Learning program, learning method, and information processing apparatus

Abstract

To accurately estimate received power of a communication device.SOLUTION: An information processing device:: calculates, by using a first model that estimates a value of received power of a radio wave received by a communication device from a base station based on position information of the base station and position information of the communication device, an estimated value of the received power of the radio wave; and trains a second model that outputs a correction value for correcting the estimated value of the first model based on an actually measured value of the received power of the communication device in an operation environment using the first model, and the estimated value.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a learning program and the like.

Background Art

[0002] In 5G, compared with 4G, high-speed and large-capacity communication can be achieved.

[0003] However, the power consumption of 5G is expected to reach 4 to 9 times that of 4G in the future. From the perspective of reducing greenhouse gas emissions, etc., it is desirable to achieve power saving while ensuring communication quality.

[0004] In order to achieve both communication quality and power saving, it is required to estimate the received power of the UE based on the information of the BS (Base Station) and UE scattered in various places, and appropriately control the BS based on the estimation result.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Summary of the Invention

Problems to be Solved by the Invention

[0006] Also, for example, as a technology related to the above, there is a conventional technology for estimating the received power of the UE from the information of the BS and UE based on an estimation model learned by machine learning.

[0007] However, in the above prior art, there is a problem that it is not an estimation model that can accurately estimate the received power of a communication device such as a UE, and the received power of a communication device such as a UE cannot be accurately estimated in actual operation.

[0008] In one aspect, an object of the present invention is to provide a learning program, a learning method, and an information processing apparatus for accurately estimating the received power of a communication device.

Means for Solving the Problems

[0009] In the first aspect, a computer executes the following processes. The computer determines an estimated value of the received power of radio waves received by the communication device from the base station by using a first model that estimates the value of the received power of radio waves received by the communication device from the base station based on the position information of the base station and the position information of the communication device. The computer trains a second model that outputs a correction value for correcting the estimated value of the first model based on the measured value and the estimated value of the received power of the communication device in the operating environment using the first model.

Effects of the Invention

[0010] The received power of the communication device can be accurately estimated.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Embodiment for Carrying Out the Invention

[0012] Hereinafter, embodiments of the learning program, learning method, and information processing apparatus disclosed in the present application will be described in detail with reference to the drawings. Note that the present invention is not limited by this embodiment.

Embodiment

[0013] An example of the processing of the information processing apparatus according to this embodiment will be described. The information processing apparatus according to this embodiment is denoted as "information processing apparatus 100". For example, the information processing apparatus 100 executes processing for collecting measured data, processing for learning a correction model, processing for creating a propagation model, and processing related to the design and learning of a control law.

[0014] First, an example of the process of collecting measured data will be described. FIG. 1 is a diagram for explaining the process of collecting measured data. The information processing apparatus 100 collects information on a base station (BS) and a communication apparatus (UE) as "measured data" while operating a network in an actual operation environment according to a predetermined control rule. The predetermined control rule is a control rule set in advance.

[0015] For example, let the area corresponding to the actual operation environment be Area 5. Area 5 includes a plurality of base stations and a plurality of communication apparatuses. The information processing apparatus 100 performs control to switch the RU (Radio Unit) of each base station to the "active mode" or the "sleep mode" according to a predetermined control rule. The above-mentioned measured data includes the position information of each base station, the position information of each communication apparatus, the received power (information on the received power) of each communication apparatus, and the like.

[0016] The information processing apparatus 100 may communicate with each base station and each communication apparatus included in Area 5 to collect measured data, or may collect measured data via an external server or the like that communicates with each base station and each communication apparatus. The information processing apparatus 100 collects measured data at predetermined time intervals and stores the collected measured data in the measured DB 142.

[0017] Note that the information processing apparatus 100 may further collect, as measured data, the height at which the base station is installed, the power consumption (transmission power) required when the base station transmits radio waves, the frequency of the radio waves, the height of the communication apparatus, and the like.

[0018] Subsequently, an example of the process of learning (machine learning) a correction model will be described. FIG. 2 is a diagram for explaining the process of learning a correction model. When the information processing apparatus 100 learns the correction model M2, it uses the input data 142a, the output data 142b, and the physical model M1.

