Abnormality management device and abnormality management method

Abstract

The present invention aims to more easily manage abnormalities in the movement path of a moving body. [Solution] The abnormality management device 1 includes a second learning unit 13 configured to: treat normal data converted by the conversion unit 12, which reflects the dependency between the sequences of travel routes, as true normal data; keep the generator parameters of a generator 131 fixed; generate pre-conversion pseudo-normal data that is sufficiently deviated from the distribution of true normal data; and update the classifier parameters of a classifier 132 that distinguishes true normal data from pseudo-normal data in a direction that maximizes an objective function based on the probability distribution of normal data and the probability distribution of abnormal data determined from the ratios estimated by the first learning unit 11, thereby training a generative model including the generator 131 and the classifier 132. The pseudo-normal data is also obtained by converting the pre-conversion pseudo-normal data so as to reflect the dependency.

Description

[Technical Field]

[0001] The present invention relates to an abnormality management device and an abnormality management method. [Background technology]

[0002] Conventional vehicle position estimation technologies for autonomous driving include signals from positioning satellites such as GPS, a six-axis inertial sensor (Inertial Measurement Unit: IMU) that detects vehicle behavior to complement signals from positioning satellites, and a distance measuring instrument (DMI) that measures the distance traveled by the vehicle by measuring the number of tire rotations.

[0003] In poor communication conditions, such as when a vehicle is traveling through a tunnel, it may not be possible to receive GPS signals from positioning satellites. However, even the IMU and DMI used to complement GPS may not provide sufficient accuracy in estimating vehicle position depending on the situation. For example, vehicle position estimation using an IMU has the disadvantage of being prone to error accumulation. When estimating vehicle position using DMI, measurement accuracy may decrease when the vehicle speed or direction changes.

[0004] Therefore, as in Patent Document 1, a technology has been proposed in which base stations are deployed along roads and the vehicle sends a location registration signal (notifying the number of the base station in which the vehicle is located) every time the vehicle crosses a base station area, thereby allowing the vehicle's location to be constantly notified. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2025-045963 Summary of the Invention [Problem to be solved by the invention]

[0006] However, although the network could grasp the vehicle's route by sending a location registration signal (notifying the base station number in which the vehicle is located), there was no way to determine whether the route deviated from the normal route.

[0007] The present invention has been made to solve the above-mentioned problems, and has an object to more easily manage abnormalities in the movement paths of moving bodies including vehicles. [Means for solving the problem]

[0008] In order to solve the above-described problems, the anomaly management device according to the present invention includes a first learning unit configured to learn parameters representing a ratio between a probability distribution of normal data indicating a sequence of communication areas passed by a mobile unit moving on a normal route in a moving space of the mobile unit defined by a plurality of communication areas and a probability distribution of abnormal data indicating a sequence of communication areas passed by a mobile unit moving on a route including an abnormal route deviating from the normal route, and to estimate the ratio based on the learned parameters; a first conversion unit configured to convert the normal data into normal data reflecting the dependency relationship using a sequence learning model having pre-set model parameters representing the dependency relationship between the sequences of communication areas passed by a mobile unit moving on a normal route; and a generator configured to generate pre-conversion pseudo-normal data before the dependency relationship is reflected, the pseudo-normal data corresponding to pseudo-normal data sufficiently deviating from the distribution of the true normal data, using the normal data reflecting the dependency relationship converted by the first conversion unit as true normal data. a second learning unit configured to update classifier parameters of a classifier that distinguishes between the true normal data and the pseudo-normal data in a direction that maximizes an objective function based on the probability distribution of the normal data and the probability distribution of the abnormal data determined from the ratio estimated by the first learning unit, while keeping generator parameters of a classifier fixed, thereby training a generative model including the generator and the classifier; a generation unit configured to generate the pre-conversion pseudo-normal data by the generator included in the generative model trained by the second learning unit; a second conversion unit configured to convert the pre-conversion pseudo-normal data generated by the generation unit into the pseudo-normal data reflecting the dependency relationship, using the sequence learning model; and a storage unit configured to store data that has been converted by the second conversion unit and reflects the dependency relationship, and that has been further identified as pseudo-normal data by the classifier included in the generative model trained by the second learning unit.

[0009] In addition, the abnormality management device of the present invention may further include an adjustment unit configured to adjust the model parameters of the sequence learning model according to the state of learning of the generation model by the second learning unit, wherein the first conversion unit is configured to convert the normal data into normal data reflecting the dependency relationship using the sequence learning model having the model parameters adjusted by the adjustment unit, and the second conversion unit is configured to convert the pre-conversion pseudo-normal data generated by the generation unit into the pseudo-normal data reflecting the dependency relationship using the sequence learning model having the model parameters adjusted by the adjustment unit.

[0010] In addition, the abnormality management device of the present invention may further include a collection unit configured to collect managed object data, which is a sequence of communication areas through which a mobile object passes during the movement of the managed object; a third conversion unit configured to convert the managed object data into managed object data reflecting the dependency using the sequence learning model; and a judgment unit configured to judge that the route of movement of the managed object is abnormal when the managed object data reflecting the dependency matches one of the pseudo-normal data stored in the memory unit and reflecting the dependency.

[0011] In addition, the abnormality management device of the present invention may further include a notification unit configured to notify the moving body that moved the managed object when the determination unit determines that the route of movement of the managed object is abnormal.

[0012] In the abnormality management device according to the present invention, the sequence learning model may be a recurrent neural network model.

[0013] In order to solve the above-mentioned problems, the anomaly management method of the present invention includes a first learning step of learning parameters representing a ratio between a probability distribution of normal data indicating a sequence of communication areas passed by a mobile unit moving along a normal route in a moving space of the mobile unit defined by a plurality of communication areas, and a probability distribution of abnormal data indicating a sequence of communication areas passed by a mobile unit moving along a route including an abnormal route deviating from the normal route, and estimating the ratio based on the learned parameters; a first conversion step of converting the normal data into normal data reflecting the dependency relationship using a sequence learning model having pre-set model parameters representing the dependency relationship between the sequences of communication areas passed by a mobile unit moving along a normal route; and a generator of generating pre-conversion pseudo-normal data before the dependency relationship is reflected, which corresponds to pseudo-normal data sufficiently deviated from the distribution of the true normal data, using the normal data reflecting the dependency relationship converted in the first conversion step as true normal data. a second learning step of learning a generative model including the generator and the classifier by updating classifier parameters of a classifier that distinguishes between the true normal data and the pseudo-normal data in a direction that maximizes an objective function based on the probability distribution of the normal data and the probability distribution of the abnormal data determined from the ratio estimated in the first learning step while keeping the generator parameters fixed; a generation step of generating the pre-conversion pseudo-normal data by the generator included in the generative model trained in the second learning step; a second conversion step of converting the pre-conversion pseudo-normal data generated in the generation step into the pseudo-normal data reflecting the dependency relationship using the sequence learning model; and a storage step of storing, in a storage unit, data of the pseudo-normal data converted in the second conversion step and reflecting the dependency relationship that is further identified as pseudo-normal data by the classifier included in the generative model trained in the second learning step.

[0014] Furthermore, the anomaly management method according to the present invention may further include an adjustment step of adjusting the model parameters of the sequence learning model according to the state of learning of the generative model in the second learning step, wherein the first conversion step uses the sequence learning model having the model parameters adjusted in the adjustment step to convert the normal data into normal data reflecting the dependency, and the second conversion step uses the sequence learning model having the model parameters adjusted in the adjustment step to convert the pre-conversion pseudo-normal data generated in the generation step into the pseudo-normal data reflecting the dependency.

[0015] In addition, the abnormality management method of the present invention may further include a collection step of collecting managed object data, which is a sequence of communication areas through which a mobile object passes during movement of the managed object; a third conversion step of converting the managed object data into managed object data reflecting the dependency relationship using the sequence learning model; and a determination step of determining that the route of movement of the managed object is abnormal if the managed object data reflecting the dependency relationship matches one of the pseudo-normal data stored in the memory unit and reflecting the dependency relationship.

