Control system, control device, control method, and control program

Abstract

A control system (10) comprises: a first identification unit (11) for identifying, by means of a first identification model, first control information for controlling a wireless network; a second identification unit (12) for identifying, by means of a second identification model, second control information for controlling the wireless network; a validity determination unit (13) for determining the validity of the first control information according to the communication performance of the wireless network, said communication performance being predicted based on the first control information; and a selection unit (14) that selects the control information for controlling the wireless network from the first control information and the second control information according to the results of the validity determination.

Description

Control system, control device, control method, and non-transitory computer-readable medium

[0001] The present disclosure relates to a control system, a control device, a control method, a control program, and a non-transitory computer-readable medium.

[0002] In recent years, artificial intelligence (AI) / machine learning (ML) have been utilized to achieve optimal control in various control systems. Related technologies are known, for example, from Patent Literature 1 and Non-Patent Literature 1.

[0003] Patent Literature 1 describes a RAN Intelligent Controller (RIC) that utilizes AI / ML to perform intelligent control in an O-RAN (Open RAN), which opens up the Radio Access Network (RAN). Non-Patent Literature 1 also summarizes guidelines for quality control of machine learning.

[0004] Japanese Patent Application Laid-Open No. 2022-105306

[0005] National Institute of Advanced Industrial Science and Technology, "Machine Learning Quality Management Guidelines", 2nd Edition (revision 2.1.0), July 5, 2021, DigiARC-TR-2021-01 / CPSEC-TR-2021001, [online], Internet<https: / / www.digiarc.aist.go.jp / publication / aiqm / AIQM-Guideline-2.1.0.pdf>

[0006] As described in Non-Patent Document 1, a model generated by machine learning does not necessarily guarantee stable operation. For this reason, it is desirable to achieve stable control when performing control using a model based on machine learning or the like in a control system such as O-RAN RIC.

[0007] In view of such problems, one of the objects of the present disclosure is to provide a control system, a control device, a control method, a control program, and a non-transitory computer-readable medium that are capable of performing stable control.

[0008] The control system of the present disclosure comprises a first identification means for identifying first control information for controlling a wireless network using a first identification model, a second identification means for identifying second control information for controlling the wireless network using a second identification model, a validity determination means for determining the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information, and a selection means for selecting control information for controlling the wireless network from the first control information and the second control information according to the validity determination result.

[0009] The control device according to the present disclosure comprises a first identification means for identifying first control information for controlling a wireless network using a first identification model, a second identification means for identifying second control information for controlling the wireless network using a second identification model, a validity determination means for determining the validity of the first control information in accordance with the communication performance of the wireless network predicted based on the first control information, and a selection means for selecting control information for controlling the wireless network from the first control information and the second control information in accordance with the validity determination result.

[0010] The control method according to the present disclosure identifies first control information for controlling a wireless network using a first identification model, identifies second control information for controlling the wireless network using a second identification model, determines the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information, and selects control information for controlling the wireless network from the first control information and the second control information according to the result of the validity determination.

[0011] The non-transitory computer-readable medium of the present disclosure is a non-transitory computer-readable medium having stored thereon a control program for causing a computer to execute a process of identifying first control information for controlling a wireless network using a first specific model, identifying second control information for controlling the wireless network using a second specific model, determining the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information, and selecting control information for controlling the wireless network from the first control information and the second control information according to the result of the validity determination.

[0012] According to the present disclosure, it is possible to provide a control system, a control device, a control method, a control program, and a non-transitory computer-readable medium that are capable of performing stable control.

[0013] 1 is a configuration diagram showing an overview of a control system according to an embodiment. 2 is a configuration diagram showing an example configuration of a control device according to an embodiment. 3 is a configuration diagram showing another example configuration of a control device according to an embodiment. 4 is a flowchart showing an overview of a control method according to an embodiment. 5 is a configuration diagram showing an example configuration of a RAN system according to a first embodiment. 6 is a configuration diagram showing a basic example configuration of a Near-RT RIC and an E2 node according to the first embodiment. 7 is a configuration diagram showing a specific example configuration of a Near-RT RIC according to the first embodiment. 8 is a flowchart showing an operation example of a Near-RT RIC according to the first embodiment. 9 is a diagram for explaining an example of handover control according to the first embodiment. 10 is a diagram for explaining an example of handover control according to the first embodiment. 11 is a diagram for explaining an example of beam control according to the first embodiment. 12 is a configuration diagram showing a basic example configuration of a Near-RT RIC and an E2 node according to a second embodiment. 13 is a configuration diagram showing a specific example configuration of a Near-RT RIC according to the second embodiment. 14 is a configuration diagram showing a specific example configuration of a Near-RT RIC according to a third embodiment. 15 is a configuration diagram showing a specific example configuration of a Near-RT RIC according to a fourth embodiment. FIG. 1 is a configuration diagram showing an overview of hardware of a computer according to an embodiment.

[0014] Hereinafter, embodiments will be described with reference to the drawings. In the drawings, the same elements are denoted by the same reference numerals, and redundant description will be omitted as necessary.

[0015] As described in Non-Patent Document 1, methods for quality control and assurance of machine learning are still under development. Although machine learning models can output appropriate results for sufficiently trained inputs, they behave unstably for insufficiently trained inputs, and therefore, expected results cannot be guaranteed. Possible methods for solving this problem include, for example, adding noise to the training data to increase the variety of the training data, artificially creating training data to increase the coverage of the training data, and using a machine learning model that can explain inference results.

[0016] However, while these methods can improve the stability of machine learning models to a certain extent, they cannot prevent runaway machine learning models, i.e., unstable behavior in response to insufficiently trained inputs. For example, in systems that require high reliability, such as those controlling AGVs (Automatic Guided Vehicles) and robots, it is difficult to ensure the required quality. Therefore, in the present embodiment, stable control is possible even when the inference accuracy of machine learning is insufficient due to insufficient learning, etc.

[0017] (Outline of the embodiment) First, an outline of the embodiment will be described. Fig. 1 shows a schematic configuration of a control system 10 according to the embodiment. For example, the control system 10 constitutes a system that controls a wireless network such as a RAN. For example, the control system 10 may include either or both of a Near-RT RIC and a Non-RT RIC, but is not limited thereto.

[0018] As shown in FIG. 1, the control system 10 includes a first specifying unit 11, a second specifying unit 12, a validity determining unit 13, and a selecting unit 14.