[0019] The input data 142a has the position information of the base station and the position information of the communication apparatus among the information included in the measured data.

[0020] The output data 142b has the received power of the communication device among the information included in the measured data. The information processing device 100 calculates the "measured value P" of the received power by performing data preprocessing on the output data 142b. For example, the information processing device 100 performs data preprocessing such as calculating the average value of the received power of the target communication device.

[0021] The physical model M1 is a model that estimates the received power of a communication device that receives radio waves from a base station based on a distance attenuation formula of radio waves, etc. in a preset first communication simulation environment. The physical model M1 is generated in advance. For example, the information processing device 100 calculates the "estimated value P'" of the received power by inputting the input data 142a into the physical model M1. The physical model M1 corresponds to the "first model".

[0022] The correction model M2 is a model that takes the input data 142a as input and outputs a correction value Δ. The correction model M2 is a DNN (Deep Neural Network) or the like. The information processing device 100 updates the parameters of the correction model M2 so that the value of "P - (P' + Δ)" approaches 0. For example, the information processing device 100 updates the parameters of the correction model M2 (performs machine learning on the correction model M2) using the error backpropagation method or the like. The correction model M2 corresponds to the "second model".

[0023] Note that when the information processing device 100 executes the training of the correction model M2, it may further use, as the input data 142a, the height at which the base station is installed, the power consumption (transmission power) required when the base station transmits radio waves, the frequency of the radio waves, the height of the communication device, etc.

[0024] Subsequently, an example of the process of creating a propagation model will be described. FIG. 3 is a diagram for explaining the process of creating a propagation model. As shown in FIG. 3, the information processing device 100 creates a propagation model M3 by combining the physical model M1 and the correction model M2. FIG. 3 shows an example of calculating an estimated result 7 of the received power based on the input data 6 and the propagation model M3.

[0025] The input data 6 includes the location information of any base station in the area corresponding to the operation environment and the location information of any communication device. For example, in the input data 6 shown in FIG. 3, the triangular mark indicates the location of any base station, and the circular mark indicates the location of each communication device.

[0026] When the information processing device 100 inputs the input data 6 into the propagation model M3, the input data 6 is input into the physical model M1 and the correction model M2 respectively. When the physical model M1 receives the input data 6, it outputs an estimated value of the received power of any communication device. When the correction model M2 receives the input data 6, it outputs a correction value of the received power of any communication device. The propagation model M3 estimates, as the received power of any communication device, the result of adding the estimated value output from the physical model M1 and the correction value output from the correction model M2.

[0027] For example, as shown in FIG. 3, when there are multiple communication devices for which the received power is to be estimated, the information processing device 100 sets a set of the location information of one base station and the location information of one communication device as input data and inputs it into the propagation model M3. The information processing device 100 repeatedly executes the above process while changing the target communication device, calculates the received power of each communication device, and calculates the estimation result 7. Note that the information processing device 100 may also set the location information of the base station and the location information of a plurality of communication devices together in the input data and calculate the estimation result 7.

[0028] Subsequently, an example of the process related to the design and learning of the control rule will be described. FIG. 4 is a diagram for explaining the process related to the design and learning of the control rule. In the example shown in FIG. 4, the information processing device 100 uses the propagation model M3 to construct a new communication simulation environment and derives a new control rule for network operation. Hereinafter, the new communication simulation environment is referred to as the "second communication simulation environment".

[0029] In the first communication simulation environment shown in FIG. 2, the received power of the communication device was estimated using only the physical model M1. However, in the second communication simulation environment, the received power of the communication device is estimated using the propagation model M3. In the propagation model M3, since the estimated value of the physical model M1 is corrected by the correction value of the correction model M2, the estimation accuracy of the received power of the communication device is improved.

[0030] In the second communication simulation environment, the information processing device 100 executes reinforcement learning to learn a new control rule for performing outage control of the base station. For example, based on the state s of the second simulation environment and the control rule, the information processing device 100 generates an action α, performs outage control of the base station, and acquires the state s and the reward r after the outage control. The control rule is a control rule that determines the action α based on the state s.