[0016] Furthermore, the abnormality management method according to the present invention may further include a notification step of notifying the moving body that moved the managed object if the route of movement of the managed object is determined to be abnormal in the determination step.

[0017] In the anomaly management method according to the present invention, the sequence learning model may be a recurrent neural network model. [Effects of the Invention]

[0018] According to the present invention, normal data converted by the first conversion unit, which reflects the inter-sequence dependency of communication areas, is regarded as true normal data, and while fixing generator parameters of a generator that generates pre-conversion pseudo-normal data that is sufficiently deviated from the distribution of true normal data, a classifier parameter of a classifier that distinguishes true normal data from pseudo-normal data is updated in a direction that maximizes an objective function based on the probability distribution of normal data and the probability distribution of abnormal data determined from the ratios estimated by the first learning unit, thereby training a generative model including the generator and the classifier. Therefore, anomalies can be managed even when there is little abnormal data. [Brief explanation of the drawings]

[0019] [Figure 1] FIG. 1 is a block diagram showing the configuration of an abnormality management system including an abnormality management device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram for explaining a movement space targeted by the abnormality management device according to this embodiment. [Figure 3] FIG. 3 is a schematic diagram showing an example of the configuration of a sequence learning model used by the abnormality management device according to this embodiment. [Figure 4] FIG. 4 is a diagram for explaining the second learning unit and the conversion unit included in the abnormality management device according to this embodiment. [Figure 5] FIG. 5 is a diagram for explaining the second learning unit included in the abnormality management device according to this embodiment. [Figure 6] FIG. 6 is a diagram for explaining the second learning unit included in the abnormality management device according to this embodiment. [Figure 7] FIG. 7 is a diagram for explaining the generating unit included in the abnormality management device according to this embodiment. [Figure 8] FIG. 8 is a block diagram showing the hardware configuration of the abnormality management device according to this embodiment. [Figure 9] FIG. 9 is a flowchart showing the operation of the abnormality management device according to this embodiment. [Figure 10]FIG. 10 is a flowchart showing the operation of the abnormality management device according to this embodiment. [Figure 11A] FIG. 11A is a flowchart showing the operation of the abnormality management device according to this embodiment. [Figure 11B] FIG. 11B is a flowchart showing the operation of the abnormality management device according to this embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0020] Preferred embodiments of the present invention will now be described in detail with reference to FIGS. 1 to 11B.

[0021] 1 is a block diagram showing the configuration of an abnormality management system including an abnormality management device 1 according to an embodiment of the present invention. The abnormality management system according to this embodiment determines an abnormality in the route traveled by a mobile object 2 in a movement space defined by a plurality of communication areas A1 to An, using a location registration signal sent by the mobile object 2 when the mobile object 2 crosses one of the communication areas A1 to An.

[0022] [Configuration of anomaly management system] First, an overview of an abnormality management system including an abnormality management device 1 according to an embodiment of the present invention will be described. As shown in Fig. 1, the abnormality management system includes, for example, an abnormality management device 1 compatible with an SA 5G wireless communication system, a mobile object 2, base stations BS1 to BSn, and a core network 3.

[0023] The base stations BS1 to BSn are composed of wireless base stations compatible with the 5G system, and relay communications between mobile units 2 located in the communication areas A1 to An and the core network 3. Hereinafter, when the base stations BS1 to BSn and the communication areas A1 to An are not distinguished from one another, they may be collectively referred to as the base stations BS and the communication areas A, respectively.

[0024] As shown in Fig. 1, communication areas A1 to An of base stations BS1 to BSn define a moving space in which a mobile unit 2 moves. Each of the communication areas A1 to An is arranged to cover a section of a road, such as a main road, along which the mobile unit 2 moves from an initial point to a destination point. In this embodiment, the communication areas A1 to An are assumed to have cells of the same size.

[0025] Figure 2 is a diagram that schematically illustrates a mobile space targeted by the anomaly management system. Each dotted circle in Figure 2 represents a communication area A1 to A16 covered by each base station BS1 to BS16 arranged in the mobile space. A mobile unit 2 moves, for example, from an initial point S to the position of the communication area A16 of a destination point G, using the positions of the communication areas A1 to A16 of base stations BS1 to BS16 as waypoints.

[0026] In this embodiment, the mobile object 2 may turn right, turn left, or go straight for each of the communication areas A1 to A16 according to route instructions from some system, and arrive at the destination point G from the initial point S. The route instructions may also be to turn right or left or go straight based on the direction of travel of the mobile object 2. For example, if a route instruction to go straight is given at the position of communication area A1, the mobile object 2 goes straight from the position of communication area A1 and arrives at the next communication area A2. Furthermore, if a route instruction to turn left is given at the position of communication area A2, the mobile object 2 turns left and arrives at the next communication area A6. In this way, the mobile object 2 can move through the movement space according to the route instructions for each of the communication areas A1 to A16.

[0027] The mobile object 2 is equipped with a communication terminal 20. In this embodiment, the mobile object 2 is assumed to be a vehicle including an automobile, a motor vehicle, a motorcycle, etc. In another embodiment, the mobile object 2 may be a person traveling on foot who carries the communication terminal 20, or the communication terminal 20 itself, but is not limited to this. The communication terminal 20 includes a processor, a main memory device, an auxiliary memory device, a communication interface, etc., and is realized as a terminal device equipped in the mobile object 2, or a mobile communication terminal such as a smartphone of a user who uses the mobile object 2, a tablet computer, etc.

[0028] Specifically, the communication terminal 20 includes a SIM 21. The mobile unit 2 is uniquely identified by an International Mobile Subscriber Identity (IMSI) of the SIM 21 included in the communication terminal 20.

[0029] When the communication terminal 20 crosses over the communication areas A1 to An as the mobile object 2 moves, the processor of the communication terminal 20 transmits a location registration signal (TAU) to the base station BS1 to BSn of the new communication area A1 to An.

[0030] The moving object 2 also includes an ECU (Electronic Control Unit) 22, which can process route instructions from some system as steering control, drive control, brake control, and automatic driving control of the moving object 2. The moving object 2 can also include a GPS module with a GPS function (not shown), a car navigation system, and various sensors such as a camera and LiDAR.

[0031] The fault management device 1 and the core network 3 are connected via a network NW such as a LAN or a WAN. In addition, the base stations BS1 to BSn constituting the wireless access network are connected to the core network 3 via a network L such as a backhaul link.

[0032] The core network 3 includes an Access and Mobility Management Function (AMF) 30, a Unified Data Management (UDM) 31, and a Unified Data Repository (UDR) 32, which are nodes in the C-plane. The core network 3 also includes a User Plane Function (UPF) 33, which is a node in the U-plane. The fault management device 1 acquires location information of the communication area A associated with a location registration signal from the mobile object 2 via a communication interface 32a of the UDR 32. The fault management device 1 can also send instructions for the determined route to some system of the mobile object 2 via a communication interface 33a of the UPF 33.

[0033] For example, as shown in Figure 1, when a mobile unit 2 crosses from the communication area A1 of base station BS1 to the communication area A2 of base station BS2, the communication terminal 20 transmits a location registration signal to make a location registration request to UDM31 via base station BS2 and AMF30.

[0034] The AMF 30 transmits the received signal to the UDM 31, and the UDM 31 performs location registration using the terminal identification information of the communication terminal 20 provided in the mobile object 2. Furthermore, the location registration signal and terminal identification information transmitted by the communication terminal 20 provided in the mobile object 2 are stored in the UDR 32 together with identification information relating to the base station BS and communication area A in which the mobile object 2 is located, and a transmission timestamp (information indicating the date and time) of the location registration signal.

[0035] The abnormality management device 1 according to this embodiment acquires, via the network NW, identification information of each base station BS that has received a location registration signal or identification information of the communication area A, which is associated with the location registration signal stored in the UDR 32. The abnormality management device 1 can also store, as setting information, in a storage unit 19 described below, location information such as GPS coordinates of the longitude and latitude of the base stations BS and communication area A arranged in the mobile space. In this embodiment, the location of the mobile unit 2 is treated as the location of the base station BS in the communication area A in which the mobile unit 2 is located. In the following description, the location of the communication area A refers to the location of the corresponding base station BS.

