Control device, processing device, control method, processing method, and program for o-ran configuration and operation

Abstract

In the present invention, a control device functioning as a Non-RealTime RAN Intelligent Controller (Non-RealTime RIC) in an Open-Radio Access Network (O-RAN) identifies processing to be executed on second information acquired by a Near-RealTime RIC in order to acquire first information used as input for machine learning and a trained model obtained through said machine learning, generates the trained model through the machine learning that uses the first information obtained by the identified processing as input, communicates third information indicating the identified processing to the Near-RealTime RIC, and provides the trained model to the Near-RealTime RIC.

Description

Control device, processing device, control method, processing method and program for setting up and operating O-RAN

[0001] The present invention relates to a technique for setting up and operating an O-RAN (Open-Radio Access Network).

[0002] Standardization of O-RAN (Open-RAN), which makes radio access networks (RANs) more open and intelligent, is underway. In O-RAN, a RAN Intelligent Controller (RIC) is defined to intelligently configure and operate the RAN by utilizing machine learning and other techniques using information measured within the RAN. Two types of RIC are defined: non-real-time RIC and quasi-real-time RIC, which have different control periods. Non-Patent Document 1 describes a method (interface, etc.) for exchanging messages between the non-real-time RIC and the quasi-real-time RIC.

[0003] O-RAN Working Group 2, “A1 Interface: Application Protocol v04.00”, O-RAN Alliance, 2023

[0004] The present invention provides a technique for improving the accuracy of RAN control in a communication system using O-RAN.

[0005] A control device according to one aspect of the present invention is a control device that functions as a Non-Real-Time RAN Intelligent Controller (non-real-time RIC) in an Open-Radio Access Network (O-RAN), and includes: an identification means that identifies a processing process to be performed on second information acquired by a near-real-time RIC in order to obtain first information to be used as input for machine learning and a trained model obtained by the machine learning; a generation means that generates the trained model by machine learning using as input the first information acquired by the identified processing process; a communication means that communicates third information indicating the identified processing process with the near-real-time RIC; and a provision means that provides the trained model to the near-real-time RIC.

[0006] A processing device according to one aspect of the present invention is a processing device that functions as a Near-Real-Time RAN Intelligent Controller (near-real-time RIC) in an Open-Radio Access Network (O-RAN), and includes: an acquisition means for acquiring second information including at least one of information collected from an E2 node or information generated in the processing device; a communication means for communicating third information indicating a processing process to be performed on the second information with a non-real-time RIC to acquire first information to be used as input for machine learning and a trained model obtained by the machine learning; a processing means for performing the processing process on the second information to acquire first information; an acquisition means for acquiring a trained model from the non-real-time RIC; and a control means for controlling the RAN using an inference result obtained by inputting the first information into the trained model.

[0007] According to the present invention, it is possible to improve the accuracy of RAN control in a communication system using O-RAN.

[0008] Other features and advantages of the present invention will become apparent from the following description taken in conjunction with the accompanying drawings, in which the same or similar elements are designated by the same reference numerals.

[0009] The accompanying drawings are incorporated in and constitute a part of the specification, illustrate embodiments of the present invention, and together with the description are used to explain the principles of the present invention. A diagram showing an example of the configuration of a communication system. A diagram showing an example of a sequence between a control device and a processing device. A diagram showing an example of the operation of the control device. A diagram showing an example of the operation of the processing device. A diagram showing an example of the operation of the control device. A diagram showing an example of the operation of the processing device. A diagram showing an example of the hardware configuration of the control device and the processing device. A diagram showing an example of the functional configuration of the control device. A diagram showing an example of the functional configuration of the processing device.

[0010] Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the scope of the invention as claimed, and not all combinations of features described in the embodiments are necessarily essential to the invention. Two or more of the features described in the embodiments may be combined in any desired manner. Furthermore, the same reference numerals are used to designate identical or similar components, and redundant descriptions will be omitted.