[0031] The state s includes the time, the traffic volume of each grid, the load of the base station, etc. The grid is a grid obtained by dividing the target area into predetermined regions. In the action α, the RU of each base station is set to the active mode or the sleep mode. The reward r includes the total reduction amount of the power consumption of each base station.

[0032] The information processing device 100 estimates the received power of each communication device based on the propagation model M3, and estimates the power consumption of the corresponding base station based on the estimation result. In the following description, the power consumed when the RU of the base station is in the active mode or the sleep mode is simply referred to as "power consumption".

[0033] For example, the information processing apparatus 100 estimates the power consumption by using the relationship that the base station to which each communication apparatus is connected is determined according to the received power magnitude of the radio waves transmitted from each base station, and the greater the sum of the traffic demand amounts of the connected communication apparatuses, the greater the power consumption of the corresponding connected base station. The information processing apparatus 100 calculates the total sum of the power consumption of each base station by summing up the power consumption of each base station. The information processing apparatus 100 calculates the reward r by subtracting the total sum of the power consumption of each base station after performing the outage control by the action α from the total sum of the power consumption of each base station before performing the outage control by the action α.

[0034] The information processing apparatus 100 repeatedly executes the process of generating the action α from the state s while changing the control rule so that the reward r is maximized (performs reinforcement learning). The information processing apparatus 100 applies the new control rule obtained after reinforcement learning to the actual operation environment to perform outage control of the base station. For example, the actual operation environment is the area 5 shown in FIG. 1.

[0035] The information processing apparatus 100 repeatedly executes the processes of collecting the above-described measured data, learning the correction model, creating the propagation model, and designing and learning the control rule.

[0036] In the description of FIG. 4, the information processing apparatus 100 derived the control rule using reinforcement learning, but is not limited thereto. For example, the information processing apparatus 100 may derive a new control rule using mathematical optimization, model predictive control, etc. Also, when deriving the control rule, the information processing apparatus 100 may perform transfer learning using the parameters of the previously derived control rule.

[0037] Also, the information processing apparatus 100 used the total reduction amount of the power consumption of each base station as the reward when performing reinforcement learning, but is not limited thereto. For example, as the reward, the received power of each communication apparatus, the SINR (Signal-to-Interference-plus-Noise Ratio) of each communication apparatus may be used as the reward.

[0038] As described above, when estimating the received power of the communication device, the information processing apparatus 100 according to the present embodiment learns a correction model M2 that generates a correction value Δ for correcting the estimated value P' of the physical model M1, and uses a propagation model M3 including the physical model M1 and the correction model M2. Thereby, compared with the case of estimating the received power of the communication device using only the physical model M1 or the case of estimating the received power of the communication device using only the estimation model learned by machine learning, the received power of the communication device can be estimated with high accuracy.

[0039] Further, in the second communication simulation environment, the information processing apparatus 100 repeatedly executes a process of generating an action α from a state s while changing a control law so that a reward r is maximized (executes reinforcement learning). As a result, it is possible to avoid the occurrence of constraint violations and performance degradation in network operation due to unfavorable (non-optimal) action selection during learning, and to derive an appropriate control law. In addition, when the control law learned in the second communication simulation environment is introduced into the actual operation environment, the same performance as during learning can be achieved. Further, by repeating the above process, even if it is not achieved, it is possible to reduce the man-hours related to adjustment of necessary hyperparameters and re-learning.

[0040] Next, a configuration example of the information processing apparatus 100 that executes the above-described process will be described. FIG. 5 is a functional block diagram showing the configuration of the information processing apparatus according to the present embodiment. As shown in FIG. 5, the information processing apparatus 100 includes a communication unit 110, an input unit 120, a display unit 130, a storage unit 140, and a control unit 150.

[0041] The communication unit 110 executes data communication with a base station, a communication device, an external device, etc. via a network. The communication unit 110 is a NIC (Network Interface Card) or the like.

[0042] The input unit 120 is an input device that inputs various information to the control unit 150 of the information processing apparatus 100. For example, the input unit 120 corresponds to a keyboard, a mouse, a touch panel, or the like.

[0043] The display unit 130 is a display device that displays information output from the control unit 150.

