40E at regular intervals after the recovery from the abnormality in Case 1. Here, it is assumed that the value Value of the PM data on the OPR side of the optical transmission device 40C is "0.2 ± 0.1", and the value Value of the PM data on the OPR side of the optical transmission device 40D is "0.4 ± 0.1" (steps S11, S12).

[0073] Next, the retraining determination function unit 251 of the controller 20X determines whether retraining of the AI models on the OPR side is necessary for each of the optical transmission devices 40B, 40C, 40D, and 40E (step S13).

[0074] At this time, the retraining determination function unit 251 of the controller 20X uses the value Value of the PM data after the recovery from the abnormality in Case 1, the average value Av (= 0.5) and the tolerance R (= 0.3) of the PM data held in the database 26, and determines whether retraining of the AI model is necessary according to the above formula 1. According to the above formula 1, the threshold range of the value Value of the PM data after the recovery from the abnormality is 0.2 ≤ Value ≤ 0.8.

[0075] Here, the value Value of the PM data on the OPR side of the optical transmission device 40C is "0.2 ± 0.1", and the value Value of the PM data on the OPR side of the optical transmission device 40D is "0.4 ± 0.1", and both are within the above threshold range. Therefore, the retraining determination function unit 251 of the controller 20X determines that retraining of the AI models on the OPR side of the optical transmission devices 40C and 40D is unnecessary. Also, although the description is omitted, the retraining determination function unit 251 of the controller 20X also determines that retraining of the AI models on the OPR side of the optical transmission devices 40B and 40E is unnecessary.

[0076] Thus, although the PM data after the recovery of the OPR side of the optical transmission device 40C has changed from the PM data during normal operation, it is within the above threshold range. Therefore, the AI model on the OPR side of the optical transmission device 40C is determined not to require re-learning.

[0077] (2) Case 2 Next, a description will be given of Case 2 in which an abnormality occurs where the optical fiber is bent strongly.

[0078] First, referring to FIG. 8, an example of PM data when the abnormality of Case 2 occurs in the network system 1 will be described. As shown in FIG. 8, during normal operation, the optical transmission devices 40A, 40B, 40C, 40D, and 40E adjust the optical power level on the OPT side to the target OPT by means of the VOA.

[0079] Here, it is assumed that an abnormality occurs where the optical fiber between the optical transmission devices 40B and 40C is bent strongly. During the occurrence of the above abnormality, the optical power level on the OPR side of the optical transmission device 40C drops significantly to "-4.0 ± 0.1". Then, the optical transmission device 40B detects the drop in the optical power level on the OPR side of the optical transmission device 40C and raises the optical power level on the OPT side to the upper limit of the adjustment amount of the VOA to "0.9 ± 0.1".

[0080] Also, the subsequent optical transmission devices 40D and 40E are also affected by the above abnormality, and the optical power levels on the OPR sides of the optical transmission devices 40D and 40E have also dropped. Therefore, the optical transmission devices 40C and 40D also raise the optical power level on the OPT side to the upper limit of the adjustment amount of the VOA.

[0081] After that, it is assumed that the above abnormality has been recovered. At this time, the optical power level after the recovery from the abnormality on the OPR side of the optical transmission device 40C is "-0.8 ± 0.1", which has changed from the normal power level of "0.4 ± 0.1". Also, the optical power level after the recovery from the abnormality on the OPR side of the optical transmission device 40D is "-0.3 ± 0.1", which has changed from the normal power level of "0.4 ± 0.1". Further, the optical power level after the recovery from the abnormality on the OPR side of the optical transmission device 40E is "0.2 ± 0.1", which has changed from the normal power level of "0.4 ± 0.1". On the other hand, the optical power level after the recovery from the abnormality on the OPR side of the other optical transmission device 40B has not changed from the normal power level.

[0082] Next, in the network system 1, an example of the operation after the recovery from the abnormality when the abnormality of case 2 occurs will be described.

[0083] After the recovery from the abnormality of case 2, the controller 20X determines whether re - learning of the AI model is necessary for each of the optical transmission devices 40A, 40B, 40C, 40D, 40E and for each of the OPT side and the OPR side.

[0084] Hereinafter, with reference to FIG. 9, among the above operations performed by the controller 20X after the recovery from the abnormality of case 2, the operation of determining whether re - learning of the AI model on the OPR side of the optical transmission devices 40B, 40C, 40D, 40E is necessary will be described.