[0036] In this way, by acquiring the location registration signal from the mobile object 2 from the UDR 32, the abnormality management device 1 can acquire the location of the mobile object 2 at the time indicated by the timestamp of the location registration signal.

[0037] [Function block of the abnormality management device] Next, functional blocks of the fault management device 1 according to this embodiment will be described with reference to the block diagram of Fig. 1. As shown in Fig. 1, the fault management device 1 includes a collection unit 10, a first learning unit 11, conversion units (first conversion unit, second conversion unit, third conversion unit) 12, a second learning unit 13, an adjustment unit 14, a generation unit 15, a fault path database 16, a determination unit 17, a notification unit 18, and a storage unit 19.

[0038] The collection unit 10 acquires the location of communication area A associated with a location registration signal from the mobile object 2, which is transmitted when the mobile object 2 crosses communication area A, as the current location of the mobile object 2. More specifically, the collection unit 10 acquires the location registration signal and information associated with the location registration signal transmitted by the communication terminal 20 of the mobile object 2 from the communication interface 32a of the UDR 32 via the network NW. The information associated with the location registration signal includes the SIM 21 of the communication terminal 20, the identification information of the base station BS that received the location registration signal or the identification information of the communication area A, and the timestamp of the time the location registration signal was transmitted.

[0039] The collection unit 10 can refer to the location information of the GPS coordinates of the communication areas A1 to An associated with the identification information of the base stations BS1 to BSn or the identification information of the communication areas A1 to An stored in the memory unit 19, and obtain this as the location of the mobile unit 2 at the time the location registration signal was transmitted.

[0040] The collection unit 10 collects normal data, which is a sequence of communication areas passed by a mobile unit 2 traveling along a normal route. For example, assuming that the route along the arrow in FIG. 2 is the normal route, the sequence of communication areas passed by a mobile unit 2 traveling along the normal route, i.e., the normal data, may be data of [1, 2, 6, 10, 14, 15, 16], but is not limited to this. Note that the collection unit 10 can also acquire the communication area of ​​the initial point of the route ("1" in the above example) from a past location registration signal. The collection unit 10 also collects abnormal data, which is a sequence of communication areas passed by a mobile unit 2 traveling along a route that deviates from the normal route and includes an abnormal route. The collection unit 10 can collect, from multiple mobile units 2, the sequence of communication areas passed by the mobile units. The data of the sequence of communication areas is associated with information identifying the mobile unit 2 (for example, the terminal identification information of the communication terminal 20 of the mobile unit 2, but is not limited to this).

[0041] The collection unit 10 can label the collected communication area sequences based on a rule base or a statistical threshold, classify the data into normal data and abnormal data, and collect these data.

[0042] The collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a route from a predetermined initial point (which may be a predetermined communication area; the same applies hereinafter) to a destination point (which may be a predetermined communication area; the same applies hereinafter), where the travel time is within a predetermined time. In other words, a route that allows travel from the predetermined initial point to the destination point within a predetermined time may be considered to be a normal route. Alternatively, the collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a route from the predetermined initial point to the destination point where the number of passed communication areas is a predetermined number or less. In other words, a route that allows travel from the predetermined initial point to the destination point by passing through a predetermined number or less of communication areas may be considered to be a normal route. Alternatively, the collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a route from the predetermined initial point to the destination point where the route passes through predetermined communication areas. In other words, a route that allows travel from the predetermined initial point to the destination point via predetermined communication areas may be considered to be a normal route. Alternatively, the collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a route from a predetermined initial point to a destination point that does not pass through a predetermined communication area. In other words, a route that allows travel from the predetermined initial point to the destination point without passing through a predetermined communication area may be considered a normal route. Alternatively, the collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a route from a predetermined initial point to the destination point that does not include overlapping communication areas. In other words, a route that allows travel from the predetermined initial point to the destination point that does not include overlapping communication areas may be considered a normal route. Alternatively, the collection unit 10 can collect, as normal data, a sequence of communication areas passed by the mobile unit 2 along a specified route from a predetermined initial point to the destination point.Alternatively, the collection unit 10 can collect, as normal data, a series of communication areas that are calculated to be passed by the mobile object 2 on a specified route from a predetermined initial point to a destination point, rather than a series of communication areas actually collected from the mobile object 2. In other words, the specified route from a predetermined initial point to a destination point can be considered to be a normal route. The above-described series of communication areas that can be collected as normal data are merely examples, and the series of communication areas that can be collected as normal data are not limited to these.

[0043] The collection unit 10 can collect, as abnormal data, a series of communication areas through which the mobile object 2 has passed that is not collected as normal data.

[0044] It is possible to collect a large amount of normal data, but abnormal data occurs very rarely, making it difficult to collect a sufficient amount. Therefore, the amount of abnormal data collected by the collection unit 10 is significantly smaller than the amount of normal data (normal data >> abnormal data).

[0045] The collection unit 10 collects, as training data for the first learning unit 11, sequence data of communication areas passed by a mobile object 2 traveling on a normal route (the normal data) and sequence data of communication areas passed by a mobile object 2 traveling on a route including a route deviating from the normal route (the abnormal data). The collection unit 10 also collects the sequence data of communication areas passed by a mobile object 2 traveling on a normal route (the normal data) as data to be converted by the conversion unit 12. The collection unit 10 also collects sequence data of communication areas passed by a mobile object 2 traveling on a normal route (managed object data) during the movement of the managed object, which is the target of abnormality judgment by the judgment unit 17.

[0046] The first learning unit 11 learns a parameter representing the ratio between the probability distribution of normal data, which is a series of communication areas through which the moving body 2 moving along a normal path passes, and the probability distribution of abnormal data, which is a series of communication areas through which the moving body 2 moving along a path deviating from the normal path passes, and estimates the ratio based on the learned parameter. The first learning unit 11 uses a series of communication areas through which the moving body 2 moving along a normal path that can be obtained in large quantities passes, and a series of communication areas through which the moving body 2 moving along a path including a path deviating from the normal path that can be obtained relatively in small quantities, to learn a parameter representing the density ratio of the probability density functions, which are the probability distributions of normal data and abnormal data. Also, the first learning unit 11 estimates the density ratio from the learned parameter.

[0047] Here, let the set of training data that is normal data be D = {x (1) , x (2) , …, x (N)}, and the set of training data that is abnormal data be D’ = {x’ (1) , x’ (2) , …, x’ (N’)}. Each observed data x (n) (1 ≤ n ≤ N) may be data collected by the collection unit 10 at the n-th time, and n may represent the more past as it is smaller. Also, each observed data x (n) (1 ≤ n ≤ N) is M-dimensional, and x (n) = (x1 (n) , x2 (n) , …, x M (n) ). An example of M is 10000, but it is not limited to this. Each component x m (n) (1 ≤ m ≤ M) is information (for example, it may be the corresponding base station number, but it is not limited to this.) for specifying the m-th communication area passed by the moving body 2 when moving from a predetermined initial point to a target point. In addition, when the number of communication areas passed by the moving body 2 until it reaches a predetermined target point is m’ (< M), x m (n)(m'≦m≦M) may be filled with a predetermined value (for example, it may be zero, but is not limited to this). (n‘) The same applies to (1≦n'≦N').