[0019] The first identification unit 11 identifies first control information for controlling the wireless network using the first identification model. The second identification unit 12 identifies second control information for controlling the wireless network using the second identification model. The first and second identification models are, for example, included in the control system 10, but may also be located outside the control system 10. For example, the first identification model is a learning model that has been machine-learned to acquire control information corresponding to wireless quality information acquired from the wireless network. The second identification model is a model that has higher reliability of the control information it identifies than the first identification model. A highly reliable model is a model that can output stable control information for a longer period of time, i.e., a model that does not output abnormal control information. In other words, the second identification model is less likely to output abnormal control information than the first identification model. For example, the second identification model may identify the second control information based on a predetermined rule, or may identify the second control information through theoretical calculation or simulation. The second identification model may also be a learning model that has been machine-learned to acquire control information corresponding to wireless quality information acquired from the wireless network.

[0020] The validity determination unit 13 determines the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information identified by the first identification model. That is, the validity determination unit 13 predicts the communication performance of the wireless network when the wireless network is controlled by the first control information. The validity determination unit 13 may predict the communication performance of the wireless network according to the first control information using a prediction model. For example, the prediction model may predict the communication performance by theoretical calculation or simulation, or may predict the communication performance based on a predetermined rule. Furthermore, the prediction model may be a learning model that machine-learns the communication performance according to the first control information.

[0021] The selector 14 selects control information for controlling the wireless network from the first control information and the second control information according to the result of the determination of the validity of the first control information by the validity determiner 13. That is, the selector 14 selects control information to be transmitted to the wireless network. For example, when the selector 14 determines that the first control information is invalid, the selector 14 switches the control information to be transmitted to the wireless network from the first control information to the second control information.

[0022] The control system 10 may be configured with one device or multiple devices. Fig. 2 shows an example configuration of a control device according to an embodiment. As shown in Fig. 2, the control device 20 may include the first identification unit 11, the second identification unit 12, the validity determination unit 13, and the selection unit 14 shown in Fig. 1. For example, the control device 20 may be either a Near-RT RIC or a Non-RT RIC.

[0023] 3 shows another example of the configuration of a control device according to an embodiment. As shown in FIG. 3, a control device 21 may include a first specifying unit 11, a second specifying unit 12, and a selecting unit 14, and a control device 22 may include a validity determining unit 13. The control device 21 may be a Near-RT RIC, and the control device 22 may be a Non-RT RIC.

[0024] Furthermore, part or all of the control system 10 may be placed at an edge or in the cloud using virtualization technology or the like. Part or all of the control system 10 may be placed at a specific location, or may be distributed across multiple locations. The edge is a location or platform on the base station side, and the cloud is a location or platform on the core network side away from the base station. For example, the first identification unit 11, the second identification unit 12, and the selection unit 14 may be placed at the edge, and the validity determination unit 13 may be placed in the cloud. Furthermore, the first identification unit 11, the second identification unit 12, the validity determination unit 13, and the selection unit 14 may each be distributed.

[0025] Fig. 4 shows a control method according to an embodiment. For example, the control method in Fig. 4 is executed by the control system 10 in Fig. 1, the control device 20 in Fig. 2, or the control devices 21 and 22 in Fig. 3.

[0026] As shown in FIG. 4 , the first identification unit 11 identifies first control information for controlling the wireless network using a first identification model (S11), and the second identification unit 12 identifies second control information for controlling the wireless network using a second identification model (S12). Note that S11 and S12 may be executed in parallel, or may be executed in the order of S11-S12, or vice versa. Next, the validity determination unit 13 determines the validity of the first control information according to the communication quality of the wireless network predicted based on the first control information identified using the first identification model (S13). Next, the selection unit 14 selects control information for controlling the wireless network from the first control information and the second control information according to the result of the determination of the validity of the first control information (S14).

[0027] As described above, in the embodiment, the communication quality of the wireless network is predicted from first control information identified by a first specific model such as a machine learning model, and the validity of the first control information is determined according to the predicted communication quality. Furthermore, the control information to be used for controlling the wireless network is selected according to the result of determining the validity of the first control information. As a result, even if the machine learning model exhibits unstable behavior in response to input that has not been sufficiently trained, for example, control information of another model can be selected, thereby enabling stable control of the wireless network.

[0028] (First Embodiment) Next, a first embodiment will be described. In this embodiment, an example will be described in which the validity of control information identified by a machine learning model is determined and the control information for controlling the RAN is switched. Note that, in this embodiment, an example in which radio control is performed in an O-RAN will be described as an example, but this embodiment may also be applied to a control system that performs other types of control.

[0029] 5 shows an example of the configuration of a RAN system 1 according to this embodiment. As shown in FIG. 5, the RAN system 1 includes a Near-RT RIC 100, a Non-RT RIC 200, and an E2 node 300.

[0030] The Non-RT RIC 200 and the Near-RT RIC 100, and the Non-RT RIC 200 and the E2 node 300 are communicatively connected via the O1 interface. The O1 interface is an interface for sending and receiving data and messages required mainly for operation and management. Note that the interface is a connection interface defined by a communication protocol for sending and receiving data and messages, and includes logical transmission paths and networks, as well as physical transmission paths and networks.

[0031] The Non-RT RIC 200 and the Near-RT RIC 100 are communicatively connected via an A1 interface. The Near-RT RIC 100 and the E2 node 300 are connected via an E2 interface. The A1 interface and the E2 interface are interfaces for sending and receiving data and messages required mainly for control purposes.

[0032] The E2 node 300 is a node that constitutes the RAN and includes an O-RAN Distributed Unit (O-DU) and an O-RAN Central Unit (O-CU). Note that either the O-DU or the O-CU, or both, may be referred to as the E2 node 300. The RAN is a wireless network accessed by a User Equipment (UE) and is connected to a core network such as a 5G Core network (5GC) or an Evolved Packet Core (EPC). The RAN may also include an O-RAN Remote Unit (O-RU) that constitutes an antenna. The UE is a terminal device that connects to the RAN and performs wireless communication, and may be a mobile phone, smartphone, tablet terminal, IoT (Internet of Things) terminal, or the like. The UE may also be an application device such as a robot, drone, or self-driving car that implements terminal functions.

[0033] The E2 node 300 including the O-DU and O-CU provides base station functionality. The base station is, for example, a gNB (next generation Node B) or an eNB (evolved Node B), but is not limited to these. The O-DU and O-CU are examples of nodes that provide base station functionality, and may be other network nodes.

[0034] The O-DU is a logical node that provides the radio signal control function and Layer 2 control function of the base station. The O-DU accommodates the O-RU, controls the antenna radio signal (beam) in the O-RU, and performs protocol processing such as MAC (Media Access Control) and RLC (Radio Link Control) required between the O-RU and O-CU.