[0085] As shown in FIG. 9, the PM management function unit 233 of the controller 20X collects the PM data on the OPR side of the optical transmission devices 40B, 40C, 40D, 40E at regular intervals after the recovery from the abnormality of case 2. Here, it is assumed that the value Value of the PM data on the OPR side of the optical transmission device 40C is "-0.8 ± 0.1", the value Value of the PM data on the OPR side of the optical transmission device 40D is "-0.3 ± 0.1", and the value Value of the PM data on the OPR side of the optical transmission device 40E is "0.2 ± 0.1" (steps S21 to S23).

[0086] Next, the relearning determination function unit 251 of the controller 20X determines whether relearning of the AI model on the OPR side is necessary for each of the optical transmission devices 40B, 40C, 40D, and 40E (step S24).

[0087] At this time, the relearning determination function unit 251 of the controller 20X uses the value Value of the PM data after the recovery from the abnormality in Case 2, the average value Av (= 0.5) of the PM data held in the database 26, and the allowable error R (= 0.3) to determine whether relearning of the AI model is necessary according to the above formula (1). According to the above formula (1), the threshold range of the value Value of the PM data after the recovery from the abnormality is 0.2 ≤ Value ≤ 0.8.

[0088] Here, the value Value of the PM data on the OPR side of the optical transmission device 40C is "-0.8 ± 0.1", and the value Value of the PM data on the OPR side of the optical transmission device 40D is "-0.3 ± 0.1", both of which are outside the above threshold range. Therefore, the relearning determination function unit 251 of the controller 20X determines that the AI models on the OPR sides of the optical transmission devices 40C and 40D need to be relearned. On the other hand, the value Value of the PM data on the OPR side of the optical transmission device 40E is "0.2 ± 0.1", which is within the above threshold range. Therefore, the relearning determination function unit 251 of the controller 20X determines that the AI model on the OPR side of the optical transmission device 40E does not need to be relearned. Also, although the explanation is omitted, the relearning determination function unit 251 of the controller 20X also determines that the AI model on the OPR side of the optical transmission device 40B does not need to be relearned.

[0089] In this way, although the PM data on the OPR side of the optical transmission device 40E has changed from the normal PM data after the recovery from the abnormality, it is within the above threshold range. Therefore, the AI model on the OPR side of the optical transmission device 40E is determined not to need to be relearned.

[0090] Thereafter, the relearning determination function unit 251 of the controller 20X instructs the AI function unit 252 of the controller 20X to relearn the AI models on the OPR sides of the optical transmission devices 40C and 40D (steps S25 and S26).

[0091] As described above, according to the first embodiment, for each optical transmission device 40 provided in the connected optical transmission network 30, and for each of the OPR side and the OPT side, the controller 20 holds the average value and the tolerance of the PM data on the OPR side or the OPT side of the optical transmission device 40. Then, when there is an optical transmission device 40 in which the PM data on the OPR side or the OPT side after recovery from an abnormality is out of the threshold range derived from the average value and the tolerance, the controller 20 determines that the OPR side or the OPT side AI model of the corresponding optical transmission device 40 needs to be re-learned.

[0092] Therefore, even if there is a slight fluctuation in the PM data after recovery from an abnormality, if the PM data is within the threshold range, re-learning of the AI model is not required. As a result, since the number of AI models to be re-learned is reduced, it is possible to reduce the re-learning cost of the AI model. In addition, since the number of AI models to be re-learned is reduced, it is also possible to contribute to reducing the load on computing resources, stabilizing the AI model, and automating the operation and maintenance work of the network.

[0093] <Second Embodiment> The second embodiment corresponds to an embodiment in which the first embodiment described above is generalized to a higher concept. With reference to FIG. 10, a configuration example of the controller 200 will be described. As shown in FIG. 10, the controller 200 includes a collection unit 201, a learning unit 202, an abnormality detection unit 203, a re-learning determination unit 204, and a database 205.

[0094] The collection unit 201 collects PM (Performance Monitoring) data representing the optical power level of the optical signal of each of a plurality of optical transmission devices provided in the optical transmission network. The learning unit 202 includes a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices and learning the temporal variation of the PM data of the optical transmission device using the PM data of the optical transmission device as learning data.