[0048] In the following, D is called normal data and D' is called abnormal data, and the probability density function of normal data D is called p(x) and the probability density function of abnormal data D' is called p'(x). The probability density function p(x) of normal data D indicates the distribution of the probability that a sequence x is observed in a communication area through which a mobile unit 2 passes when traveling on a normal route. Furthermore, the probability density function of abnormal data D' indicates the distribution of the probability that a sequence x is observed in a communication area through which a mobile unit 2 passing on a route including an abnormal route passes. The density ratio r(x) between the probability density function p(x) of normal data D and the probability density function p'(x) of abnormal data D' is expressed by the following equation (1).

number

[0049]

number

[0050] The basis function ψ(x) is defined by the RBF (Radial Basis Function) kernel and expressed by the following equation (3).

number

[0051] Here, based on the above formula (1), the specific form of the above formula (3) in which the number of bases b is the number of training data N (b=N) is given by the following formula (4).

number

[0052] The above equation (4) expresses the density ratio as a linear sum of RBFs centered on all training points. Here, we introduce the generalized Kullback-Leibler divergence (KL divergence), which measures the information-theoretic distance between non-negative functions f and g, as shown in the following equation (5).

number

[0053] In density ratio estimation, f=p(x) and g=r θ Substituting p'(x) into the above equation (5), the following equation (6) is used as the objective function.

number

[0054] In the above equation (6), each x n , x' n’ The optimization objective function obtained by approximating the integral with an empirical distribution that sets values ​​other than θ to 0, ignoring terms that do not depend on the parameter θ, and dropping constants is expressed as the following equation (7).

number

[0055] By minimizing J(θ) in equation (7), the density ratio r θ Since J(θ) is a convex function, the first learning unit 11 updates the parameter θ from the initial value until convergence using the parameter θ update formula by the gradient descent method of the following formula (8).

number

[0056] The result of specifically calculating the gradient of the above formula (7) is expressed by the following formula (9).

number

[0057] The first term in the above equation (9) represents the contribution from abnormal data, and the second term represents the contribution from normal data. Since the second term is dominant in the above equation (9), stable estimation is possible when the number of normal data N is large. In other words, even when the amount of abnormal data N' is small, the parameter θ can be stably calculated.

[0058] Before calculating the optimal solution of the parameter θ using the above equations (7) to (9), the first learning unit 11 calculates an appropriate value for the bandwidth h in the above equation (4) by cross-validation, information criterion (KL divergence minimization criterion), etc. Based on the optimal solution of the parameter θ calculated by the KL density ratio estimation method, the first learning unit 11 calculates the density ratio r for an arbitrary input x using the above equation (2). θ The density ratio estimated by the first learning unit 11 is passed to the second learning unit 13, which will be described later.

[0059] The converter 12 is used as a feature conversion model for converting the sequence data of communication areas handled by a generative model in the second learning unit 13, which will be described later, into data that reflects features that represent the dependency relationships between the sequence data. The converter 12 uses a sequence learning model that includes pre-set model parameters that represent the dependency relationships between the sequence data of communication areas to convert normal data into normal data that reflects the dependency relationships.

[0060] Furthermore, the converter 12 uses a sequence learning model to convert pseudo-normal data (pre-conversion pseudo-normal data) generated by the generator 15 (described later) into pseudo-normal data that reflects the dependency relationships between the sequence data of the communication areas. Furthermore, the converter 12 uses a sequence learning model having model parameters adjusted by the adjuster 14 (described later) to convert normal data into normal data that reflects the dependency relationships between the sequence data of the communication areas. Similarly, the converter 12 converts the pre-conversion pseudo-normal data generated by the generator 15 into pseudo-normal data that reflects the dependency relationships between the sequence data of the communication areas. Furthermore, the converter 12 (third converter) converts the managed object data, which is a sequence of communication areas through which the mobile object 2 passed during the movement of the managed object collected by the collector 10, into managed object data that reflects the dependency relationships between the sequence data of the communication areas, and passes the converted data to the determiner 17.

[0061] The sequence learning model is a model that sequentially inputs sequence data of communication areas that constitute an input sequence, and generates feature representations that consider the relationship between the preceding and following sequence data while sequentially updating the internal state. The sequence learning model is a model equipped with a recurrent neural network (RNN), a long short-term memory (LSTM), or a self-attention mechanism, but is not limited to these.

[0062] When a sequence (normal data) of communication areas passed through when a mobile object 2 moves along a normal route is input to the sequence learning model, the sequence learning model performs calculations on the input sequence based on the model parameters, and outputs as an output sequence normal data represented as data in which dependencies in the sequence have been extracted or emphasized while preserving the sequence features contained in the input sequence. As shown in Figure 4, by providing the sequence learning model on the output side of generator 131 and on the input side of classifier 132 included in the generative model described below, it becomes possible to generate data in the generative model while preserving the features between the sequence data.

[0063] Here, an example will be described in which an RNN is used as a sequence learning model. FIG. 3 is a schematic diagram showing the network structure of an RNN. As shown in FIG. 3, the RNN is configured as a neural network consisting of an input layer X, a hidden layer H which is a memory cell, and an output layer Y. Each node in FIG. 3 represents an input x at each time t (which in this embodiment may be considered as an "order"; the same applies below). (t) , hidden state (internal state) h (t) , and output y (t) First, input x (t) and the hidden state h (t-1) are multiplied by the corresponding weight matrices U and W, respectively, and a bias term b is added, thereby performing the linear operation shown in the following equation (10).

[0064]

number

[0065] In the above equation (10), a (t) is the intermediate value for updating the hidden state at time t, and represents the weighted sum of the input and past state information. (t) By applying the nonlinear activation function σ(·) to the new hidden state h, as shown in the following equation (11), (t) is obtained.

[0066]

number

[0067] The hidden state h in the above equation (11) (t) is an internal representation that retains past sequence information while reflecting features based on the current input, and is carried over to the processing at the next time t+1. Furthermore, this hidden state h (t) Based on this, the output y (t) is calculated.

[0068]

number

[0069] In the above equation (12), V is the weight matrix from the hidden state to the output, and c is the bias term for the output layer. In this way, RNNs recursively use the hidden state from the previous time in the input processing of the next time, transmitting information from the hidden state from the previous time to the hidden state from the next time, so they are structured to generate outputs that reflect the temporal dependencies inherent in time-series data.

[0070] Pre-set values ​​are used for the model parameters of the sequence learning model. In the RNN of FIG. 3, values ​​set in advance based on empirical rules are used for the model parameters U, W, V, b, and c. When the model parameter values ​​are adjusted by the adjustment unit 14 (described later), the conversion unit 12 performs conversion processing based on the sequence learning model in which the adjusted model parameters (U', W', V', b', c') are set.

[0071] The second learning unit 13 regards the normal data converted by the conversion unit 12 (first conversion unit) reflecting the dependency between the series data in the communication areas as true normal data, and updates the classifier parameters of the classifier 132 that distinguishes between true normal data and pseudo-normal data in the direction of maximizing an objective function based on the probability density function (probability distribution) of the normal data and the probability density function (probability distribution) of the abnormal data determined from the density ratio (ratio) estimated by the first learning unit 11, while keeping the generator parameters of the generator 131 that generates pre-conversion pseudo-normal data that is sufficiently deviated from the distribution of the true normal data fixed, thereby learning a generative model.

[0072] As shown in Fig. 4, the second learning unit 13 executes a Max learning phase of a GAN (Generative Adversarial Network) as a generative model having a generator 131 and a classifier 132, in which only the parameters of the classifier 132 are updated while the generator 131 is fixed. When the sequence of communication areas converted by the conversion unit 12 is regarded as true normal data, the second learning unit 13 aims to generate pseudo-normal data that is sufficiently deviated from the distribution of normal data, that is, a sequence of abnormal communication areas. For this reason, the Min learning phase in the normal GAN ​​adversarial learning procedure, in which the generator 131 is updated in the minimization direction, is not performed.

[0073] As shown in FIG. 4, the generation model according to this embodiment, which includes a generator 131 and a classifier 132, generates a sequence of communication areas (hereinafter also referred to as a "sequence of normal communication areas") x=(x1, x2, ..., x M ) is used as training data for learning. Here, pseudo-normal data that deviates sufficiently from the distribution of normal data is a generated sequence that, in terms of statistical properties, is located in a region that significantly deviates from the normal range of normal data, compared to normal data that indicates a sequence from a normal communication area. When the index used for distance or deviation from normal data is log-likelihood, a sequence with a smaller likelihood corresponds to the pseudo-normal sequence. When the index used is cross-entropy, the larger the entropy value, the more deviation the sequence. Furthermore, when the KL distance is used as the index, a sequence with a large deviation in the overall distribution is treated as a sequence that deviates sufficiently.