[0035] The O-CU is a logical node that provides the base station's radio resource control function and data processing function higher than Layer 2. The O-CU accommodates the O-DU and performs data transmission and reception via the accommodated O-DU, QoS (Quality of Service) control, cell / UE management, handover control, and protocol processing such as PDCP (Packet Data Convergence Protocol), SDAP (Service Data Adaptation Protocol), and RRC (Radio Resource Control) required between the O-DU and the core network.

[0036] The E2 node 300 may include any number of O-DUs and O-CUs, one or more. In other words, it may include multiple base stations. The number of O-DUs and O-CUs does not necessarily have to be the same. The O-DUs and O-CUs may be located in different locations or in the same location. Furthermore, the O-DUs and O-CUs may be implemented by different virtual machines running on an edge virtualization platform, or by the same virtual machine. The O-DUs and O-CUs may be virtualized distributed units (vDUs) and virtualized central units (vCUs), and may constitute a virtual base station. The O-DUs and O-CUs may also be physical DUs and CUs. Furthermore, the E2 node 300 may be a base station device including the functions of the O-DUs and O-CUs.

[0037] The Near-RT RIC 100 is a logic function that controls and optimizes the RAN in near real time. The Near-RT RIC 100 controls the RAN in a short control period, for example, greater than or equal to 10 ms (milliseconds; the same applies hereinafter) and less than 1 s (seconds; the same applies hereinafter). The Near-RT RIC 100 collects and analyzes radio information from the E2 node 300, which includes either or both of the O-DU and O-CU, via the E2 interface, and controls the E2 node 300 according to the radio information. The Near-RT RIC 100 includes a machine learning model, which is a trained model, and analyzes the radio information and determines RAN control using the machine learning model. For example, the Near-RT RIC 100 performs control according to the radio information in accordance with a control policy acquired from the Non-RT RIC 200 via the A1 interface. The control policy is a policy related to RAN control, such as an A1 policy. The A1 policy is a guideline for RAN optimization defined on the A1 interface. The Near-RT RIC 100 is located in the same location as either or both of the O-DU and O-CU, or in a location close to either or both of the O-DU and O-CU. For example, the Near-RT RIC 100 may be implemented in the same edge virtual machine as either or both of the O-DU and O-CU.

[0038] The Non-RT RIC 200 is a logic function that controls and optimizes the RAN in non-real time. The Non-RT RIC 200 controls the RAN at a long control period of, for example, 1 second or more. The Non-RT RIC 200 manages control policies, manages the operation of the E2 node 300 and the Near-RT RIC 100, and performs learning (training) and updating of machine learning models. For example, the Non-RT RIC 200 generates control policies and notifies the Near-RT RIC 100 of the generated control policies via the A1 interface. The Non-RT RIC 200 also manages and sets the configuration information (Configuration) of the E2 node 300 based on data acquired from the E2 node 300 and the Near-RT RIC 100 via the O1 interface. The Non-RT RIC 200 is arranged in a Service Management and Orchestration (SMO) that manages and orchestrates the RAN. The SMO is arranged in a location away from the E2 node 300 and the Near-RT RIC 100, for example, on a cloud. The Non-RT RIC 200 may include the functionality of the SMO.

[0039] FIG. 6 shows an example of the basic configuration of the Near-RT RIC 100 and E2 node 300 according to this embodiment, and FIG. 7 shows an example of a specific configuration of the Near-RT RIC 100. In FIG. 7, part of the configuration shown in FIG. 6 is omitted. Note that this configuration is just an example, and other configurations may be used as long as the operation according to this embodiment, which will be described later, is possible. Part of the configuration of the Near-RT RIC 100 may be arranged in the Non-RT RIC 200. For example, the control determination unit 150 may be arranged in the Non-RT RIC 200. This makes it possible to distribute the processing load.

[0040] As shown in FIG. 6, the E2 node 300 includes a radio information acquisition unit 310, a radio information transmission unit 320, a control information reception unit 330, and a RAN control unit 340.

[0041] The radio information acquisition unit 310 acquires radio information of the RAN. In response to an instruction from the Near-RT RIC 100, the radio information acquisition unit 310 acquires information stored in the O-DU or O-CU, or radio information from the UE or O-RU. The radio information acquisition unit 310 acquires, for example, radio quality information collected from the UE as the radio information. For example, the radio quality information is a wideband CQI (Channel Quality Indicator) or the like. Other examples of the radio information including the radio quality information include a subband CQI, a signal to interference plus noise power ratio (SINR), a reference signal received power (RSRP), a reference signal received quality (RSRQ), a received signal strength indicator (RSSI), a block error rate (BLER), a usage record of a modulation and coding scheme (MCS) index, a rank indicator (RI), and an actually used multiplicity of a multi-input multi-output (MIMO).

[0042] The wireless information transmission unit 320 transmits the wireless information acquired by the wireless information acquisition unit 310 to the Near-RT RIC 100 via the E2 interface. For example, the wireless information transmission unit 320 transmits the wireless information in response to an instruction from the Near-RT RIC 100.

[0043] The control information receiver 330 receives control information from the Near-RT RIC 100 via the E2 interface. The control information is radio control information that controls the RAN according to the radio information, such as the MCS and radio resource allocation priority for each UE, and parameters for handover control and beam control. Other examples of the control information include the MIMO multiplicity, the frequency and timing of reference signal transmission, the frequency, timing, and type of measurement information (CSI (Channel State Information) Report) (indicating which of three CQI tables to use), whether PDCP duplication is used, and Bandwidth Part (indicating which BWP to use when multiple BWPs are available). The RAN controller 340 controls the RAN based on the received control information. For example, the RAN controller 340 sets the MCS and radio resource allocation priority for each UE included in the received control information in the MCS controllers and radio resource controllers in the O-DU and O-CU.

[0044] As shown in FIG. 6 , the Near-RT RIC 100 includes a wireless information receiving unit 110, a wireless information recording unit 120, a wireless control specifying unit 130, a wireless control alternative specifying unit 140, a control determining unit 150, a control switching unit 160, and a control information transmitting unit 170.

[0045] The radio information receiving unit 110 receives radio information from an E2 node 300 including either or both of an O-DU and an O-CU via the E2 interface. The radio information receiving unit 110 collects radio information from the E2 node 300 as identification data that the radio control identifying unit 130 and the radio control alternative identifying unit 140 use to identify control information. For example, the radio information receiving unit 110 may instruct the E2 node 300 on the data to be collected and the period for collecting the data.