[0095] The database 205 holds the average value and the allowable error of the PM data of each optical transmission device for a plurality of optical transmission devices. The abnormality detection unit 203 detects abnormalities in the optical transmission network using a plurality of AI models.

[0096] When there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, the relearning determination unit 204 determines that the AI model of the corresponding optical transmission device needs to be relearned.

[0097] As described above, according to the second embodiment, the controller 200 holds the average value and the allowable error of the PM data of each optical transmission device for a plurality of optical transmission devices. Then, when there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, the controller 200 determines that the AI model of the corresponding optical transmission device needs to be relearned.

[0098] Therefore, even if there is some fluctuation in the PM data after abnormality recovery, if the PM data falls within the threshold range, relearning of the AI model is not required. As a result, since the number of AI models to be relearned is reduced, it is possible to reduce the relearning cost of the AI model.

[0099] Note that the collection unit 201 may collect PM data at the receiving end (corresponding to the above OPR) and the transmitting end (corresponding to the above OPT) of each of the plurality of optical transmission devices. Also, a plurality of AI models may be provided for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end, and may learn the temporal variation of the PM data at the receiving end or the transmitting end of the optical transmission device. Further, the database 205 may hold the average value and the allowable error of the PM data at the receiving end or the transmitting end of the optical transmission device for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end. Also, when there is an optical transmission device whose PM data at the receiving end or the transmitting end after abnormality recovery is outside the threshold range, the relearning determination unit 204 may determine that the AI model at the receiving end or the transmitting end of the corresponding optical transmission device needs to be relearned.

[0100] In addition, the relearning determination unit 204 may instruct relearning for the AI model of the corresponding optical transmission device determined to require relearning. In addition, for an optical transmission device in which the PM data after abnormality recovery is within the threshold range, the relearning determination unit 204 may determine that relearning of the AI model of the optical transmission device is unnecessary.

[0101] In addition, the threshold range may be a range having, as an upper limit, a value obtained by adding an allowable error to the average value and, as a lower limit, a value obtained by subtracting the allowable error from the average value. In addition, the collection unit 201 may periodically collect the PM data of each of the plurality of optical transmission devices.

[0102] In addition, an abnormality in the optical transmission network may be an abnormality in any physical port of a plurality of optical transmission devices. Alternatively, the abnormality in the optical transmission network may be an abnormality in an optical fiber between a plurality of optical transmission devices.

[0103] <Hardware Configuration of Controller According to Embodiment> With reference to FIG. 11, a hardware configuration example of a computer 900 that realizes the above-described controllers 20 and 200 will be described. As shown in FIG. 11, the computer 900 includes a processor 901 and a memory 902. The processor 901 and the memory 902 are coupled to each other.

[0104] The processor 901 may be, for example, a microprocessor, an MPU (Micro Processing Unit), or a CPU (Central Processing Unit). The processor 901 may include a plurality of processors.

[0105] The memory 902 is configured by a combination of a volatile memory and a non-volatile memory. The memory 902 may include storage disposed apart from the processor 901. In this case, the processor 901 may access the memory 902 via an I (Input) / O (Output) interface (not shown).

[0106] The memory 902 may store a software module (computer program) including a group of instructions and data for performing the processing by the above-described controllers 20, 200.

[0107] Also, in some implementations, the processor 901 may be configured to perform the processing of the above-described controllers 20, 200 by reading and executing the software module from the memory 902.

[0108] Also, the above-described program may be stored in a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, the computer-readable medium or the tangible storage medium includes RAM, ROM, flash memory, SSD, or other memory technologies, CD (Compact Disc)-ROM, DVD (Digital Versatile Disc), Blu-ray (registered trademark) disc, or other optical disc storage, magnetic cassette, magnetic tape, magnetic disk storage, or other magnetic storage devices. The program may be transmitted on a transitory computer-readable medium or a communication medium. By way of example and not limitation, the transitory computer-readable medium or the communication medium includes electrical, optical, acoustic, or other forms of propagated signals.

[0109] As described above, the present disclosure has been described with reference to the embodiments, but the present disclosure is not limited to the above-described embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. And each embodiment can be combined with other embodiments as appropriate.

[0110] Also, each drawing is merely an illustration for explaining one or more embodiments. Each drawing is not associated with only one specific embodiment, but may be associated with one or more other embodiments. As those skilled in the art can understand, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings to create, for example, embodiments that are not explicitly illustrated or described. Not all of the features or steps shown in any one drawing for explaining exemplary embodiments are necessarily essential, and some features or steps may be omitted. The order of the steps described in any drawing may be changed as appropriate.