[0011] (System Configuration) Fig. 1 shows an example of the configuration of a communication system according to this embodiment. This communication system is configured using an Open-Radio Access Network (O-RAN), which is being standardized to open up and intelligentize radio access networks (RANs). The O-RAN is configured to include, for example, an SMO 101, a control device 102 and a processing device 103 in the SMO 101, and an E2 node 104. Note that SMO is an abbreviation for Service Management and Orchestration. The SMO 101 monitors, manages, and controls the entire system. The control device 102 and the processing device 103 work together to design and set RAN parameters, automate and optimize RAN operations, and so on. The control device 102 may be a Non-RealTime RAN Intelligent Controller (non-real-time RIC) defined in the O-RAN standard. Furthermore, the processing device 103 may be a Near-RealTime RIC (quasi-real-time RIC) defined in the O-RAN standard. The control device 102 and the processing device 103 may be separate entities or may be implemented in a single piece of hardware. The control device 102 determines policies for controlling the long-term behavior of the entire system. For example, the control device 102 generates a trained model for predicting the communication quality (throughput, delay, etc.) of the RAN using machine learning, and generates policies for controlling the RAN. The processing device 103 controls the short-term behavior of the E2 node 104. For example, the processing device 103 predicts the communication quality of the RAN using the trained model generated by the control device 102, and controls the RAN in accordance with the policy generated by the control device 102. The E2 node 104 is, for example, an O-RAN Central Unit (O-CU), an O-RAN Distributed Unit (O-DU), etc. The O-CU has functions such as establishing and disconnecting connections between the O-DU, O-RU, and User Equipment (UE) connected thereto, and controlling communications. The O-CU performs, for example, communication control such as control signaling, and transfer processing for communicating user plane data with terminal devices.In one example, when the O-CU performs upper layer processing of communication processing with the UE, the O-DU can perform lower layer processing of the communication processing below the medium access control (MAC) layer.

[0012] The interface between the SMO 101 and each of the processing device 103 and the E2 node 104 may be referred to as an O1 interface. The SMO 101 may acquire performance information, fault information, signaling information, information on user data characteristics, and the like accumulated in the E2 node 104 from the E2 node 104 using the O1 interface. The control device 102 included in the SMO 101 may generate configuration parameters optimized for the wireless environment and traffic load, for example, using the information acquired by the SMO 101. Note that the processing device 102 may be separate from the SMO 101. The SMO 101 may provide the configuration parameters generated by the control device 102 to the E2 node 104 using the O1 interface.

[0013] The interface between the control device 102 and the processing device 103 may be referred to as an A1 interface. As an example, the control device 102 may generate a policy for controlling the RAN using information acquired by the SMO 101, and provide the policy to the processing device 103 using the A1 interface. In this embodiment, the policy generated by the control device 102 for controlling the RAN is referred to as an A1 policy. The A1 policy may be called by other names. The A1 policy may also include performance targets for communication quality (throughput, delay time, etc.) for specific users, slices, cells, etc. The control device 102 may also provide the learned model generated by itself to the processing device 103 using the A1 interface.

[0014] The interface between the processing device 103 and the E2 node 104 may be referred to as an E2 interface. The processing device 103 may acquire information from the E2 node 104 using the E2 interface and perform analysis using the acquired information. Furthermore, the processing device 103 may control the E2 node 104 using the results of its own analysis in accordance with the A1 policy provided by the control device 102. For example, the processing device 103 may issue a handover instruction to the E2 node 104. In this way, information exchange and control between the devices constituting the O-RAN are performed using interfaces defined between them. Note that information exchange and control between the devices may also be performed using interfaces other than those described above, such as a Y1 interface or an O2 interface.

[0015] As described above, the control device 102 and the processing device 103 cooperate with each other to configure and operate the RAN. However, the information available to the control device 102 and the processing device 103 differs. For example, with respect to measurement information such as communication quality in a cell to be controlled, the processing device 103 can acquire measurement information for each UE from the E2 node 104 by using the E2 interface. On the other hand, the measurement information that the control device 102 can acquire via the O1 interface is information for each cell. In other words, the information available to the control device 102 may be coarser in granularity than the information available to the processing device 103. Here, consider the case where the control device 102 generates a trained model through machine learning using the acquired measurement information as input, the processing device 103 predicts the communication quality of the RAN by performing inference using the trained model, and the processing device 103 then controls the E2 node 104. In this case, if the training data available as input for machine learning by the control device 102 is cell-level measurement information (e.g., average or total values ​​of throughput and delay time), the communication quality that can be predicted in the future may also be cell-level information. Therefore, for example, in a situation where fine control on a UE-by-UE basis is required, a learned model with the required granularity cannot be generated, and highly accurate RAN control may not be possible. Furthermore, as an example, the A1 policy generated by the control device 102 may include a performance target on a UE-by-UE basis. In this case, if the measurement information within the RAN that the control device 102 can acquire is only cell-by-cell information, the control device 102 cannot grasp the achievement status of the performance target for each UE in the actual RAN. Therefore, there may be a discrepancy between the performance target for each UE set by the control device 102 and the actual communication quality status in the RAN. As such, the limited information available to the control device 102 may result in a situation where the RAN is not configured or operated according to the required accuracy.

[0016] In view of such circumstances, this embodiment provides an interface between the control device 102 and the processing device 103 so that the control device 102 can receive information held by the processing device 103. An example of the exchange of messages and measurement information between the control device 102 and the processing device 103 in this embodiment will be described below.