[0074] 4, a conversion unit 12 equipped with a sequence learning model is provided on the input side of the classifier 132 of the generative model and on the output side of the generator 131. An adjustment unit 14 that adjusts the model parameters of the sequence learning model is provided between the objective function 135 of the generative model and the conversion unit 12.

[0075] 5 and 6 are diagrams schematically illustrating the neural network configuration of the generator 131 and the classifier 132 of the generative model used by the second learning unit 13. As shown in FIG. 5, the generator 131 is configured as a neural network having an input layer, a hidden layer, and an output layer. The generator 131 is a model that generates pseudo-normal data from random noise. For example, m randomly sampled vectors of Gaussian noise (z1 to z m ).

[0076] The generator 131 outputs the output G(z) after performing a product-sum operation on the input and weight parameters and threshold processing using an activation function. The output G(z) from the generator 131 is pseudo-normal data that deviates from the distribution of true normal data. CNN or ResNet can be used as the neural network that constitutes the generator 131.

[0077] The classifier 132 shown in Fig. 6 is configured with a neural network having an input layer, a hidden layer, and an output layer. In the example of Fig. 6, as the training data input, a sequence of normal communication areas converted by the conversion unit 12 and reflecting the dependency between the sequence data is given as true normal data. Each input node shown in Fig. 6 is provided with a sequence (x (1) ,x (2) ,…,x (N) ) of each communication area sequence x (n) Output values ​​x1~x of the sequence learning model for (1≦n≦N) M are input respectively.

[0078] The classifier 132 outputs a probability value between 0 and 1 after performing a product-sum operation on the input and weight parameters and threshold processing using an activation function. When the classifier 132 correctly identifies the training data related to the input true normal data as true normal data, it outputs a value close to the output y=1. On the other hand, when the classifier 132 correctly identifies the training data related to the input pseudo normal data as pseudo normal data, it outputs a value close to the output y=0. In this way, the classifier 132 is a model that distinguishes the model distribution generated by the generator 131 from the data distribution of the training data, which is the true distribution. A CNN can be used as the neural network that constitutes the classifier 132.

[0079] As shown in the block diagram of FIG. 4, the generator 131 of the generative model adopted by the second learning unit 13 is represented as a function G, and the classifier 132 is represented as a function D. Furthermore, true normal data is represented as x, the predicted value output by the classifier 132 is represented as y, and the correct label is represented as t. The correct label t is set to 1 for true normal data and 0 for pseudo-normal data generated by the generator 131. In this case, the classifier 132 calculates the cross entropy E CE It can be expressed as:

[0080]

number

[0081] The first term in the brace of the above equation (13) represents t n lny n In this case, the predicted value y n is the correct label of the true normal data, t n = 1. On the other hand, the second term in the braces represents (1-t n )ln(1-y n ), the predicted value y n is the correct label value (1-t n ) = 0. In this way, the cross entropy E CE is the maximum value when the predicted value matches the correct label value.

[0082] Here, the generator 131 that configures the generative model uses parameters w G ,θ G and the function G(w G ,θ G ) The classifier 132 uses the parameter w D ,θ D and function D(w D ,θ D ) The cross entropy E in the above equation (13) CE The loss function (objective function E) of the generative model including the generator 131 and the discriminator 132 based on the above can be expressed by the following equation (14).

number

[0083] The first term of the above equation (14) represents E D(x)=1 lnD(w D ,θ D ) is the expected value at which the classifier 132 classifies true normal data as true normal data. D(x)=0 ln(1-D(G(w G ,θ G ),w D ,θ D )) is the expected value at which the classifier 132 classifies the pseudo-normal data generated by the generator 131 as pseudo-normal data. Here, the expected value of the above formula (14) can be expressed as the following formula (15) using a probability distribution.

[0084]

number

[0085] Here, for the probability density function p(x) of normal data and the probability density function p'(x) of abnormal data in the above equation (1), in order to form a probabilistic labeled classification problem for Max optimization learning of GAN, let p(x)≡ρ(x|y=1) and p'(x)≡ρ(x|y=0). The density ratio r estimated by the first learning unit 11 is θ (x) is defined by the following equation (16).

number

[0086] The probability density function ρ(x|y=1), which is the conditional probability distribution of normal data in the above equation (16), can be calculated from a large amount of normal data. Using the calculated probability density function ρ(x|y=1) of normal data, the probability density function ρ(x|y=0), which is the conditional probability distribution of abnormal data, can be expressed by the following equation (17).

number

[0087] The probability density function ρ(x|y=1) of normal data and the probability density function ρ(x|y=0) of abnormal data in the above formula (17) are substituted into the objective function E in the above formula (15), and set values ​​are used for the prior probability ρ(y=1) of the normal (y=1) class and the prior probability ρ(y=0) of the abnormal (y=0) class. For example, the prior probability values ​​are arbitrarily set as ρ(y=1):ρ(y=0)=0.99:0.01, and these prior probability values ​​can be adjusted as needed. Furthermore, the posterior probability ρ(y=1|x) of normality for the observed communication area sequence x is calculated as D(w D ,θ D ) and the posterior probability ρ(y=0|x) of an anomaly for the sequence x in the communication area is 1-D(G(w G ,θ G ),w D ,θ D ) Each posterior probability corresponds to the density ratio r θ and can be obtained from the prior probability.

[0088] Here, when the generator 131 is fixed, the objective function E becomes a maximization problem of the following equation (18) with respect to the discriminator 132.

number

[0089] In training the generative model of this embodiment, as described above, only Max optimization of the objective function E is performed, and the parameters of the discriminator 132 with the generator 131 fixed are trained. This prevents the output of the generator 131 from converging to the distribution of normal data, and instead maintains a sequence that is sufficiently deviated from the distribution of normal data. Once the update of the discriminator 132 has converged, the pseudo-normal data sequence output by the fixed generator 131 and converted by the conversion unit 12 has a low likelihood compared to a sequence in a normal communication area. At this time, the generator 131 can generate pseudo-normal data having the statistical properties expressed by the following equation (19).

number

[0090] The adjustment unit 14 adjusts the model parameters of the sequence learning model according to the state of learning of the generative model by the second learning unit 13. As described above, the model parameters of the sequence learning model are set as initial values ​​based on predetermined values ​​such as empirical rules. As shown in FIG. 4, the adjustment unit 14 monitors the loss transition and output distribution of the classifier 132 during the optimization learning process of the generative model by the second learning unit 13, and adjusts the values ​​of the model parameters of the sequence learning model based on the evaluation (the arrow between the objective function 135 and the conversion unit 12 in FIG. 4). The adjustment unit 14 performs an evaluation, for example, every time the loss of the classifier 132 is calculated, and adaptively adjusts the values ​​of the model parameters of the sequence learning model based on the evaluation result, independently of the gradient propagation of the generative model. Adjusting the model parameters of the sequence learning model by the adjustment unit 14 can improve the learning accuracy of the generative model.

[0091] The generation unit 15 generates pseudo-normal data (pre-conversion pseudo-normal data) using a generator 131' included in the generative model trained by the second learning unit 13. As shown in FIG. 7, the generation unit 15 inputs noise to the trained generator 131' to generate a large amount of pre-conversion pseudo-normal data. The pre-conversion pseudo-normal data generated by the generation unit 15 is passed to the conversion unit 12. Furthermore, the conversion unit 12 connected to the output side of the trained generator 131' converts the pre-conversion pseudo-normal data into pseudo-normal data that reflects the inter-sequence dependency between communication areas through the calculation of the sequence learning model.