[0046] The wireless information recording unit 120 is a database that records, i.e., stores, wireless information received from the E2 node 300. The wireless information recording unit 120 accumulates the wireless information as time-series data. Note that the wireless information receiving unit 110 may output the received wireless information to the wireless control identifying unit 130 and the wireless control alternative identifying unit 140.

[0047] The radio control specifying unit 130 specifies control information C1 for controlling the E2 node 300, including either or both of the O-DU and the O-CU, based on the radio information received from the E2 node 300 using the radio information receiving unit 110 and recorded in the radio information recording unit 120. The specified control is control of the operation of the RAN, and includes control of the radio resource allocation scheduler, beam, handover, etc., which are possible by setting the O-DU or the O-CU. The radio control specifying unit 130, for example, predicts future radio quality around the UE from the radio quality, and specifies control information C1 including the MCS and radio resource allocation priority for each UE to be set in the E2 node 300 according to the predicted radio quality.

[0048] 7, the radio control specification unit 130 includes an ML model 131 that specifies the control information C1. The radio control specification unit 130 inputs radio information collected from the E2 node 300 into the ML model 131 and specifies the control information C1 of the E2 node 300 according to the radio information. The ML model 131 is a trained model that has been machine-learned to learn the control information according to the radio information. The ML model 131 is, for example, a first specification model stored in the storage unit of the Near-RT RIC 100. The ML model 131 is a machine-learning model that specifies, i.e., infers, the control information C1 that controls the E2 node 300, which includes either or both of the O-DU and the O-CU, according to the radio information. The ML model 131 is, for example, a model that can analyze and predict time-series data. The ML model 131 may be a convolutional neural network (CNN), a recurrent neural network (RNN), a long-short term model (LSTM), or any other neural network. The ML model 131 is not limited to a neural network and may be any other machine learning model.

[0049] The wireless control alternative identification unit 140 identifies control information C2 for controlling the E2 node 300 instead of the wireless control identification unit 130. Similar to the wireless control identification unit 130, the wireless control alternative identification unit 140 identifies control information C2 for controlling the E2 node 300 based on wireless information received from the E2 node 300 using the wireless information receiving unit 110 and recorded in the wireless information recording unit 120.

[0050] 7, the radio control alternative identification unit 140 includes an alternative model 141 that identifies control information C2. The radio control alternative identification unit 140 inputs radio information collected from the E2 node 300 into the alternative model 141, and identifies control information C2 of the E2 node 300 according to the radio information. The alternative model 141 is, for example, a second identification model stored in the storage unit of the Near-RT RIC 100. Similar to the ML model 131, the alternative model 141 is any model that can identify control of the E2 node 300 including either or both of O-DU and O-CU according to the radio information.

[0051] For example, the alternative model 141 is a model with higher reliability than the ML model 131. That is, the alternative model 141 can output more stable control information than the ML model 131. Note that, since it is only necessary for the alternative model 141 to output stable control information, it may output control information with lower accuracy than the ML model 131, for example. The alternative model 141 may identify the control information based on a predetermined rule or a predetermined algorithm.

[0052] In one example, the alternative model 141 may specify control information for controlling the MCS so that a fixed target BLER is achieved for each communication requirement. For example, a correspondence table is set in advance in which target BLERs are associated with each requirement, such as a target BLER of 10% when the communication delay requirement is 100 ms, and a target BLER of 1% when the communication delay requirement is 10 ms, and the control information is specified according to the values ​​in the correspondence table. For example, when the target BLER is 10%, the MCS may be set according to the CQI reported from the UE, and when the target BLER is 1%, the MCS index may be lowered based on the CQI value reported from the UE so that the BLER becomes 1%.

[0053] In another example, the alternative model 141 may specify control information such that all the radio resource allocation priorities for the UEs are the same, thereby causing the E2 node 300 to perform radio resource allocation operations in accordance with proportional fairness scheduling, which is typically used in base stations.

[0054] In yet another example, the alternative model 141 may identify the control information by performing theoretical calculations or simulations corresponding to the RAN. For example, the alternative model 141 may perform simulations with several control parameters, calculate values ​​of a retransmission rate (BLER) or a queuing delay, and identify the best parameters as the control information.

[0055] Furthermore, like the ML model 131, the alternative model 141 may be a trained model that has learned control information corresponding to radio information through machine learning. For example, the ML model 131 may be a specialized model specialized for a specific environment, while the alternative model 141 may be a general-purpose model that can be adapted to any environment. The specialized model is a model that learns the relationship between radio information and control information in, for example, a specific base station or a specific region, and is adapted to local characteristics. The general-purpose model is a model that learns the relationship between radio information and control information in, for example, many base stations or a wide region. For example, the ML model 131 may be a model that learns control information corresponding to radio information acquired only from the RAN to be controlled, while the alternative model 141 may be a model that learns control information corresponding to radio information acquired from other RANs. Furthermore, the ML model 131 may be a short-term characteristic tracking model that learns the relationship between radio information and control information over a short period, i.e., a predetermined period, while the alternative model 141 may be a long-term general-purpose model that learns the relationship between radio information and control information over a long period, i.e., a period longer than the predetermined period. For example, the alternative model 141 may be a model incorporating a predetermined algorithm, i.e., a pre-trained model that is applied when the system is introduced. Alternatively, the alternative model 141 may be a model that is selected as the model that generates the most stable control information by measuring the performance of multiple trained models trained in different environments in advance.

[0056] The control determination unit 150 determines the validity of the control information C1 identified and output from the radio control determination unit 130. The control determination unit 150 predicts the communication performance of the RAN when the RAN (E2 node) is controlled using the control information C1 identified by the ML model 131, and determines the validity of the control information C1 based on the predicted communication performance. Determining the validity of the control information C1 also determines the validity of the operation (behavior) of the ML model 131 that identified the control information C1. Note that in this example, the validity of the control information C1 is determined, but it is also possible to determine the validity of the control information C1 and the control information C2 and output more valid control information, such as control information that shortens the delay time, as the determination result. Furthermore, the most valid control information may be determined from the control information identified by three or more models, not just two models, the ML model 131 and the alternative model 141.

[0057] 7, the control determination unit 150 includes a system model 151 and a validity determination unit 152. The control determination unit 150 inputs the control information C1 identified and output by the ML model 131 to the system model 151, and predicts a performance index P1 corresponding to the control information C1. The system model 151 is, for example, a prediction model stored in a storage unit of the Near-RT RIC 100. The system model 151 is any model capable of predicting a performance index corresponding to the control information. For example, the performance index may be a BLER (retransmission rate), a queuing delay (queuing amount), or the like, or a delay time corresponding to the BLER or queuing amount, or may be throughput, frequency utilization efficiency (PRB (Physical Resource Block) utilization rate), or the like.