[0017] (Data Request by Control Device) Fig. 2 shows an example of a sequence between the control device 102 and the processing device 103. This sequence can be executed using, for example, the A1 interface. Note that this sequence may also be executed using an O1, Y1, A1-EI interface, or the like, other than the A1 interface. Also, Fig. 2 shows an example in which messages are directly exchanged between the control device 102 and the processing device 103, but messages may also be exchanged indirectly, for example, via other components in the O-RAN.

[0018] First, the control device 102 transmits a request message to the processing device 103 requesting the provision of measurement information held by the processing device 103 (S201). The request message may be called by another name, for example, a Data Collection Request. Upon receiving the request message, the processing device 103 determines whether or not to provide the measurement information and transmits the determination result to the control device 102 as a response message (S202). The response message may be called by another name, for example, a Data Collection Response. If the processing device 103 accepts the provision of the measurement information, it transmits the response message and then starts providing the measurement information in accordance with the request included in the request message (S203). Note that if the processing device 103 accepts the request, it may start providing the measurement information without transmitting a response message. Alternatively, the response message and the measurement information may be transmitted simultaneously or in the same message. If the processing device 103 does not accept the request included in the request message, it may transmit a response message including a reason for refusal. The processing device 103 may notify the control device 102 of, for example, the reason for rejection, that the processing device 103 does not have the requested measurement information, that the processing device 103 does not have the functionality to perform the processing required to provide the requested measurement information, etc. Note that if the processing device 103 does not have the requested measurement information at the time of receiving a request message from the control device 102, it may perform new measurements or collect information to provide the information to the control device 102. Also, the processing device 103 may send a message to start a sequence for providing the measurement information before the control device 102. For example, the processing device 103 may periodically poll the control device 102. Also, the request message or the response message may be transmitted integrated with another message.

[0019] (Specifying Measurement Information in Request Message) The request message may include, for example, information specifying a period during which measurement information is measured (which may be referred to as a measurement period, or measurement time). The measurement period may include information specifying one or more of the time when measurement starts (which may be referred to as a measurement start time), the time when measurement ends (which may be referred to as a measurement end time), and the interval at which measurement is performed (which may be referred to as a measurement period). The request message may also include, for example, information specifying a period during which measurement information is provided (which may be referred to as a reporting time). The provision period may include information specifying one or more of the time when provision starts (which may be called a reporting start time), the time when provision ends (which may be called a reporting end time), and the interval at which provision is performed (which may be called a reporting period). The control device 102 may specify an appropriate cycle or period for the measurement period or provision period depending on the purpose of using the measurement information, the characteristics of the measurement information, etc., and notify the processing device 103 of the same using a request message.

[0020] The request message may include, for example, information specifying a procedure for providing the measurement information (which may be referred to as a provision procedure or collection method). For example, as an example of a provision procedure, the processing device 103 may periodically transmit the measurement information over a specified provision period. Such a provision procedure may be referred to as streaming. For example, when the control device 102 modifies or adds an A1 policy to reflect the RAN status, it may be appropriate for the measurement information to be provided in real time. In such a case, the control device 102 may request, in the request message, the periodic provision of the measurement information so that the measurement information can be used in a timely manner for generating, modifying, or adding an A1 policy. As another example of a provision procedure, the processing device 103 may compile information collected over a specified measurement period into one or more files and provide them to the control device 102 all at once. Such a provision procedure may be referred to as file-based. For example, measurement information used for machine learning may have a large data size, and therefore the A1 interface may be occupied for a long period of time when this measurement information is provided. Furthermore, the measurement information used for machine learning may require processing such as preprocessing. In such a case, the control device 102 may request that the measurement information be provided after the measurement period ends and after performing processing such as compression of the measurement information or preprocessing for machine learning, or during a time when the A1 interface is not used. Furthermore, as will be described later, a method for processing the measurement information (which may be referred to as a processing method) may be specified in the request message. Note that the measurement information may be provided by a method other than the above-described providing procedure. The processing device 103 provides the measurement information to the control device 102 according to the specified procedure.