[0092] The communication area anomaly sequence database 16 stores pseudo-normal data that reflects the dependency relationships between communication area sequences converted by the conversion unit 12 (second conversion unit). More specifically, the communication area anomaly sequence database 16 stores pseudo-normal data that the conversion unit 12 converts pre-conversion pseudo-normal data that is generated in large quantities by the learned generator 131' in the generation unit 15. The communication area anomaly sequence database 16 accumulates the pseudo-normal data as sequences of abnormal communication areas and constructs a database of abnormal communication area sequences used in the anomaly determination process. The communication area anomaly sequence database 16 can store only pseudo-normal data that is determined as 0 (abnormal) >> 1 (normal) from the pseudo-normal data converted by the conversion unit 12 by the learned classifier 132'.

[0093] The determination unit 17 determines that the movement route of the managed object is abnormal when the communication area series of the managed object (managed object data) converted by the conversion unit 12 and reflecting the dependency relationships between the communication area series matches one piece of pseudo-normal data reflecting the dependency relationships between the communication area series and stored in the communication area abnormality series database 16. The determination unit 17 determines that there is a match when the difference between the managed object data reflecting the dependency relationships and one piece of pseudo-normal data reflecting the dependency relationships and stored in the communication area abnormality series database 16 is zero or within a certain range, and can determine that the movement route of the managed object is abnormal.

[0094] Pseudo-normal data (1 to L) (L is a positive integer of 2 or more) composed of a series of communication areas of each 1 to D (D is a positive integer of 2 or more), which are stored in the abnormal communication area series database 16, and one of the series of communication areas (converted management target data) in which the dependency is reflected regarding the movement of the management target is represented as follows, respectively. [Number]

[0095] The determination unit 17 compares the communication area indicated by 1 to D of one of the converted management target data with the communication areas indicated by 1 to D of each of the pseudo-normal data, and if they match, determines that the path of the movement of the management target is abnormal. It is possible to output a determination result by comparing the communication areas in a part of the processes from 1 to n (n < D) among the processes of 1 to D.

[0096] When the determination unit 17 determines that the path of the movement of the management target is abnormal, the notification unit 18 sends out the notification. The notification unit 18 can send out, for example, a notification that the path of the movement of the management target is abnormal to the communication terminal 20 of the moving body 2 that has moved the management target via the network NW.

[0097] The storage unit 19 stores the parameter θ and the density ratio r estimated by the learning by the first learning unit 11. θ The storage unit 19 also stores the generator 131 and the discriminator 132 included in the learned generation model, and the model parameters of the sequence learning model adjusted by the adjustment unit 14.

[0098] [Hardware Configuration of the Abnormal Management Device] Next, an example of the hardware configuration for realizing the abnormal management device 1 having the above-described functions will be described using FIG. 8.

[0099] 8, the fault management device 1 can be realized by, for example, a computer including a processor 102, a main memory device 103, a communication interface 104, an auxiliary memory device 105, and an input / output (I / O) 106 connected via a bus 101, and a program for controlling these hardware resources. Furthermore, the fault management device 1 includes a display device 107.

[0100] The processor 102 is a circuit or device that performs arithmetic processing, and is realized by, for example, a general-purpose central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), etc. Alternatively, the processor 102 may be configured by combining some or all of these.

[0101] The main memory device 103 is configured, for example, by a volatile random access memory (RAM), and pre-stores programs for the processor 102 to perform various controls and calculations. The processor 102 and the main memory device 103 realize the functions of the abnormality management device 1, such as the collection unit 10, first learning unit 11, conversion unit 12, second learning unit 13, adjustment unit 14, determination unit 17, and notification unit 18 shown in FIG.

[0102] The communication interface 104 is an interface circuit for connecting the abnormality management device 1 to various external electronic devices via a network.

[0103] The auxiliary storage device 105 is composed of a readable / writable storage medium and a drive for reading and writing various information such as programs and data from and to the storage medium. The auxiliary storage device 105 can use non-volatile storage such as a hard disk or flash memory as the storage medium.

[0104] The auxiliary storage device 105 has a program storage area for storing an abnormality management program. The auxiliary storage device 105 also has a program storage area for storing parameters representing the density ratio of the probability density functions of normal data and abnormal data executed by the abnormality management device 1, and a first learning program for estimating the density ratio. The auxiliary storage device 105 also has a program storage area for storing a program for the conversion unit 12 to perform calculations of a sequence learning model. The auxiliary storage device 105 also has a program storage area for storing a second learning program for training the classifier 132 of the generative model executed by the abnormality management device 1.

[0105] The auxiliary storage device 105 realizes the communication area anomaly series database 16 and the storage unit 19 described in Fig. 1. Furthermore, for example, it may have a backup area for backing up the above-mentioned data, programs, etc.

[0106] The input / output I / O 106 is an input / output device that inputs signals from external devices and outputs signals to external devices.

[0107] The display device 107 is configured by an organic EL display, a liquid crystal display, etc. The display device 107 can display any information such as the travel route of the moving object 2 and the series of communication areas on the screen.

[0108] [Operation of the abnormality management device] Next, the operation of the abnormality management device 1 having the above-described configuration will be described with reference to the flowcharts of FIGS. 9 to 11B.

[0109] As shown in FIG. 9, first, the collection unit 10 collects normal data, which is a sequence of communication areas passed by a mobile unit 2 traveling on a normal route, and abnormal data, which is a sequence of communication areas passed by a mobile unit 2 traveling on a route including an abnormal route (step S1). The collection unit 10 collects data corresponding to a normal route or data that can be considered to correspond to a normal route as normal data using the method described above. Furthermore, the collection unit 10 collects a small amount of abnormal data relative to the normal data. The collection unit 10 can collect, via the network NW, communication areas associated with a location registration signal from the mobile unit 2 that is transmitted when the mobile unit 2 crosses a communication area.

[0110] Next, the first learning unit 11 performs a first learning process (step S2). FIG. 10 is a flowchart illustrating the first learning process of step S2 in more detail. As shown in step S30 of FIG. 10, the first learning unit 11 calculates a density ratio r θ (Step S20). The first learning unit 11 updates the parameter θ by gradient descent or the like using the above equations (7) to (9) to find an optimal solution for the parameter θ.

[0111] Next, the first learning unit 11 calculates the density ratio r from the learned parameter θ calculated in step S20 based on the above formula (2). θ is estimated (step S21).

[0112] 9, the conversion unit 12 performs a conversion process using the sequence learning model, and the second learning unit 13 performs a second learning process (step S3). In step S3, the second learning unit 13 regards normal data converted by the conversion unit 12 based on the calculation of the sequence learning model and reflecting the dependency relationship between sequences in the communication area as true normal data, and calculates the density ratio r estimated by the first learning unit 11 while keeping the generator parameters of the generator 131 fixed. The generator 131 generates pseudo-normal data that is sufficiently deviated from the distribution of true normal data. θOnly the classifier parameters of the classifier 132 that distinguishes between true normal data and pseudo normal data are updated in the direction of maximizing the objective function E based on the probability density function ρ(x|y=1) of normal data and the probability density function ρ(x|y=0) of abnormal data (the above equation (17)) determined from the above equation.

[0113] 11A and 11B are flowcharts for explaining the conversion process and the second learning process in step S3. First, the second learning unit 13 converts the density ratio r θ The probability density function ρ(x|y=1) of normal data and the probability density function ρ(x|y=0) of abnormal data (the above formula (17)) determined from the above are set as the objective function E (formula (15)) of the GAN (step S300). More specifically, the second learning unit 13 calculates the probability density function ρ(x|y=1), which is the conditional probability distribution of normal data, from the large amount of normal data collected by the collection unit 10 using a maximum likelihood estimation method or the like. In addition, the second learning unit 13 determines the probability density function ρ(x|y=0), which is the conditional probability distribution of abnormal data expressed by the above formula (17), from the calculated probability density function ρ(x|y=1) of normal data.

[0114] The prior probability ρ(y=1) of normality and the prior probability ρ(y=0) of abnormality are set to values ​​previously set, for example, ρ(y=1):ρ(y=0)=0.99:0.01, in the above equation (12), because the number of normal data is overwhelmingly large. Furthermore, in the optimal solution of the objective function E in the above equation (15), D(w D ,θ D ) is the posterior probability ρ(y=1|x) that the observed communication area sequence x is normal, and 1-D(G(w G ,θ G ),w D ,θ D ) is the posterior probability ρ(y=0|x) of anomaly for the communication area sequence x. These posterior probabilities ρ(y=1|x) and ρ(y=0|x) are calculated based on the estimated density ratio r θ and can be obtained from the prior probabilities ρ(y=1), ρ(y=0).