[0044] The storage unit 140 has control rule data 141, a measured database 142, a physical model M1, a correction model M2, and a propagation model M3. The storage unit 140 is a memory or the like.

[0045] The control rule data 141 is the information of the control rule described above. The information processing apparatus 100 performs standing wave control of the base station based on the control rule data 141.

[0046] The measured database 142 stores the measured data collected by the information processing apparatus 100. Other explanations regarding the measured data and the measured database 142 are the same as the above explanations.

[0047] The physical model M1 is a model that estimates the received power of a communication device that receives radio waves from a base station based on a distance attenuation formula of radio waves or the like in a preset first communication simulation environment. Other explanations regarding the physical model M1 are the same as the above explanations.

[0048] The correction model M2 is a model (NN) that takes the input data 142a as input and outputs a correction value Δ. Other explanations regarding the correction model M2 are the same as the above explanations.

[0049] The propagation model M3 is a model that combines the above physical model M1 and the trained correction model M2. Other explanations regarding the propagation model M3 are the same as the above explanations.

[0050] The control unit 150 includes a standing wave control unit 151, a collection unit 152, a learning unit 153, a creation unit 154, and a reinforcement learning unit 155. The control unit 150 is a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or the like.

[0051] The wave stop control unit 151 controls the RUs of each base station included in the area of the actual operation environment based on the control rule data 141. In the following description, the area of the actual operation environment will be described as "Area 5".

[0052] For example, the wave stop control unit 151 uses the communication unit 110 to communicate with the base stations included in Area 5 and identify the state of Area 5. The state includes time, traffic volume of each grid, base station load, etc. The wave stop control unit 151 generates a control signal for setting the RUs of the base stations included in Area 5 to the active mode or the sleep mode based on the state of Area 5 and the control rule data 141 (the latest control rule), and transmits it to the base stations. The wave stop control unit 151 repeatedly executes such processing at predetermined time intervals.

[0053] The collection unit 152 uses the communication unit 110 to communicate with the base stations and communication devices included in Area 5 and collect the measured data. The collection unit 152 registers the collected measured data in the measured DB 142. The collection unit 152 repeatedly executes such processing at predetermined time intervals during a certain period.

[0054] The description of other processes for the collection unit 152 to collect the measured data corresponds to the description of the processes performed in FIG. 1.

[0055] The learning unit 153 executes the learning (machine learning) of the correction model M2 based on the measured DB 142 and the physical model M1. The description of the process for the learning unit 153 to learn the correction model M2 corresponds to the description of the processes performed in FIG. 2.

[0056] The creation unit 154 generates the propagation model M3 by combining the physical learning model M1 and the trained correction model M2. The description of the propagation model M3 is the same as the description of the propagation model M3 described in FIG. 3.

[0057] The reinforcement learning unit 155 constructs a new communication simulation environment (the second communication simulation environment) using the propagation model M3, executes reinforcement learning of control rules, and thereby derives new control rules for network operation. The reinforcement learning unit 155 updates the control rule data 141 according to the derived control rules.

[0058] The explanation of the process by which the reinforcement learning unit 155 derives control rules is the same as the explanation of the process related to the design and learning of the control rules described in FIG. 4.

[0059] Next, an example of the processing procedure of the information processing apparatus 100 according to the present embodiment will be described. FIG. 6 is a flowchart showing the processing procedure of the information processing apparatus according to the present embodiment. As shown in FIG. 6, the outage control unit 151 of the information processing apparatus 100 executes outage control of the base stations in the area corresponding to the actual operation environment based on the control rule data 141 (step S101).

[0060] The collection unit 152 of the information processing apparatus 100 operates the network for a certain period, communicates with the base stations and communication devices in the area corresponding to the actual operation environment, collects measured data, and registers it in the measured DB 142 (step S102).

[0061] The learning unit 153 of the information processing apparatus 100 executes learning of the correction model M2 based on the measured data registered in the measured DB 142 and the physical model M1. The creation unit 154 of the information processing apparatus 100 creates the propagation model M3 based on the physical model M1 and the correction model M2 (step S104).

[0062] The information processing apparatus 100 constructs the second communication simulation environment by incorporating the propagation model M3 into the first communication simulation environment (step S105).