[0021] The request message may include, for example, information specifying the type of measurement information (which may be referred to as a data name). As an example, the measurement information may include measurement information specified in other O-RAN interfaces other than the A1 interface (which may be referred to as O-RAN interface data), measurement information newly specified in the A1 interface (which may be referred to as interface specific data), and measurement information specified independently by a product development vendor (which may be referred to as vendor specific data). For example, measurement information specified in other O-RAN interfaces other than the A1 interface is measurement information exchanged using an interface between the processing device 103 and other devices, such as measurement information specified in interfaces such as E2, A1-EI, and Y1. When the measurement information requested by the request message is the same as measurement information specified in another interface, the parameter names associated with the measurement information can be reused rather than newly specifying the measurement information in the A1 interface, thereby avoiding the complication of the same measurement information being given different names. The parameter names of such measurement information specified in the existing O-RAN interface can be written as "OR.DataName". "OR" is a prefix indicating that the measurement information is measurement information specified in another O-RAN interface, and "DataName" is the parameter name specified in the other interface. Furthermore, the measurement information newly specified in the A1 interface is measurement information not specified in other O-RAN interfaces, and new parameter names can be specified in the A1 interface. For example, newly generated information by the processing device 103 analyzing measurement information or newly generated information by processing measurement information can be linked to new parameter names and provided to the control device 102. The parameter names of such measurement information newly specified in the A1 interface can be written as "IS.DataName". "IS" is a prefix indicating that it is newly defined, and "DataName" is the parameter name associated with the measurement information.Furthermore, measurement information that is not specified as an O-RAN interface but is independently specified by a product development vendor may be linked to a free parameter name defined by the product vendor. For example, in the processing device 103, information generated by a process unique to the product vendor may be linked to a parameter name uniquely generated by the product vendor and provided to the control device 102. The parameter name of such measurement information uniquely specified by the product vendor may be written as "VS.DataName." "VS" is a prefix indicating that it has been uniquely specified by the product vendor, and "DataName" is the parameter name associated with the measurement information. Note that the type of measurement information may include types other than those described above. Furthermore, one request message may include one or more pieces of information identifying the type of measurement information. The information identifying the type of measurement information may be information identifying any one of a parameter name, a category, and a prefix, or a combination of these.

[0022] The request message may include, for example, information (which may be referred to as a Scope Identifier) ​​that specifies the unit (or range) of the measurement information. The information that specifies the unit of the measurement information may include an identifier that uniquely identifies a cell (cell ID), an identifier that uniquely identifies a UE (UE ID), a condition for identifying multiple UEs that satisfy a predetermined condition (UE group), an identifier that uniquely identifies a slice (slice ID), and an identifier that uniquely identifies a QoS flow (QoS ID). For example, the slice ID may be S-NSSAI (Single-Network Slice Selection Assistance Information). Furthermore, the QoS ID may be a Quality flow Indicator (QFI) or a 5G QoS Indicator (5QI). The information that specifies the unit of the measurement information may be other than the above. A single request message may include multiple pieces of information specifying the unit of measurement information. For example, the control device 102 may request the same type of measurement information for the same period in multiple different units in order to generate a learned model corresponding to each unit of measurement information or a learned model combining these units, depending on the intended use. Two or more pieces of information specifying the unit of measurement information may also be combined. For example, a combination of a UE ID and a slice ID provides only measurement information for a slice corresponding to a specified S-NSSAI in a UE corresponding to the specified UE ID. In this case, it is possible to generate a more accurate learned model based on the traffic characteristics of the specified slice and to set an A1 policy with finer granularity. Furthermore, a combination of a UE ID and a QoS SID, for example, allows the use of measurement information with even finer granularity than a slice. On the other hand, by combining a cell ID with a slice ID or a QoS SID, the control device 102 can use the average or total for each slice or QoS flow for multiple UEs. Using more measurement information makes it possible, for example, to generate a learned model or A1 policy that is less dependent on the UE's location.Furthermore, when generating an A1 policy or a learned model for UEs that satisfy a specific condition, the control device 102 may request measurement information of a UE group that is a group of UEs that satisfy this condition. Note that the UE group may be combined with a slice ID or a QoS ID.

[0023] In this way, the control device 102 can use the measurement information held by the processing device 103 for control purposes to improve the accuracy of RAN control by receiving the information in a procedure or unit appropriate for the purpose. Also, by specifying the type, unit, procedure, etc. of the measurement information to be requested, it is possible to reduce the amount of traffic transferred using the A1 interface and to level out the utilization rate of the A1 interface.

[0024] (RAN Control Operation Using Measurement Information) The control device 102 performs processing for configuring and operating the RAN using measurement information provided from the processing device 103. FIG. 3 shows an example of the operation of the control device 102 in RAN control using measurement information. In this embodiment, as an example, the control device 102 generates a trained model for predicting the future throughput of each UE using measurement information related to the communication quality of each UE. Examples of measurement information related to communication quality may include received signal strength indicator (RSSI), signal-to-interference-and-noise ratio (SINR), throughput, and delay time. For example, assume that the control device 102 sets the throughput of each UE in a certain slice to be 10 Mbps or more as an A1 policy and provides this A1 policy to the processing device 103 (S301). At this time, the control device 102 requests the processing device 103 to provide time-series measurement information related to the communication quality of each UE in this slice (S302). For example, in the request message, the control device 102 specifies the S-NSSAI corresponding to the slice as the slice ID and the ID corresponding to the UE using the slice as the UE ID. Furthermore, in the request message, the control device 102 may specify measurement information related to communication quality, such as throughput, RSSI, and SINR, as the type of measurement information. Since the control device 102 uses this measurement information as input for machine learning, the control device 102 may specify a procedure for providing the measurement information, such as providing the information all at once after the measurement is completed, and may also specify a method for processing the measurement information. The control device 102 then performs machine learning using the time-series measurement information provided from the processing device 103 as input, and generates a trained model (S303). For example, the control device 102 acquires from the processing device 103 time-series measurement information (first measurement information) of the communication quality of each UE over the past 20 seconds from a certain reference time, and time-series measurement information (second measurement information) of the throughput of each UE over the past 10 seconds from the reference time. The control device 102 performs machine learning using training data in which the first measurement information is input data and the second measurement information is correct data, thereby generating a trained model. Note that the training data used by the control device 102 and the trained model generated are not limited to the measurement information related to communication quality, and machine learning may be performed in combination with other available information.The generated trained model is provided to the processing device 103 via the A1 interface (S304).