[0058] The system model 151 may predict the performance index based on a predetermined rule or a predetermined algorithm. For example, the system model 151 may calculate the performance index P1 by theoretically calculating or simulating the operation of the RAN including the E2 node 300. For example, a simulation is performed using the control information C1 to calculate the BLER, queuing delay, and the like. The system model 151 may also identify the performance index P1 based on a predetermined rule, such as a correspondence table that previously associates control information with performance indexes. The system model 151 may be a trained model that has learned performance indexes corresponding to the control information through machine learning.

[0059] The validity determination unit 152 determines the validity of the performance index P1 predicted by the system model 151. For example, a threshold value of a predetermined range is set, and the validity is determined based on whether the performance index P1 is within the predetermined range. For example, the threshold value for determining validity may be set by the non-RT RIC 200. The validity determination unit 152 outputs the validity determination result to the control switching unit 160.

[0060] The control switching unit 160 switches (selects) the control information to be transmitted to the E2 node 300, i.e., the control information for controlling the RAN, depending on the result of the determination by the control determination unit 150 (validity determination unit 152) of the validity of the control information C1. If the control switching unit 160 determines that the control information C1 identified by the radio control specification unit 130 is valid, the control switching unit 160 selects the control information C1 as the control information to be transmitted to the E2 node 300, and if the control information C1 is determined to be invalid, the control switching unit 160 selects the control information C2 identified by the radio control alternative specification unit 140 as the control information to be transmitted to the E2 node 300.

[0061] The control information transmitter 170 transmits the control information C1 identified by the radio control specifying unit 130 or the control information C2 identified by the radio control alternative specifying unit 140 to the E2 node 300 in response to switching by the control switching unit 160. The control information transmitter 170 transmits the control information selected by the control switching unit 160 to the E2 node 300 including either or both of an O-DU and an O-CU via the E2 interface.

[0062] 8 shows an example of the operation of the Near-RT RIC 100 according to this embodiment. As shown in FIG. 8, the Near-RT RIC 100 receives radio information from the E2 node 300 (S101). The radio information receiving unit 110 receives radio information such as a wideband CQI from the E2 node 300 via the E2 interface. The radio information recording unit 120 records the radio information received from the E2 node 300.

[0063] Next, the Near-RT RIC 100 identifies control information C1 and C2 based on the received wireless information (S102). The wireless control identification unit 130 identifies control information C1 corresponding to the wireless information using the ML model 131. The wireless control alternative identification unit 140 identifies control information C2 corresponding to the wireless information using the alternative model 141. Note that the identification process of control information C2 by the wireless control alternative identification unit 140 is not limited to S102, and may be executed at any timing from S102 to S106.

[0064] For example, in an example where delay control is performed, the radio control specifying unit 130 and the radio control alternative specifying unit 140 specify control information for controlling retransmission delay and queuing delay, which are included in delay factors. The retransmission delay is a delay caused by retransmission of data, and the queuing delay is a delay caused by queuing transmission data in a transmission queue. For example, to control the retransmission delay, an MCS (target BLER) may be specified, or to control the queuing delay, a radio resource allocation priority for a UE, such as a priority based on a radio resource allocation ratio or a residence time in a transmission queue, may be specified.

[0065] As another example, handover control may be performed. FIGS. 9 and 10 show an example of handover control. FIG. 9 shows the handover procedure (S21 to S23), and FIG. 10 shows the radio wave strength at the UE at each time corresponding to S21 to S22. As shown in FIGS. 9 and 10, in S21, when the radio wave strength of the base station A to which the UE belongs deteriorates below a predetermined threshold TH1, the UE starts transmitting a Measurement Report indicating radio wave quality information of the cell to which the UE belongs (base station A) and the neighboring cell (base station B). Next, in S22, base station A determines whether to perform a handover based on the radio wave quality information indicated in the Measurement Report transmitted from the UE and a predetermined threshold TH2. If it determines that a handover is necessary, it instructs the UE to perform a handover to base station B. Next, in S23, the UE performs a handover to base station B instructed by base station A.

[0066] In such an example of handover control, the radio control specifying unit 130 and the radio control alternative specifying unit 140 may specify, as control information, a trigger threshold TH1 at which the UE starts transmitting a Measurement Report. For example, the threshold TH1 is a threshold for the radio wave quality (RSRP (Reference Signal Received Power), RSRQ (Reference Signal Received Quality)) value of the home station or the difference value with a neighboring cell. Furthermore, as control information, a neighbor cell relation table (NCRT) may be specified, which is neighbor cell information. By specifying the neighbor cell information, it is possible to narrow down the neighbor cells that report radio wave quality in the Measurement Report. By specifying the neighbor cell information using a movement pattern (handover pattern), etc., it is possible to exclude base stations that have not experienced handover in the past from the candidate base stations. Furthermore, as control information, a threshold TH2 at which the base station determines whether to perform a handover may be specified. For example, the threshold TH2 is a threshold for the radio wave quality (RSRP, RSRQ) of the associated station or the difference between the radio wave quality and that of an adjacent cell.

[0067] As another example, beam control may be performed. FIG. 11 shows an example of beam control. As shown in FIG. 11, multiple beams are transmitted from one base station. Forming beams can increase the cell radius, and is particularly intended for use in high frequency bands. The base station transmits reference signals corresponding to multiple beams in an SS (Synchronization Signal) Block. Each reference signal includes an SSB Index, which is an identifier for the beam. The UE measures the strength of each beam and reports it to the base station. The base station instructs the UE to use the strongest beam, and then adjusts the directivity of the antennas of the base station and UE using CSI-RS, a reference signal transmitted for each UE, so that the directivity faces each other. The beams facing each other are called a beam-pair. When the UE moves and the intensity of a beam with another SSB index becomes stronger, the UE reports this to the base station via a CSI-Report, and the base station determines whether to switch beams based on a predetermined threshold and switches the beams.

[0068] In such an example of beam control, the radio control specifying unit 130 and the radio control alternative specifying unit 140 may specify, as control information, the direction of the beam to be included in the SSB transmitted by the base station. Furthermore, as control information, the frequency and timing of reporting beam strength by the UE's SI-Report may be specified. There are periodic and aperiodic CSI-Reports, and in the case of periodic, the cycle can be controlled, and in the case of aperiodic, the reporting timing can be controlled. Furthermore, as control information, a threshold value for determining beam switching by the base station may be specified. For example, this threshold value is the value or difference of radio wave strength for each beam.