[0063] The reinforcement learning unit 155 of the information processing apparatus 100 performs reinforcement learning in the second communication simulation environment and derives a new control rule for performing outage control of the base station (step S106). The reinforcement learning unit 155 updates the control rule data 141 according to the new control rule (step S107).

[0064] The outage control unit 151 executes outage control of the base stations in the area corresponding to the actual operation environment based on the updated control rule data 141 (step S108).

[0065] Next, a specific example of the target area and an example of the estimated result of the received power at each point by the propagation model M3 are shown.

[0066] FIG. 7 is a diagram showing the target area. In the example shown in FIG. 7, an example is shown in which the target area 15 includes one first base station (Macro cell Base Station: MBS), a plurality (for example, 11) of second base stations (Small cell Base Station: SBS), and a plurality (for example, 64) of communication devices (UE). The first base station is indicated by a square (□). The second base station is indicated by a triangle (△). Note that, for example, a second base station with overlapping positions may be indicated by one triangle. The communication device is indicated by a circle (●). Six obstacles 16 are set in the target area 15. The influence of the obstacles 16 on the radio wave environment is unknown.

[0067] Note that the learning conditions of the correction model M2 using the DNN by the learning unit 153 are set to the following conditions (1) to (3).

[0068] Condition (1): The input data of the training data (measured data) is the position information of the base station and the position information of the communication device. The output data is the received power of the communication device. Condition (2): The model structure and the number of parameters are as follows. The number of layers is 4. The layers include an input layer, two intermediate layers, and an output layer. The number of neurons in the intermediate layer is 64. The activation function of the intermediate layer is the ReLU function. Condition (3): Regarding the learning algorithm, the optimization method is "Adam", and the loss function is "mean absolute error".

[0069] Figures 8 to 11 are diagrams showing the comparison results between the estimated values and the theoretical values by the propagation model. The theoretical value is the correct value.

[0070] In Figure 8, the theoretical value 21a and the estimated value 21b of the received power of each communication device (location) when the first base station 21 transmits radio waves are shown. In Figure 9, the theoretical value 22a and the estimated value 22b of the received power of each communication device (location) when the second base station 22 transmits radio waves are shown. In Figure 10, the theoretical value 23a and the estimated value 23b of the received power of each communication device (location) when the second base station 23 transmits radio waves are shown. In Figure 11, the theoretical value 24a and the estimated value 24b of the received power of each communication device (location) when the second base station 24 transmits radio waves are shown. Diagrams of the comparison results between the estimated values and the theoretical values of the received power of each communication device when other base stations transmit radio waves are omitted.

[0071] Referring to the comparison results of the estimated values and the theoretical values shown in Figures 8 to 11, it can be seen that the estimated values estimated by the propagation model M3 are close to the theoretical values. That is, by using the propagation model M3, the received power of the communication device can be accurately estimated.

[0072] Next, the verification results of the outage control to the base station will be described. When performing the verification, based on the control rules derived by the information processing device 100, the outage control to the base station was executed for 10 days for the target area 15 shown in Figure 7. The acquisition of traffic data and the switching of the active / sleep of the RUs of each base station were executed every 30 minutes. However, when a communication device that cannot be connected to any base station occurs within the target area 15 when the outage control is executed, all the RUs of all the base stations will be set to active by the fail-safe function and reconnection will be performed.

[0073] FIG. 12 is a diagram showing a comparison result of the average power consumption of the base station. The vertical axis of graph G1 is an axis corresponding to the power consumption of the base station. Bar graph 31 shows the control result (average power consumption) by the control rule learned in the first communication simulation environment. Bar graph 32 shows the control result (average power consumption) by the control rule learned in the second communication simulation environment. For reference, bar graph 33 shows the control result (average power consumption) of the control rule learned in the communication simulation in which the true radio environment is known and simulated.

[0074] In the example shown in FIG. 12, in the control by the control rule learned in the first communication simulation environment, the fail-safe occurred 24 times. On the other hand, in the control by the control rule learned in the first communication simulation environment, the fail-safe did not occur even once. Thereby, in the second communication simulation environment, the average power consumption of the base station can be reduced as compared with the first communication simulation environment. In addition, it is possible to avoid the occurrence of a communication device without a connection destination during the execution of the outage control caused by the difference in the radio environment between the simulation environment at the time of deriving the control rule and the actual operation environment, and to reduce the man-hours related to the re-derivation of the control rule.