[0025] The processing device 103 controls the E2 node using the trained model provided from the control device 102. FIG. 4 shows an example of the operation of the processing device 103 in this embodiment. First, the processing device 103 receives the trained model from the control device 102 (S401). For example, the trained model may be provided as a program executable by a computer included in the processing device 103. The processing device 103 collects measurement information to be input to the trained model (S402). The measurement information to be input to the trained model may be acquired, for example, from the E2 node via the E2 interface. Alternatively, measurement information stored in a storage medium included in the processing device 103 may be used. Then, the processing device 103 performs future prediction of the communication quality of the RAN by inference using the trained model (S403). In this embodiment, as an example, the processing device 103 infers the throughput of each UE over a 10-second period from the current time using time-series measurement information covering the past 20 seconds from the current time, for the same type and unit of measurement information as the measurement information used by the control device 102 as training data. The processing device 103 controls the E2 node 104 based on the results of the inference analysis (S404). In this embodiment, if the inference results in a UE whose throughput falls below the setting in the A1 policy, the processing device 103 instructs the E2 node 104 to hand over the UE to another adjacent cell. In this way, the control device 102 generates a trained model using communication quality for each UE. The processing device 103 can use this trained model to predict communication quality for each UE and perform fine-grained RAN control based on this prediction. Note that the control device 102 may periodically receive measurement information and update the trained model. By updating the trained model, the processing device 103 can maintain appropriate predictions of communication quality in response to changes in the communication quality of the RAN. Furthermore, when measurement information from multiple processing devices 103 is available, the control device 102 can perform machine learning using more measurement information. By using measurement information received from a wide range of areas, a general-purpose trained model that is less affected by a specific area can be generated.Even when using only the measurement information of one processing device, the control device 102 is generally implemented on a computer with higher performance than the processing device 103. Therefore, having the control device 102 analyze the measurement information has the advantage of being able to process large amounts of measurement information at high speed. On the other hand, when the control device 102 receives measurement information from each UE for each slice or QoS flow, it may generate multiple trained models for each unit. Using trained models generated for each slice or QoS flow makes it possible to perform control based on highly accurate predictions that take into account differences in communication traffic characteristics, etc.

[0026] FIG. 5 illustrates another example of the operation of the control device 102 in RAN control using measurement information. In this embodiment, the control device 102 may change or add the A1 policy upon receiving measurement information related to communication quality for each UE from the processing device 103. For example, assume that 30% of the resource blocks in a certain cell are allocated to a certain slice. A resource block is a unit of available radio resources allocated to each UE, slice, etc. in a cell. For example, assume that the control device 102 sets the A1 policy as a performance target for this slice, ie, that the RAN throughput for each UE is 10 Mbps or more, and provides this A1 policy to the processing device 103 (S501). At this time, the control device 102 requests the processing device 103 to provide measurement information related to the number of resource blocks used by each UE in this slice and the throughput (S502). For example, in the request message, the control device 102 specifies the S-NSSAI corresponding to the slice as the slice ID and the ID corresponding to the UE using the slice as the UE ID. Furthermore, in the request message, the control device 102 may specify, as types of measurement information, the number of resource blocks in use, throughput, and measurement information related to communication quality such as RSSI and SINR. Since the control device 102 uses this measurement information to verify the A1 policy, the control device 102 specifies periodic provision as the measurement information providing procedure. The control device 102 then verifies the A1 policy using the measurement information provided from the processing device 103 and generates the A1 policy as necessary (S503). For example, the control device 102 may calculate the resource block usage rate of this slice in this cell from the total number of resource blocks in use by each UE provided with the information. Furthermore, the control device 102 may predict the throughput of each UE from the throughput measurement information for each UE, for example, by analysis using a predetermined algorithm or inference using a trained model.Here, if the total resource block usage rate of this slice exceeds a predetermined threshold (for example, 27% of the resource blocks in the entire cell are used) and the predicted throughput of a specific UE does not satisfy the A1 policy (for example, less than 10 Mbps), the control device 102 generates a new A1 policy for load balancing and provides it to the processing device 103 (S504). For example, the new A1 policy specifies this cell and this slice, and sets the resource block usage rate to 24% or less.