[0069] Next, the Near-RT RIC 100 predicts a performance index P1 based on the identified control information C1 (S103). The control determination unit 150 uses the system model 151 to predict the performance index P1 (performance index value) when the E2 node 300 uses the control information C1.

[0070] For example, in an example where delay control is performed, the system model 151 may predict, as a performance index, a BLER (retransmission rate) related to retransmission delay, or may predict a queuing amount in a transmission queue related to queuing delay. The queuing amount may be something that can be inferred, such as a short-term throughput average and a long-term throughput average. The system model 151 may predict a delay time according to the BLER or the queuing amount.

[0071] 9 and 10, the system model 151 may predict the time required for handover, events such as Handover Failure (RLF: Radio Link Failure) and Ping-Pong, and radio wave quality values ​​after handover as performance indicators. The system model 151 may predict the time required for handover according to each event and radio wave quality value.

[0072] 11, the system model 151 may predict, as performance indicators, a Beam-Failure event due to loss of a beam pair, an event such as Ping-Pong, and a radio wave quality value after beam switching. The system model 151 may also predict the time required for beam switching according to each event and radio wave quality value.

[0073] Next, the Near-RT RIC 100 determines whether the predicted performance index P1 is within a predetermined range (S104). The validity determination unit 152 determines whether the performance index P1 predicted from the control information C1 by the system model 151 is within a predetermined range.

[0074] The Near-RT RIC 100 selects the control information C1 if the predicted performance index P1 is within a predetermined range (S105), and selects the control information C2 if the predicted performance index P1 is outside the predetermined range (S106). If the performance index P1 predicted from the control information C1 is within the predetermined range, the control switching unit 160 inputs the control information C1 identified by the ML model 131 to the control information transmission unit 170, and if the performance index P1 predicted from the control information C1 is outside the predetermined range, the control switching unit 160 switches to input the control information C2 identified by the alternative model 141 to the control information transmission unit 170.

[0075] Next, the Near-RT RIC 100 transmits control information to the E2 node 300 (S107). The control information transmission unit 170 transmits the selected control information C1 or C2 to the E2 node 300 via the E2 interface. Thereafter, steps S101 to S107 are repeatedly executed. For example, even if the control information to be transmitted is switched from control information C1 to control information C2, if the control information C1 subsequently becomes an appropriate value, the control information to be transmitted is returned to control information C1.

[0076] As described above, in this embodiment, control information identified by the machine learning model is input into a wireless communication system model, a performance index related to wireless quality is calculated by theoretical calculation or simulation, and the validity of the control information is determined based on the calculated performance index. If the control information is determined to be invalid, the control information is switched to control information identified using an alternative model based on predetermined rules or theories. This allows stable wireless control to be performed even if the machine learning model behaves unstable, thereby suppressing deterioration in the quality of wireless communication.

[0077] (Second Embodiment) Next, a second embodiment will be described. In this embodiment, an example will be described in which parameters of a system model that determines the validity of control information are corrected using performance indicators collected from the E2 node. Note that this embodiment can be implemented in combination with the first embodiment, and the configurations shown in the first embodiment may be used as appropriate.

[0078] Fig. 12 shows a basic configuration example of the Near-RT RIC 100 and E2 node 300 according to this embodiment, and Fig. 13 shows a specific configuration example of the Near-RT RIC 100. In Fig. 13, some of the configuration shown in Fig. 12 is omitted.

[0079] As shown in FIG. 12 , the E2 node 300 according to this embodiment includes a performance indicator transmission unit 350 in addition to the configuration of the first embodiment. The performance indicator transmission unit 350 transmits a performance indicator P2 to the Near-RT RIC 100 via the E2 interface. The performance indicator transmission unit 350 acquires or measures the performance indicator P2 based on information stored in the O-DU or O-CU and information collected from the UE or O-RU, and transmits the acquired or measured performance indicator P2 (performance indicator value) to the Near-RT RIC 100. For example, after the RAN control unit 340 performs control based on control information received from the Near-RT RIC 100, the performance indicator transmission unit 350 transmits a performance indicator as a result of the actual control. The acquired and transmitted performance indicator P2 may be instructed by the Near-RT RIC 100 or may be registered in the E2 node 300. The performance indicator P2 is a performance indicator that can be observed at least in the RAN. For example, similar to the performance index P1 shown in the system model 151 of the first embodiment, the performance index P2 is, in an example where delay control is performed, the actual values ​​of BLER and MCS, the amount of queuing, etc.; in an example where handover control is performed, the performance index P2 is, in an example where handover control is performed, the time required for handover, events such as Handover Failure and Ping-Pong, the radio wave quality value after handover, etc.; and in an example where beam control is performed, the performance index P2 is, in an example where beam control is performed, the event of Beam-Failure due to loss of beam pair, events such as Ping-Pong, the radio wave quality value after beam switching, etc.

[0080] 12, the Near-RT RIC 100 according to this embodiment includes a performance index receiving unit 180 in addition to the configuration of the first embodiment. The performance index receiving unit 180 receives a performance index P2 from an E2 node 300 that includes either or both of an O-DU and an O-CU controlled by control information via the E2 interface. The performance index receiving unit 180 outputs the received performance index P2 to the control determining unit 150.

[0081] As shown in FIG. 13 , the control determination unit 150 according to this embodiment includes a parameter correction unit 153 in addition to the configuration of the first embodiment. The parameter correction unit 153 corrects the parameters of the system model 151 based on the performance index P2 acquired from the E2 node 300. The parameter correction unit 153 adjusts (calibrates) the internal parameters used to calculate the performance index P1 in the system model 151 using the performance index P2, which is an actual measurement value. For example, the internal parameters are parameters in a predetermined mathematical formula used in the calculation of the system model 151. When control information is input to the system model 151, the parameter correction unit 153 corrects the parameters so as to output a performance index that is the same as or close to the actual measurement value. For example, the correction may be performed using Bayesian estimation, a Kalman filter, or the like. The other configurations are the same as those of the first embodiment.

[0082] As described above, in addition to the configuration of the first embodiment, actual performance indicators may be collected from the E2 node, and the parameters of the system model may be corrected based on the collected performance indicators. This improves the accuracy of performance indicator predictions by the system model, allowing for more accurate determination of the validity of control information.

[0083] (Third Embodiment) Next, a third embodiment will be described. In this embodiment, an example will be described in which the performance index predicted by the ML model is verified using the performance index collected from the E2 node. Note that this embodiment can be implemented in combination with the first or second embodiment, and the configurations shown in the first or second embodiment may be used as appropriate.