[0111] Also, some or all of the above embodiments may be described as follows, but are not limited thereto. (Appendix 1) A collection unit that collects PM (Performance Monitoring) data representing the optical power level of the optical signals of each of a plurality of optical transmission devices provided in an optical transmission network; A plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, which use the PM data of the optical transmission device as learning data to learn the temporal variation of the PM data of the optical transmission device; For each of the plurality of optical transmission devices, a database that holds the average value and allowable error of the PM data of the optical transmission device; An abnormality detection unit that detects an abnormality in the optical transmission network using the plurality of AI models; A relearning determination unit that determines that relearning of the AI model of the corresponding optical transmission device is necessary if there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error; A controller comprising the above. (Appendix 2) The collection unit collects the PM data at the receiving end and the transmitting end of each of the plurality of optical transmission devices; The plurality of AI models are provided for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end, and learn the temporal variation of the PM data at the receiving end or the transmitting end of the optical transmission device. The database holds the average value and the allowable error of the PM data at the receiving end or the transmitting end of the optical transmission device for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end. When there is an optical transmission device in which the PM data at the receiving end or the transmitting end after abnormality recovery is outside the threshold range, the re-learning determination unit determines that re-learning of the AI model at the receiving end or the transmitting end of the corresponding optical transmission device is necessary. The controller according to Supplementary Note 1. (Supplementary Note 3) The re-learning determination unit instructs re-learning for the AI model of the corresponding optical transmission device that has been determined to require re-learning. The controller according to Supplementary Note 1. (Supplementary Note 4) For an optical transmission device in which the PM data after abnormality recovery is within the threshold range, the re-learning determination unit determines that re-learning of the AI model of the optical transmission device is unnecessary. The controller according to Supplementary Note 1. (Supplementary Note 5) The threshold range is a range having, as an upper limit, a value obtained by adding the allowable error to the average value and, as a lower limit, a value obtained by subtracting the allowable error from the average value. The controller according to Supplementary Note 1. (Supplementary Note 6) The collection unit periodically collects the PM data of each of the plurality of optical transmission devices. The controller according to Supplementary Note 1. (Supplementary Note 7) An abnormality in the optical transmission network is an abnormality in any physical port of the plurality of optical transmission devices. The controller according to Supplementary Note 1. (Supplementary Note 8) An abnormality in the optical transmission network is an abnormality in the optical fiber between the plurality of optical transmission devices. The controller according to Supplementary Note 1. (Supplementary Note 9) A learning cost reduction method executed by a controller, comprising: Collecting PM (Performance Monitoring) data representing the optical power level of each optical signal of a plurality of optical transmission devices provided in an optical transmission network; Using a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, learning the time-series variation of the PM data of the optical transmission device by using the PM data of the optical transmission device as learning data; For each of the plurality of optical transmission devices, storing in a database the average value and the allowable error of the PM data of the optical transmission device; Detecting an abnormality in the optical transmission network by using the plurality of AI models; When there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, determining that the AI model of the corresponding optical transmission device needs to be re-learned; A learning cost reduction method including the above steps. (Appendix 10) A program for causing a computer to: Execute a procedure for collecting PM (Performance Monitoring) data representing the optical power level of each optical signal of a plurality of optical transmission devices provided in an optical transmission network; Execute a procedure for learning the time-series variation of the PM data of an optical transmission device by using a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices and using the PM data of the optical transmission device as learning data; Execute a procedure for storing in a database the average value and the allowable error of the PM data of each of the plurality of optical transmission devices; Execute a procedure for detecting an abnormality in the optical transmission network by using the plurality of AI models; Execute a procedure for determining that the AI model of an optical transmission device needs to be re-learned when there is an optical transmission device whose PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error; A program for causing a computer to execute the above steps.

[0112] Also, some or all of the elements (e.g., configurations and functions) described in Appendices 2 to 8 that are subordinate to Appendix 1 may be subordinate to Appendices 9 and 10 in the same subordinate relationship as Appendices 2 to 8. Some or all of the elements described in any appendix may be applicable to various hardware, software, recording means for recording software, systems, and methods.