[0075] FIG. 13 is a diagram showing a comparison result of the maximum load of the base station. The vertical axis of graph G2 is an axis corresponding to the load of the base station. Bar graph 41 shows the control result (maximum load) by the control rule learned in the first communication simulation environment. Bar graph 42 shows the control result (maximum load) by the control rule learned in the second communication simulation environment. For reference, bar graph 43 shows the control result (maximum load) of the control rule learned in the communication simulation in which the true radio environment is known and simulated.

[0076] As shown in FIG. 13, for the maximum load of the base station, almost the same value was obtained under any conditions.

[0077] Next, the effects of the information processing device 100 according to this embodiment will be described. When estimating the received power of a communication device, the information processing device 100 learns a correction model M2 that generates a correction value Δ for correcting the estimated value P' of the physical model M1, and uses a propagation model M3 that includes the physical model M1 and the correction model M2. This makes it possible to estimate the received power of the communication device with higher accuracy than when the received power of the communication device is estimated using only the physical model M1 or when the received power of the communication device is estimated using only an estimation model learned by machine learning.

[0078] In the second communication simulation environment, the information processing device 100 repeatedly executes a process of generating an action α from the state s while changing the control law so as to maximize the reward r (performing reinforcement learning). This makes it possible to derive an efficient control law, avoiding constraint violations and performance degradation in network operation due to undesirable (non-optimal) action selection during learning. Furthermore, when the control law learned in the second communication simulation environment is introduced into an actual operating environment, it is possible to achieve performance equivalent to that achieved during learning. Furthermore, by repeating the above process, it is possible to reduce the man-hours required for adjusting hyperparameters, re-learning, etc., even if the target is not achieved.

[0079] The information processing device 100 learns the correction model M2 based on the difference between the estimated value P' and the actual measurement value P. This makes it possible to generate the correction model M2 that can correct the output result of the physical model M1.

[0080] The information processing device 100 creates a propagation model M3 by combining the physical model M1 and the correction model M2. By using the propagation model M3, it is possible to accurately estimate the received power of the communication device.

[0081] When training the correction model M2, the information processing device 100 may further use, as input data 142a, the height at which the base station is installed, the power consumption (transmission power) required when the base station transmits radio waves, the frequency of the radio waves, the height of the communication device, etc. This makes it possible to estimate the reception power of the communication device with even greater accuracy.

[0082] Next, a description will be given of an example of the hardware configuration of a computer that realizes the same functions as the above-described information processing device 100. Fig. 14 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the information processing device of this embodiment.

[0083] 14, computer 200 has CPU 201 that executes various types of arithmetic processing, input device 202 that accepts data input from a user, and display 203. Computer 200 also has communication device 204 that exchanges data with external devices via a wired or wireless network, and interface device 205. Computer 200 also has RAM 206 that temporarily stores various types of information, and hard disk drive 207. Each of devices 201 to 207 is connected to bus 208.

[0084] The hard disk drive 207 stores a stoppage control program 207a, a collection program 207b, a learning program 207c, a creation program 207d, and a reinforcement learning program 207e. The CPU 201 reads out each of the programs 207a to 207e and loads them into the RAM 206.

[0085] The transmission outage control program 207a functions as the transmission outage control process 206a. The collection program 207b functions as the collection process 206b. The learning program 207c functions as the learning process 206c. The creation program 207d functions as the creation process 206d. The reinforcement learning program 207e functions as the reinforcement learning process 206e.

[0086] The processing of the shutdown control process 206a corresponds to the processing of the shutdown control unit 151. The processing of the collection process 206b corresponds to the processing of the collection unit 152. The processing of the learning process 206c corresponds to the processing of the learning unit 153. The processing of the creation process 206d corresponds to the processing of the creation unit 154. The processing of the reinforcement learning process 206e corresponds to the processing of the reinforcement learning unit 155.