[0027] FIG. 6 illustrates another example of the operation of the processing device 103 according to this embodiment. The processing device 103 receives the A1 policy from the control device 102 in S504 (S601). The processing device 103 controls the E2 node 104 using the received A1 policy. For example, the processing device 103 collects measurement information required to execute the A1 policy from the E2 node 104 (S602). For example, the processing device 103 collects measurement information for identifying resource block usage rates in cells and slices set in the A1 policy. The measurement information for identifying resource block usage rates may be the number of resource blocks used in the cell. The processing device 104 also collects information for identifying the throughput of each UE. The processing device 103 analyzes the collected measurement information and determines whether the A1 policy is satisfied (S603). This analysis may be performed, for example, by inference using a trained model that inputs the collected measurement information or by determination using a predetermined algorithm. The processing device 103 controls the E2 node based on the analysis results (S604). For example, the processing device 103 determines that some UEs should be handed over because the analysis results show that the resource blocks used by the cells and slices set in the A1 policy account for 27% of the entire cell, exceeding the threshold value set in the A1 policy (24% of the entire cell). Furthermore, the processing device 103 selects UEs whose predicted throughput does not satisfy the A1 policy (less than 10 Mbps) as targets for handover. Based on these determination results, the processing device 103 instructs the E2 node 104 to hand over the selected UEs to adjacent cells (S604). In this way, the control device 102 receives fine-grained measurement information from the processing device 103, verifies the A1 policy using this measurement information, and provides the generated A1 policy to the processing device 103. This enables accurate configuration and operation of the RAN.

[0028] (Operation when Processing Measurement Information) When the control device 102 generates a trained model using measurement information received from the processing device 103 and the processing device 103 performs inference using this trained model, the formats of the measurement information used by the control device 102 and the processing device 103 must match. In particular, if the control device 102 performs preprocessing such as normalizing the measurement information or interpolating missing data when generating the trained model, the processing device 103 must also perform inference using measurement information that has undergone similar preprocessing to obtain a correct inference result. In this embodiment, the control device 102 provides a function for notifying the processing device 103 of a method for processing the measurement information. Note that the processing device 103 may specify a method for processing the measurement information to the control device 102. For example, if the measurement information specified in the request message is to be processed, the control device 102 may notify the processing method within the request message. Furthermore, the method by which the control device 102 notifies the processing method of the measurement information is not limited to the request message, and a sequence for notifying the processing method may be newly provided in the A1 interface. By providing an independent sequence for the request message, the control device 102 can specify a wide range of measurement information to be processed. For example, the control device 102 can request the processing device 103 to always perform a specified processing process on a specific type of measurement information. Furthermore, when the control device 102 provides a trained model to the processing device 103, a method for preprocessing the measurement information to be input to the trained model may be provided. In this case, the control device 102 can select a different processing method to preprocess the measurement information each time a trained model is generated, and notify the processing device 103 of the selected processing method each time. The processing device 103 can process the measurement information according to the processing method specified in the request message or a sequence for notifying the processing method, and provide the measurement information to the control device 102. Note that the processing device 103 may not perform processing when providing the measurement information to the control device 102, and the control device 102 may process the measurement information it receives. In this case, when performing inference using the trained model provided by the control device 102, the processing device 103 processes the input data of the trained model using the processing method identified by the control device 102.In addition, the processing of measurement information is not limited to preprocessing for generating a trained model, but may also be other processing such as compression to reduce the data size of the measurement information or encryption to exchange information securely.

[0029] (Specifying Processing Method) As an example, the control device 102 may specify the processing method using a mathematical expression. The mathematical expression may include, for example, an expression using a pair of a condition and an operation. For example, if the measurement information to be processed is the throughput for each UE, the processing method may be (Condition 1) if the throughput value exceeds 10 Mbps, (Operation 1) replace it with 10 Mbps, and (Condition 2) normalize the entire measurement information between 0 and 1. In this case, the request message may notify the processing device 103 of each pair of Condition 1 and Operation 1, and Condition 2 and Operation 2. The processing device 103 determines whether each of the requested measurement information satisfies the specified condition, and if the condition is satisfied, performs the operation corresponding to the condition and provides the information to the control device 102. Note that the processing method may be deleting duplicates or filling in missing values ​​to generate training data to be used in machine learning. In the case of duplicate deletion, for example, the condition is that the identification number identifying each measurement information is the same as the identification number of other measurement information, and the operation is to delete that measurement information. In the case of missing value complementation, for example, the condition is that there is a missing piece of measurement information, and the operation is to delete this data or replace it with an average or representative value of all the measurement values. Note that the mathematical expression is not limited to conditions and operations, and can include any items or expressions that specify the conditions or operations to be processed.