Explanation of Signs

[0113] 1 Network system 10 OSS / BSS orchestrator 20, 20X, 20Y, 20Z Controller 21 NBI 22 SBI 23 Device management function unit 231 Configuration management function unit 232 Fault management function unit 233 PM management function unit 24 NW management function unit 241 Topology management function unit 242 Logical path management function unit 243 Service management function unit 25 Autonomous control function unit 251 Relearning determination function unit 252 AI (machine learning) function unit 253 Closed-loop control unit 26 Database 30, 30X, 30Y, 30Z Optical transmission network 40, 40A, 40B, 40C, 40D, 40E Optical transmission device 41, 42 AMP 43 Control unit 200 Controller 201 Collection unit 202 Learning unit 203 Abnormality detection unit 204 Relearning determination unit 205 Database 900 Computer 901 Processor 902 Memory

Claims

1. A collection unit that collects PM (Performance Monitoring) data representing the optical power level of the optical signals of each of a plurality of optical transmission devices provided in an optical transmission network; A plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, and using the PM data of the optical transmission device as learning data to learn the time-series variation of the PM data of the optical transmission device; For each of the plurality of optical transmission devices, a database that holds the average value and allowable error of the PM data of the optical transmission device; An abnormality detection unit that detects an abnormality in the optical transmission network using the plurality of AI models; When there is an optical transmission device in which the PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, a relearning determination unit that determines that relearning of the AI model of the corresponding optical transmission device is necessary; A controller comprising:

2. The collection unit collects the PM data at the receiving end and the transmitting end of each of the plurality of optical transmission devices, The plurality of AI models are provided for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end, and learn the time-series variation of the PM data at the receiving end or the transmitting end of the optical transmission device, The database holds the average value and the allowable error of the PM data at the receiving end or the transmitting end of the optical transmission device for each of the plurality of optical transmission devices and for each of the receiving end and the transmitting end, When there is an optical transmission device in which the PM data at the receiving end or the transmitting end after abnormality recovery is outside the threshold range, the relearning determination unit determines that relearning of the AI model at the receiving end or the transmitting end of the corresponding optical transmission device is necessary. The controller according to claim 1.

3. The relearning determination unit instructs relearning for the AI model of the corresponding optical transmission device, which has been determined to require relearning. The controller according to claim 1.

4. For the optical transmission devices where the PM data after abnormal recovery is within the threshold range, the relearning determination unit determines that relearning of the AI model of the optical transmission device is unnecessary. The controller according to claim 1.

5. The threshold range is a range with the value obtained by adding the tolerance to the average value as the upper limit and the value obtained by subtracting the tolerance from the average value as the lower limit. The controller according to claim 1.

6. The collection unit periodically collects the PM data of each of the plurality of optical transmission devices. The controller according to claim 1.

7. An abnormality in the optical transmission network is an abnormality in any physical port of the plurality of optical transmission devices. The controller according to claim 1.

8. An abnormality in the optical transmission network is an abnormality in the optical fiber between the plurality of optical transmission devices. The controller according to claim 1.

9. A learning cost reduction method executed by a controller, comprising: collecting PM (Performance Monitoring) data representing the optical power level of the optical signal of each of the plurality of optical transmission devices provided in the optical transmission network; using the PM data of the optical transmission device as learning data by a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, and learning the temporal variation of the PM data of the optical transmission device; for each of the plurality of optical transmission devices, holding the average value and the tolerance of the PM data of the optical transmission device in a database; using the plurality of AI models to detect an abnormality in the optical transmission network; When there is an optical transmission device in which the PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, determine that the AI model of the corresponding optical transmission device needs to be retrained, A learning cost reduction method including.

10. On a computer, A procedure for collecting PM (Performance Monitoring) data representing the optical power level of each optical signal of a plurality of optical transmission devices provided in an optical transmission network, A procedure for learning the time-series variation of the PM data of the optical transmission device by using the PM data of the optical transmission device as learning data by a plurality of AI (Artificial Intelligence) models provided for each of the plurality of optical transmission devices, A procedure for holding, in a database, the average value and the allowable error of the PM data of the optical transmission device for each of the plurality of optical transmission devices, A procedure for detecting an abnormality in the optical transmission network by using the plurality of AI models, A procedure for determining that the AI model of the corresponding optical transmission device needs to be retrained when there is an optical transmission device in which the PM data after abnormality recovery is outside the threshold range derived from the average value and the allowable error, A program for causing the above to be executed.

Images