[0087] It should be noted that each of the programs 207a to 207e does not necessarily have to be stored in the hard disk drive 207 from the beginning. For example, each of the programs may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 200. Then, the computer 200 may read and execute each of the programs 207a to 207e.

[0088] The following supplementary notes are further disclosed regarding the embodiments including the above examples.

[0089] (Supplementary Note 1) Calculating an estimated value of the received power of radio waves by utilizing a first model that estimates a value of the received power of radio waves received by the communication device from the base station based on location information of the base station and location information of the communication device; A second model is trained based on the estimated value and an actual measurement value of the received power of the communication device in an operating environment using the first model, and the second model outputs a correction value for correcting the estimated value of the first model. A learning program that causes a computer to execute a process.

[0090] (Appendix 2) The learning program according to appendix 1, wherein the training process trains the second model based on the difference between the actual measured value and the estimated value.

[0091] (Supplementary Note 3) The second model is a model that outputs the correction value when location information of the base station and the communication device is input, The learning program described in Appendix 2 is characterized in that it further causes a computer to execute a process of generating a third model that calculates an estimated value of the received power of the communication device based on an estimated value calculated based on the location information of the base station and the location information of the communication device and the first model, and the correction value.

[0092] (Appendix 4) A learning program as described in Appendix 3, characterized in that the computer is further made to execute a process of identifying an action based on the state of an area of a communication simulation environment and a control law, performing shutdown control of a base station included in the communication simulation environment based on the identified action, repeatedly executing a process of identifying a reward based on information about the base station and communication device included in the area after the shutdown control has been performed and an estimated value of the received power of the communication device calculated based on the third model, and performing learning to update the control law so as to maximize the reward.

[0093] (Supplementary Note 5) The learning program according to Supplementary Note 4, further comprising causing a computer to execute a process for performing a stoppage control for the base station based on the updated control law.

[0094] (Appendix 6) The learning program described in Appendix 1, characterized in that the process of calculating the estimated value further utilizes the height at which the base station is installed, the power consumption when the base station transmits radio waves, the frequency of the radio waves, and the height at which the communication device is installed to calculate an estimated value of the received power of the radio waves.

[0095] (Supplementary Note 7) Calculating an estimated value of the received power of radio waves by utilizing a first model that estimates a value of the received power of radio waves received by the communication device from the base station based on location information of the base station and location information of the communication device; A second model is trained based on the estimated value and an actual measurement value of the received power of the communication device in an operating environment using the first model, and the second model outputs a correction value for correcting the estimated value of the first model. A learning method characterized in that the processing is executed by a computer.

[0096] (Appendix 8) The learning method according to appendix 7, wherein the training process trains the second model based on the difference between the actual measured value and the estimated value.

[0097] (Supplementary Note 9) The second model is a model that outputs the correction value when location information of the base station and the communication device is input, The learning method described in Appendix 8, characterized in that the computer further executes a process of generating a third model that calculates an estimated value of the received power of the communication device based on an estimated value calculated based on the location information of the base station and the location information of the communication device and the first model, and the correction value.

[0098] (Appendix 10) The learning method described in Appendix 9, characterized in that the computer further executes a process of identifying an action based on the state of an area of a communication simulation environment and a control law, performing shutdown control of a base station included in the communication simulation environment based on the identified action, repeatedly executing a process of identifying a reward based on information about the base station and communication device included in the area after the shutdown control has been performed and an estimated value of the received power of the communication device calculated based on the third model, and performing learning to update the control law so as to maximize the reward.

[0099] (Supplementary Note 11) The learning method according to Supplementary Note 10, further comprising the step of: a computer executing a process for performing a stoppage control for the base station based on the updated control law.

[0100] (Appendix 12) The learning method described in Appendix 7, characterized in that the process of calculating the estimated value further utilizes the height at which the base station is installed, the power consumption when the base station transmits radio waves, the frequency of the radio waves, and the height at which the communication device is installed to calculate an estimated value of the received power of the radio waves.

[0101] (Supplementary Note 13) Calculating an estimated value of the received power of radio waves by utilizing a first model that estimates a value of the received power of radio waves received by the communication device from the base station based on location information of the base station and location information of the communication device; A second model is trained based on the estimated value and an actual measurement value of the received power of the communication device in an operating environment using the first model, and the second model outputs a correction value for correcting the estimated value of the first model. An information processing device having a control unit that executes processing.