[0030] The control device 102 may specify a processing method by specifying a function corresponding to the processing method. For example, the control device 102 may share with the processing device 103 in advance the functions used to process the measurement information and the identifiers identifying each function. In this case, the control device 102 may specify the processing method by notifying the processing device 103 of the identifiers corresponding to the functions used in the processing. By sharing functions between the control device 102 and the processing device 103 in advance, it is possible to reduce the information required to specify each processing method, even when performing complex processing. As an example, the control device 102 requests the processing device 103 to register any function containing a mathematical formula or the like along with its identifier. By checking the processing device 103's response to this request, information about functions that can be specified may be shared between the devices. Note that if an argument is required to execute processing using a function, the control device 102 may notify the argument along with the function identification.

[0031] The control device 102 may specify an xApp corresponding to the processing method in the request message, thereby identifying the processing method. The xApp is, for example, an application that analyzes and controls various types of information within the framework of the processing device 103. For example, the control device 102 may share with the processing device 103 in advance the content of the processing performed by the xApp and an identifier identifying each xApp. The control device 102 may specify the processing method by notifying the processing device 103 of an identifier corresponding to the xApp used for the processing. As an example, the control device 102 may request the processing device 103 to install the xApp. By confirming a response to this request indicating completion of installation by the processing device 103, information about the xApp that can be specified between the devices may be shared in advance. Note that if input parameters are required to execute the xApp, the control device 102 may notify the input parameters along with the identification of the xApp. Even when complex processing is performed on the measurement information, the processing method can be easily notified by identifying the xApp.

[0032] (Hardware Configuration) Next, an example configuration of the control device 102 and processing device 103 described above will be described. FIG. 7 is a diagram showing the hardware configuration of the control device 102 and processing device 103. In one example, the control device 102 and processing device 103 include a processor 701, a ROM 702, a RAM 703, a storage device 704, and a communication circuit 705. The processor 701 is a computer including one or more processing circuits, such as a general-purpose CPU (Central Processing Unit) or an ASIC (Application Specific Integrated Circuit). The processor 701 reads and executes programs stored in the ROM 702 or storage device 704 to perform overall processing of the device and each of the above-mentioned processes. The ROM 702 is a read-only memory that stores programs related to the processes executed by the control device 102 and processing device 103, various parameters, and other information. The RAM 703 functions as a workspace when the processor 701 executes a program, and is a random access memory that stores temporary information. The storage device 704 is, for example, a removable external storage device. The communication circuit 705 includes, for example, a circuit for communicating with other devices. As an example, the communication circuit 705 can function as an A1 interface, an O1 interface, an E2 interface, etc. Although one communication circuit 705 is illustrated in FIG. 7, each device can have multiple communication circuits.

[0033] (Functional Configuration) FIG. 8 is a diagram showing an example of the functional configuration of the control device 102. The control device 102 includes, as its functions, a first acquisition unit 801, a request unit 802, a second acquisition unit 803, a processing specification unit 804, a processing information sharing unit 805, a processing unit 806, a model generation unit 807, a model provision unit 808, and a RAN control unit 809. FIG. 8 shows the functional configuration of the control device 102 of this embodiment, and, for example, omits the general configuration of the control device. Note that these functional units can be realized, for example, by the processor 701 executing a program stored in the ROM 702 or the storage device 704 and controlling the communication circuit 705 as necessary. However, this is not limited to this, and, for example, dedicated hardware for realizing each function may be provided.