[0115] Next, the second learning unit 13 acquires the sequences of normal communication areas collected in step S1 as normal data (step S301) (training data 134 before conversion in FIG. 4). Next, the conversion unit 12 sets initial values ​​of model parameters of the sequence learning model (step S302). In step S302, pre-set model parameters are set, and for example, values ​​based on empirical rules or values ​​set randomly can be used. Next, the conversion unit 12 inputs the normal data acquired in step S301 into the sequence learning model, converts it into normal data that reflects the dependency relationships between the sequences of communication areas based on the model parameters, and outputs the converted normal data as true normal data (step S303).

[0116] Next, the second learning unit 13 inputs the true normal data output from the sequence learning model in step S303 to the classifier 132 as training data, and adjusts the parameter w of the classifier 132 so that the true normal data is classified as true normal data (y=1). D ,θ D (Step S304). In Step S304, the second learning unit 13 can cause the classifier 132 to learn true normal data using, for example, an error backpropagation method. In Step S304, the classifier 132 that can distinguish true normal data from true normal data is pre-trained.

[0117] Next, the second learning unit 13 generates Gaussian noise and provides a random vector of the generated Gaussian noise as an input to the generator 131 (step S305). Subsequently, the generator 131 calculates a random vector of the input z and the weight parameter w based on the provided Gaussian noise. G ,θ G and threshold processing using an activation function to generate pseudo-normal data G(z) before conversion (step S306).

[0118] Next, the conversion unit 12 provides the pseudo normal data G(z) before conversion generated in step S306 as an input to the sequence learning model, and converts it into pseudo normal data G(z) that reflects the dependency relationship between the sequences of the communication areas based on the calculations of the sequence learning model (step S307). Subsequently, as shown by the connector A in FIG. 11B, the second learning unit 13 trains the classifier 132. The training of the classifier 132 is performed by using the parameter w of the generator 131. G ,θ G First, the second learning unit 13 provides the true normal data obtained in step S303 as training data to the classifier 132 as input. Then, the second learning unit 13 adjusts the parameter w by backpropagation or the like so that the objective function E in the above equation (15) is maximized. D ,θ D (Step S308). The label of the training data is set to 1 (true normal data).

[0119] Next, the second learning unit 13 provides the pseudo-normal data generated by the generator 131 in step S306 and further converted in step S307 as input to the discriminator 132, and calculates the parameter w by backpropagation or the like so that the objective function E in the above equation (15) is maximized. D ,θ D is updated (step S309).

[0120] The learning of the classifier 132 in steps S308 and S309 corresponds to the dashed arrows in the block diagram of the second learning unit 13 shown in FIG. 4 , which indicate that a classifier error is calculated in block 135 of the objective function E based on the output 133 from the classifier 132, and then the error is backpropagated to the classifier 132.

[0121] Thereafter, if the value of the objective function E has not converged (step S310: NO), the adjustment unit 14 adjusts the model parameters of the sequence learning model (step S311), and the second learning unit 13 repeatedly learns the classifier 132. The adjustment unit 14 sets the model parameters adjusted in step S311 to the sequence learning model (step S312). Next, the conversion unit 12 again inputs the normal data acquired in step S301 to the sequence learning model for which the adjusted model parameters have been set, performs calculations on the sequence learning model, converts the data into normal data that reflects the dependency relationships between the sequences of the communication areas, and outputs the data as true normal data (step S313).

[0122] Next, the conversion unit 12 inputs the pre-conversion pseudo-normal data generated by the generator 131 based on noise into the sequence learning model in which the adjusted model parameters are set, performs calculations on the sequence learning model, converts the data into pseudo-normal data that reflects the inter-sequence dependency of the communication areas, and outputs the pseudo-normal data (step S314). Thereafter, the second learning unit 13 repeats steps S308 and S309, and when the value of the objective function E converges to the optimal solution of the above equation (15) (step S310: YES), repeats the processes from step S302 to step S314 using other normal data in turn, as shown by the connector B in the figure, until the generator 131 and the discriminator 132 are trained (step S315: NO).

[0123] Thereafter, when the classifier 132 has been trained using all normal data (step S315: YES), the storage unit 19 stores the trained generative models, the generator 131' and the classifier 132', and the model parameters of the sequence learning model adjusted by the adjustment unit 14 (step S316). Note that steps S302 to S314 can be batch processed. After that, the process proceeds to step S4 in FIG. 9.

[0124] Next, the generation unit 15 causes the generator 131' included in the trained generative model constructed by the second learning unit 13 to generate pre-conversion pseudo-normal data (step S4). Subsequently, the conversion unit 12 uses the sequence learning model in which the adjusted model parameters are set to convert the pre-conversion pseudo-normal data generated in step S4 into pseudo-normal data that reflects the dependency relationships between communication area sequences (step S5). Next, the communication area anomaly sequence database 16 stores the pseudo-normal data (step S6). In step S6, the pseudo-normal data converted in step S5 is further passed through the classifier 132' included in the trained generative model, and only pseudo-normal data for which the output of the classifier 132' is determined to be 0 (abnormal) >> 1 (normal) can be stored in the communication area anomaly sequence database 16. Note that the determination process by the classifier 132' may be performed after the pre-conversion pseudo-normal data is generated by the generator 131' in step S305.

[0125] The collection unit 10 collects sequences (management target data) of communication areas through which the mobile object 2 passes during the movement of the management target (step S7). Next, the conversion unit 12 performs calculations on the sequence learning model in which the adjusted model parameters are set, and converts the management target data collected in step S6 into management target data that reflects the dependency relationships between the communication area sequences (step S8).

[0126] Next, the determination unit 17 determines that the movement route of the managed object is abnormal if the managed object data converted in step S8 matches at least one of the pseudo-normal data stored in the communication area anomaly series database 16 (step S9). In step S9, the determination unit 17 can determine that the movement route of the managed object is abnormal when the pseudo-normal data matches the managed object data converted in step S8, sequentially from the first pseudo-normal data among the 1 to L pseudo-normal data stored in the communication area anomaly series database 16.

[0127] Next, the notification unit 18 sends a notification that the route of movement of the managed object is abnormal, for example, to the communication terminal 20 of the moving object 2 that has made the movement of the managed object (step S10).

[0128] As described above, the anomaly management device 1 according to this embodiment learns parameters representing the density ratio between the probability density function of the normal data and the probability density function of the anomalous data based on normal data, which is a sequence of communication areas passed by a mobile unit 2 traveling along a normal route and can be collected in large quantities, and anomalous data, which is a sequence of communication areas passed by a mobile unit 2 traveling along a route including an anomalous route and can be collected in small quantities. Furthermore, the probability density function of the normal data and the probability density function of the anomalous data determined based on the estimated density ratio are set as the objective function of the generative model, and the classifier 132 is updated in the direction of maximizing the objective function. Furthermore, in learning the generative model, training data that reflects the dependency between the sequence of communication areas converted by the sequence learning model is used as input data for the classifier 132. Then, a database is constructed using pseudo-normal data, which is generated in large quantities by the generator 131′ of the trained generative model and is sufficiently different from the distribution of true normal data indicating that the mobile unit 2 traveled along a normal route, and is further converted using the sequence learning model, as templates for abnormal data. Therefore, even if the amount of collected abnormality data is small, it is possible to manage abnormalities in the movement routes of the management target.

[0129] Furthermore, according to the abnormality management device 1 of this embodiment, the density ratio is estimated using abnormal data that is available even in small amounts, and the probability density function of normal data and the probability density function of abnormal data determined based on the density ratio are set as coefficients of the objective function of the generative model. Therefore, the generative model is trained by reflecting the patterns of normal data and abnormal data obtained as actual observation data, thereby improving the learning accuracy of the generative model.