[0102] (Appendix 14) The information processing apparatus according to Appendix 13, wherein the control unit trains the second model based on the difference between the measured value and the estimated value.

[0103] (Appendix 15) The second model is a model that outputs the correction value when the position information of the base station and the communication device is input. The control unit further executes a process of generating a third model for calculating an estimated value of the received power of the communication device based on the estimated value calculated based on the position information of the base station and the communication device and the first model, and the correction value. The information processing apparatus according to Appendix 14.

[0104] (Appendix 16) The control unit identifies an action based on the state of the area and the control rule in the communication simulation environment, and performs outage control of the base stations included in the communication simulation environment based on the identified action. After performing the outage control, the control unit repeatedly executes a process of identifying a reward based on the information of the base stations and communication devices included in the area and the estimated value of the received power of the communication device calculated based on the third model, and performs learning to update the control rule so that the reward is maximized. The information processing apparatus according to Appendix 15.

[0105] (Appendix 17) The information processing apparatus according to Appendix 16, wherein the control unit further executes a process of performing outage control on the base station based on the updated control rule.

[0106] (Appendix 18) The information processing apparatus according to Appendix 13, wherein the control unit further uses the height at which the base station is installed, the power consumption when the base station transmits radio waves, the frequency of the radio waves, and the height at which the communication device is installed to calculate an estimated value of the received power of the radio waves.

Explanation of Reference Numerals

[0107] 100 Information processing apparatus 110 Communication unit 120 Input unit 130 Display Unit 140 Memory Unit 141 Control Rule Data 142 Measured Data Base 150 Control Unit 151 Standing Wave Control Unit 152 Collection Unit 153 Learning Unit 154 Creation Unit 155 Reinforcement Learning Unit

Claims

1. By using a first model that estimates the value of the received power of the radio wave received by the communication device from the base station based on the location information of the base station and the location information of the communication device, a estimated value of the received power of the radio wave is determined. Training a second model that outputs a correction value for correcting the estimated value of the first model based on the measured value of the received power of the communication device in the operating environment using the first model and the estimated value. A learning program characterized by causing a computer to execute the process.

2. The training process trains the second model based on the difference between the measured value and the estimated value. The learning program according to claim 1.

3. The second model is a model that outputs the correction value when the location information of the base station and the location information of the communication device are input. The computer is further caused to execute a process of generating a third model that determines an estimated value of the received power of the communication device based on the location information of the base station and the location information of the communication device, the estimated value calculated based on the first model, and the correction value. The learning program according to claim 2.

4. Identifying an action based on the state and control rules of the area in the communication simulation environment, performing outage control of the base stations included in the communication simulation environment based on the identified action, and repeating a process of identifying a reward based on the information of the base stations and communication devices included in the area after the outage control and the estimated value of the received power of the communication device determined based on the third model, and performing learning to update the control rules so that the reward is maximized. The learning program according to claim 3, characterized in that the computer is further caused to execute the process.

5. The computer is further caused to execute a process of performing outage control on the base station based on the updated control rule. The learning program according to claim 4.

6. The process of calculating the estimated value further uses the height at which the base station is installed, the power consumption when the base station transmits radio waves, the frequency of the radio waves, and the height at which the communication device is installed to calculate the estimated value of the received power of the radio wave. The learning program according to claim 1.

7. By using a first model that estimates the value of the received power of the radio wave received by the communication device from the base station based on the location information of the base station and the location information of the communication device, an estimated value of the received power of the radio wave is calculated. Training a second model that outputs a correction value for correcting the estimated value of the first model based on the measured value of the received power of the communication device in the operating environment using the first model and the estimated value. A learning method characterized in that a computer executes the process.

8. By using a first model that determines the value of the received power of the radio wave received by the communication device from the base station based on the location information of the base station and the location information of the communication device, an estimated value of the received power of the radio wave is determined. Training a second model that outputs a correction value for correcting the estimated value of the first model based on the measured value of the received power of the communication device in the operating environment using the first model and the estimated value. An information processing apparatus having a control unit that executes the process.

Images