[0084] For example, a basic configuration example of the Near-RT RIC 100 and E2 node 300 according to this embodiment is the same as that shown in Fig. 12 of the second embodiment, and Fig. 14 shows a specific configuration example of the Near-RT RIC 100 according to this embodiment. In Fig. 14, part of the configuration shown in Fig. 12 is omitted.

[0085] 14 , in this embodiment, the ML model 131 predicts and outputs a performance index P3. That is, the ML model 131 identifies control information C1 according to input radio information and predicts a performance index P3 of the RAN according to the control information C1. The ML model 131 is a trained model that has trained to generate the control information C1 and the performance index P3 according to the radio information. The performance index P3 is a performance index that can be observed in the RAN, similar to the performance index P2 acquired from the E2 node 300.

[0086] Furthermore, the control determiner 150 according to this embodiment includes an actual value verifier 154 in addition to the configuration of the first embodiment. The actual value verifier 154 compares a performance index P3 (performance index value) predicted by the ML model 131 with a performance index P2 (performance index value), which is an actual value (actual measurement value) acquired from the E2 node 300, and verifies the predicted performance index P3. For example, the actual value verifier 154 calculates the difference between the predicted performance index P3 and the acquired performance index P2, and determines whether the difference is within a predetermined range. If the difference is within the predetermined range, it is determined that the performance index P3 and the performance index P2 match, and if the difference is outside the predetermined range, it is determined that the performance index P3 and the performance index P2 do not match.

[0087] As in the first embodiment, the validity determination unit 152 determines the performance index P1 calculated by the system model 151 from the control information C1 and determines the validity based on the verification result of the actual value verification unit 154. In this case, the validity determined is the validity of the control information C1 and also the validity of the ML model 131 that generated the control information C1. For example, if the predicted performance index P3 and the acquired performance index P2 match, the control information C1 may be determined to be valid. If the predicted performance index P3 and the acquired performance index P2 do not match, the control information C1 may be determined to be invalid. If the performance index P1 calculated by the system model 151 is within a predetermined range and the predicted performance index P3 and the acquired performance index P2 match, the control information C1 may be determined to be valid. If the performance index P1 calculated by the system model 151 is outside the predetermined range or the predicted performance index P3 and the acquired performance index P2 do not match, the control information C1 may be determined to be invalid. The other configurations are the same as those in the first embodiment.

[0088] As described above, in addition to the configuration of the first embodiment, actual performance indicators may be collected from the E2 node, the performance indicators predicted by the ML model may be verified based on the collected performance indicators, and the validity of the control information specified by the ML model may be determined based on the verification results. This makes it possible to determine the validity of the control information specified by the ML model while verifying the performance indicators predicted by the ML model, thereby making it possible to more accurately determine the validity of the control information.

[0089] (Fourth Embodiment) Next, a fourth embodiment will be described. In this embodiment, an example will be described in which the determination of the validity of control information is used for learning an ML model. Note that this embodiment can be implemented in combination with any of the first to third embodiments, and each configuration shown in any of the first to third embodiments may be used as appropriate.

[0090] For example, a basic configuration example of the Near-RT RIC 100 and E2 node 300 according to this embodiment is the same as that shown in Fig. 6 of the first embodiment, and Fig. 15 shows a specific configuration example of the Near-RT RIC 100 according to this embodiment. Fig. 15 omits some of the configuration shown in Fig. 6. Note that in this example, it is sufficient that the learning operation of the ML model 131 is possible, and therefore the control switching unit 160 and the control information transmission unit 170 may not be provided.

[0091] 15 , the radio control specification unit 130 according to the present embodiment includes a penalty assigning unit 132 in addition to the configuration of the first embodiment. The penalty assigning unit 132 is a learning unit that causes the ML model 131 to perform machine learning in accordance with the determination result of the validity determination unit 152 when learning the ML model 131. When it is determined that the control information C1 is invalid, the penalty assigning unit 132 imposes a penalty on the control information C1 that is determined to be invalid. The ML model 131 learns to generate valid control information in accordance with the radio information based on the penalty.

[0092] As described above, the validity determination result in the first embodiment may be used for training the ML model, thereby improving the accuracy with which the ML model identifies control information, and enabling stable control information to be output.

[0093] The present disclosure is not limited to the above-described embodiment, and can be modified as appropriate within the scope of the present disclosure.

[0094] Each component in the above-described embodiments may be configured with hardware, software, or both, and may be configured with a single piece of hardware or software, or may be configured with multiple pieces of hardware or software. Each device, including the Non-RT RIC and Near-RT RIC, and each function (processing) may be implemented by a computer 30 having a network interface 31, a processor 32 such as a CPU (Central Processing Unit), and a memory 33 serving as a storage device, as shown in FIG. 16 . The network interface 31 may include a network interface card (NIC) for communicating with devices including network nodes. For example, a program for performing the method (control method) in the embodiment may be stored in the memory 33, and each function may be implemented by executing the program stored in the memory 33 by the processor 32.

[0095] These programs include instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The programs may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable media or tangible storage media include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disc (DVD), Blu-ray disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The programs may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagated signals.

[0096] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure.