[0130] Furthermore, according to the anomaly management device 1 of this embodiment, a sequence learning model that extracts features that represent the dependency relationships between communication area sequences is interposed between the input side of the classifier 132 of the generative model and the output side of the generator 131, and the values ​​of the model parameters of the sequence learning model are adjusted according to the progress of learning of the generative model. This makes it possible to further improve the learning accuracy of the generative model.

[0131] In the embodiment described above, the second learning unit 13 has been described as having a generative model with a GAN configuration. However, the generative model can be configured not only based on a GAN but also based on a VAE (Variational Autoencoder), Energy-Based Models (EBMs), or the like.

[0132] In addition, in the described embodiment, the conversion unit 12 employs a neural network such as an RNN, LSTM, or self-attention as the sequence learning model. However, the sequence learning model may be an autoencoder, a transformer, or the like, as long as it learns the temporal or sequential dependencies inherent in sequence data and extracts features, converts, or reconstructs the input sequence based on the dependencies.

[0133] The above describes embodiments of the abnormality management device and abnormality management method of the present invention, but the present invention is not limited to the described embodiments, and various modifications that a person skilled in the art can conceive are possible within the scope of the invention described in the claims. [Explanation of symbols]

[0134] 1...abnormality management device, 10...collection unit, 11...first learning unit, 12...conversion unit, 13...second learning unit, 14...adjustment unit, 15...generation unit, 16...communication area abnormality sequence database, 17...determination unit, 18...notification unit, 19...memory unit, 2...mobile body, 20...communication terminal, 21...SIM, 22...ECU, 3...core network, 30...AMF, 31...UDM, 32...UDR, 33...UPF, 101...bus, 102...processor, 103...main memory device, 32a, 33a, 104...communication interface, 105...auxiliary memory device, 106...input / output I / O, 107...display device, 131...generator, 132...identifier, NW...network.

Claims

1. a first learning unit configured to learn a parameter representing a ratio between a probability distribution of normal data indicating a sequence of communication areas through which a mobile object moving along a normal route passes and a probability distribution of abnormal data indicating a sequence of communication areas through which a mobile object moving along a route including an abnormal route deviating from the normal route passes, in a movement space of the mobile object defined by a plurality of communication areas, and to estimate the ratio based on the learned parameter; a first conversion unit configured to convert the normal data into normal data that reflects the dependency relationships using a sequence learning model having pre-set model parameters that represent dependency relationships between sequences of communication areas passed by a mobile object traveling on a normal route; a second learning unit configured to: consider the normal data, in which the dependency relationship converted by the first conversion unit is reflected, as true normal data, and generate pre-conversion pseudo-normal data before the dependency relationship is reflected, corresponding to pseudo-normal data sufficiently deviating from the distribution of the true normal data, while keeping fixed generator parameters of a generator that distinguishes between the true normal data and the pseudo-normal data, in a direction that maximizes an objective function based on a probability distribution of the normal data and a probability distribution of the abnormal data determined from the ratio estimated by the first learning unit; and to learn a generative model including the generator and the classifier; a generation unit configured to generate the pre-conversion pseudo-normal data using the generator included in the generative model trained by the second learning unit; a second conversion unit configured to convert the pre-conversion pseudo-normal data generated by the generation unit into the pseudo-normal data reflecting the dependency relationship using the sequence learning model; a storage unit configured to store data that has been further identified as the pseudo-normal data by the classifier included in the generative model trained by the second learning unit, from the pseudo-normal data that has been converted by the second conversion unit and in which the dependency relationship has been reflected; An abnormality management device comprising:

2. 2. The abnormality management device according to claim 1, further comprising an adjustment unit configured to adjust the model parameters of the sequence learning model according to a state of learning of the generative model by the second learning unit; the first conversion unit is configured to convert the normal data into normal data in which the dependency is reflected, using the sequence learning model having the model parameters adjusted by the adjustment unit; The second conversion unit is configured to convert the pre-conversion pseudo-normal data generated by the generation unit into the pseudo-normal data reflecting the dependency relationship, using the sequence learning model having the model parameters adjusted by the adjustment unit. An abnormality management device characterized by:

3. 2. The abnormality management device according to claim 1, a collection unit configured to collect management object data, which is a sequence of communication areas through which a mobile object passes during movement of the management object; a third conversion unit configured to convert the target data into target data in which the dependency relationship is reflected, using the sequence learning model; a determination unit configured to determine that a route of movement of the managed object is abnormal when the managed object data in which the dependency relationship is reflected matches one of the pseudo-normal data stored in the storage unit and in which the dependency relationship is reflected; An abnormality management device comprising:

4. 4. The abnormality management device according to claim 3, Further, a notification unit configured to notify a moving body that has moved the managed object when the determination unit determines that the route of the movement of the managed object is abnormal. An abnormality management device characterized by:

5. 2. The abnormality management device according to claim 1, The sequence learning model is a recurrent neural network model. An abnormality management device characterized by:

6. a first learning step of learning a parameter representing a ratio between a probability distribution of normal data indicating a sequence of communication areas through which a mobile object moving along a normal route passes and a probability distribution of abnormal data indicating a sequence of communication areas through which a mobile object moving along a route including an abnormal route that deviates from the normal route passes, in a movement space of the mobile object defined by a plurality of communication areas, and estimating the ratio based on the learned parameter; a first conversion step of converting the normal data into normal data that reflects the dependency relationships using a sequence learning model having pre-set model parameters that represent dependency relationships between sequences of communication areas passed by a mobile object traveling on a normal route; a second learning step of learning a generative model including the generator and the classifier by updating a classifier parameter of a classifier that distinguishes between the true normal data and the pseudo-normal data in a direction that maximizes an objective function based on a probability distribution of the normal data and a probability distribution of the abnormal data determined from the ratio estimated in the first learning step, while keeping fixed a generator parameter of a generator that generates pre-conversion pseudo-normal data before the dependency relationship is reflected, where the pre-conversion pseudo-normal data corresponds to pseudo-normal data that is sufficiently deviated from the distribution of the true normal data, with the normal data that has been converted in the first conversion step and in which the dependency relationship is reflected being considered as true normal data; a generating step of generating the pre-conversion pseudo-normal data by the generator included in the generative model trained in the second learning step; a second conversion step of converting the pre-conversion pseudo-normal data generated in the generation step into the pseudo-normal data reflecting the dependency relationship using the sequence learning model; a storage step of storing, in a storage unit, data that has been further identified as the pseudo-normal data by the classifier included in the generative model trained in the second learning step, from the pseudo-normal data that has been converted in the second conversion step and in which the dependency relationship has been reflected; An abnormality management method comprising:

7. 7. The abnormality management method according to claim 6, further comprising an adjustment step of adjusting the model parameters of the sequence learning model according to a learning state of the generative model in the second learning step; the first conversion step converts the normal data into normal data that reflects the dependency relationship using the sequence learning model having the model parameters adjusted in the adjustment step; The second conversion step converts the pre-conversion pseudo-normal data generated in the generation step into the pseudo-normal data reflecting the dependency relationship, using the sequence learning model having the model parameters adjusted in the adjustment step. An abnormality management method characterized by:

8. 7. The abnormality management method according to claim 6, Furthermore, a collection step of collecting management object data which is a series of communication areas through which a mobile object passes during movement of the management object; a third conversion step of converting the object data into object data in which the dependency relationship is reflected, using the sequence learning model; a determining step of determining that the movement route of the managed object is abnormal when the managed object data in which the dependency relationship is reflected matches one of the pseudo-normal data stored in the storage unit and in which the dependency relationship is reflected; An abnormality management method comprising:

9. 9. The abnormality management method according to claim 8, Further, the method includes a notification step of notifying a moving body that has moved the managed object when the route of the managed object is determined to be abnormal in the determination step. An abnormality management method characterized by:

10. 7. The abnormality management method according to claim 6, The sequence learning model is a recurrent neural network model. An abnormality management method characterized by:

Images