[0097] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes. (Supplementary Note 1) A control system comprising: a first identification means for identifying first control information for controlling a wireless network using a first identification model; a second identification means for identifying second control information for controlling the wireless network using a second identification model; a validity determination means for determining validity of the first control information according to communication performance of the wireless network predicted based on the first control information; and a selection means for selecting control information for controlling the wireless network from the first control information and the second control information according to a result of the validity determination. (Supplementary Note 2) The control system according to Supplementary Note 1, wherein the second identification model is a model that is less likely to output abnormal control information than the first identification model. (Supplementary Note 3) The control system according to Supplementary Note 1 or 2, wherein the first identification model is a learning model obtained by machine learning control information according to wireless quality information acquired from the wireless network. (Supplementary Note 4) The control system according to any one of Supplementary Notes 1 to 3, wherein the second identification model is a model that identifies the second control information based on a predetermined rule for wireless quality information acquired from the wireless network. (Supplementary Note 5) The control system according to any one of Supplementary Notes 1 to 3, wherein the second identification model is a model that identifies the second control information by performing a theoretical calculation or a simulation corresponding to the wireless network for wireless quality information acquired from the wireless network. (Supplementary Note 6) The control system according to any one of Supplementary Notes 1 to 3, wherein the second identification model is a learning model obtained by machine learning control information corresponding to the wireless quality information acquired from the wireless network. (Supplementary Note 7) The control system according to Supplementary Note 3, wherein the first identification model is a learning model obtained by machine learning control information corresponding to wireless quality information acquired only from the wireless network, and the second identification model is a learning model obtained by machine learning control information corresponding to wireless quality information acquired from wireless networks including other wireless networks.(Supplementary Note 8) The control system according to Supplementary Note 3, wherein the first specific model is a learning model obtained by machine learning control information corresponding to wireless quality information acquired from the wireless network for a predetermined period, and the second specific model is a learning model obtained by machine learning control information corresponding to wireless quality information acquired from the wireless network for a period longer than the predetermined period. (Supplementary Note 9) The control system according to any one of Supplements 1 to 8, wherein the validity determination means predicts communication performance of the wireless network according to the first control information using a prediction model. (Supplementary Note 10) The control system according to Supplementary Note 9, wherein the prediction model is a model that predicts the communication performance by performing theoretical calculation or simulation corresponding to the wireless network for the first control information. (Supplementary Note 11) The control system according to Supplementary Note 9, wherein the prediction model is a model that predicts the communication performance for the first control information based on a predetermined rule. (Supplementary Note 12) The control system according to Supplementary Note 9, wherein the prediction model is a learning model obtained by machine learning communication performance corresponding to the first control information. (Supplementary Note 13) The control system according to any one of Supplements 9 to 12, wherein the validity determination means corrects parameters used by the prediction model to predict communication performance based on communication performance acquired from the wireless network. (Supplementary Note 14) The control system according to any one of Supplements 1 to 13, wherein the first identification means further predicts communication performance of the wireless network using the first identification model, and the validity determination means determines the validity based on the communication performance of the wireless network predicted by the first identification model and the communication performance acquired from the wireless network. (Supplementary Note 15) The control system according to Supplementary Note 3, 7, or 8, further comprising learning means that performs machine learning on a learning model of the first identification model based on a result of the determination of validity. (Supplementary Note 16) The control system according to Supplementary Note 15, wherein the learning means imposes a penalty on the first control information when the first control information is determined to be invalid.(Supplementary Note 17) The control system according to any one of Supplements 1 to 16, wherein the control system includes a Near-RT (Real Time) RIC (RAN Intelligent Controller) that controls a RAN (Radio Access Network), or a Non-RT RIC. (Supplementary Note 18) The control system according to Supplementary Note 17, wherein the Near-RT RIC includes the first specifying means, the second specifying means, and the selecting means, and the Non-RT RIC includes the validity determining means. (Supplementary Note 19) A control device comprising: first identification means for identifying first control information for controlling a wireless network using a first identification model, second identification means for identifying second control information for controlling the wireless network using a second identification model, validity determination means for determining validity of the first control information according to communication performance of the wireless network predicted based on the first control information, and selection means for selecting control information for controlling the wireless network from the first control information and the second control information according to the validity determination result. (Supplementary Note 20) A control method comprising: identifying first control information for controlling a wireless network using a first identification model, identifying second control information for controlling the wireless network using a second identification model, determining validity of the first control information according to communication performance of the wireless network predicted based on the first control information, and selecting control information for controlling the wireless network from the first control information and the second control information according to the validity determination result. (Supplementary Note 21) A non-transitory computer-readable medium storing a control program for causing a computer to execute the following processes: identifying first control information for controlling a wireless network using a first identification model; identifying second control information for controlling the wireless network using a second identification model; determining the validity of the first control information according to communication performance of the wireless network predicted based on the first control information; and selecting control information for controlling the wireless network from the first control information and the second control information according to the validity determination result.

[0098] 1 RAN system 10 Control system 11 First identification unit 12 Second identification unit 13 Validity determination unit 14 Selection unit 20, 21, 22 Control device 30 Computer 31 Network interface 32 Processor 33 Memory 100 Near-RT RIC 110 Radio information receiving unit 120 Radio information recording unit 130 Radio control identification unit 131 ML model 132 Penalty assignment unit 140 Radio control alternative identification unit 141 Alternative model 150 Control determination unit 151 System model 152 Validity determination unit 153 Parameter correction unit 154 Performance value verification unit 160 Control switching unit 170 Control information transmission unit 180 Performance index reception unit 200 Non-RT RIC 300 E2 node 310 Radio information acquisition unit 320 Radio information transmitting unit 330 Control information receiving unit 340 RAN control unit 350 Performance index transmitting unit

Claims

1. A first specifying means for specifying first control information for controlling a wireless network by a first specific model; A second specifying means for specifying second control information for controlling the wireless network by a second specific model; A validity determination means for determining the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information; A selection means for selecting control information for controlling the wireless network from the first control information and the second control information according to the determination result of the validity; A control system comprising the above.

2. The second specific model is a model with a lower probability of outputting abnormal control information than the first specific model. The control system according to Claim 1.

3. The first specific model is a learning model obtained by machine learning control information according to wireless quality information acquired from the wireless network. The control system according to Claim 1 or 2.

4. The second specific model is a model for specifying the second control information based on a predetermined rule for wireless quality information acquired from the wireless network, or a model for specifying the second control information by performing theoretical calculation or simulation corresponding to the wireless network for wireless quality information acquired from the wireless network. The control system according to Claim 1 or 2.

5. The second specific model is a learning model obtained by machine learning control information according to wireless quality information acquired from the wireless network. The control system according to Claim 1 or 2.

6. The first specific model is a learning model obtained by machine learning control information according to wireless quality information acquired only from the wireless network, The second specific model is a learning model obtained by machine learning control information according to wireless quality information acquired from a wireless network including other wireless networks. The control system according to Claim 3.

7. The first specific model is a learning model obtained by machine learning control information according to wireless quality information acquired from the wireless network for a predetermined period, The second specific model is a learning model obtained by machine learning control information according to wireless quality information acquired from the wireless network for a period longer than the predetermined period. The control system according to Claim 3.

8. a first specifying means for specifying first control information for controlling a wireless network according to a first specific model; a second specifying means for specifying second control information for controlling the wireless network according to a second specific model; a validity determination means for determining the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information; a selection means for selecting control information for controlling the wireless network from the first control information and the second control information according to the determination result of the validity; A control device comprising the above.

9. Specify first control information for controlling a wireless network according to a first specific model, Specify second control information for controlling the wireless network according to a second specific model, Determine the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information, Select control information for controlling the wireless network from the first control information and the second control information according to the determination result of the validity, Control method.

10. Specify first control information for controlling a wireless network according to a first specific model, Specify second control information for controlling the wireless network according to a second specific model, Determine the validity of the first control information according to the communication performance of the wireless network predicted based on the first control information, Select control information for controlling the wireless network from the first control information and the second control information according to the determination result of the validity, A control program for causing a computer to execute the process.