[0034] The first acquisition unit 801 acquires measurement information from the E2 node 104. For example, the first acquisition unit 801 may acquire measurement information such as RAN communication quality, which can be used by the control device 102 to control the RAN, using the O1 interface. The request unit 802 transmits a request message to the processing device 103 to request the provision of measurement information. The request message may be transmitted using the A1 interface. The request unit 802 may use the request message to notify the processing device 103 of information specifying the measurement period, provision period, provision procedure, type, unit, etc. of the measurement information. The request unit 802 may also use the request message to notify the processing device 103 of information specifying a method for processing the measurement information. The second acquisition unit 803 receives a response message and measurement information from the processing device 103 in response to the request message. If it is possible to provide measurement information according to the request, the processing device 103 may provide the measurement information together with the response message. Note that the measurement information may be provided in a message separate from the response message. The measurement information obtained by the control device 102 from the E2 node 104 via the first acquisition unit 801 and the measurement information obtained from the processing device 103 via the second acquisition unit 803 may have different granularity, with the former being statistical information on a cell-by-cell basis and the latter being individual information on a UE-by-UE basis. The second acquisition unit 803 may also be provided with information such as analysis results independently generated by the processing device 103 based on the information acquired from the E2 node. The processing specification unit 804 specifies processing to be used in the model generation unit 807 or the processing device 103. For example, the processing specification unit 804 specifies processing for performing preprocessing when the model generation unit 807 generates a trained model through machine learning. This processing may also be applied when the processing device 103 performs inference using the trained model. As an example, preprocessing specified by the processing specifying unit 804 may be performed on the measurement information acquired by the second acquisition unit 803, thereby acquiring information to be used as input for machine learning to be executed by the model generation unit 807. The processing specifying unit 804 specifies one or more processes necessary for the model generation unit 807 to execute machine learning, from processing processes such as normalization, duplicate deletion, and missing value completion.The processing operation specifying unit 804 may select a processing operation to be executed from functions, xApps, etc. shared in advance with the processing device 103. The processing information sharing unit 805 notifies the processing device 103 of the processing operation specified by the processing operation specifying unit 804. The processing information sharing unit 805 may receive a response from the processing device 103, including a rejection of the notification specifying the processing operation or a request to change the processing operation. The processing information sharing unit 805 may also receive a notification specifying the processing operation from the processing device 103. Information specifying the processing operation may be notified to the processing device 103 using a request message. Information specifying the processing operation may also be notified to the processing device 103 using a method other than a request message. For example, the processing information sharing unit 805 may request the processing device 103 to register functions, xApps, etc. to be used for the processing operation, and, if necessary, may notify the processing device 103 of the measurement information to be processed, arguments to be used for the function, input parameters to be used for the xApp, etc. The processing unit 806 executes the processing operation on the measurement information. The processing unit 806 may perform processing using the processing specified by the processing processing specification unit 804. The processed measurement information may be provided to the model generation unit 807. The model generation unit 807 generates a trained model through machine learning. The model generation unit 807 generates a trained model for, for example, predicting future communication quality of the RAN. The model generation unit 807 may use the processed measurement information provided by the processing unit 806 as input for machine learning, or, if processed measurement information is provided by the processing device 103, may use this as input for machine learning. The model providing unit 808 provides the trained model generated by the model generation unit 807 to the processing device 103. The RAN control unit 809 controls the RAN. As an example, the RAN control unit 809 generates an A1 policy for controlling the RAN using the measurement information acquired by the first acquisition unit 801 and the second acquisition unit 803, and provides the A1 policy to the processing device 103. The RAN control unit 809 can use the measurement information acquired by the second acquisition unit 803 to determine, for example, the achievement status of a performance target for each UE set as an A1 policy, and generate a new A1 policy based on the achievement status. The generated A1 policy can be provided to the processing device 103.

[0035] FIG. 9 is a diagram showing an example of the functional configuration of the processing device 103. The processing device 103 includes, as its functions, an information acquisition unit 901, a storage unit 902, a receiving unit 903, a transmitting unit 904, a processing information sharing unit 905, a processing unit 906, a model acquisition unit 907, and a RAN control unit 908, for example. FIG. 9 shows the functional configuration of the processing device 103 of this embodiment, and, for example, omits the general configuration of a processing device. Note that these functional units can be realized, for example, by the processor 701 executing a program stored in the ROM 702 or the storage device 704 and controlling the communication circuit 705 as necessary. However, this is not limited to this, and, for example, dedicated hardware for realizing each function may be provided.

[0036] The information acquisition unit 901 collects measurement information related to the RAN from the E2 node 104. Examples of information that can be collected include measurement information related to communication quality (RSSI, SINR, throughput, delay time, etc.), control function information and control history information of the E2 node, and the like, and can be specified on a cell-by-cell, slice-by-slice, UE-by-UE basis. The information acquisition unit 901 may perform analysis using the collected information to generate statistical information, etc. For example, the information acquisition unit 901 may use the measurement information collected from the E2 node to generate statistical values ​​such as the average, maximum, minimum, and variance of the throughput and delay over time for each UE, and the average, maximum, minimum, and variance of the throughput in each cell. The information acquisition unit 901 may also acquire the received power and interference power in each cell, the resource usage rate of each cell, and analyze the correlation between each piece of measurement information and each piece of statistical information. Furthermore, the information acquisition unit 901 may acquire and analyze information such as the operating status of each node (including the E2 node) constituting the RAN. The measurement information collected by the information acquisition unit 901 and the analysis results generated may be recorded in the storage unit 902. The storage unit 902 stores the measurement information and analysis results collected from the E2 node 104, the A1 policy, the learned model, the function for processing the measurement information, the xApp, and the like received from the control device 102. The receiving unit 903 receives a request message requesting the provision of measurement information from the control device 102. The transmitting unit 904 determines whether to accept the request included in the request message and transmits a response message to the control device 102. For example, if the transmitting unit 904 accepts the request, it transmits the measurement information and analysis results recorded in the storage unit 902 to the control device 102 in accordance with the request. The measurement information and analysis results transmitted by the transmitting unit 904 may be processed by the processing unit 906. If the transmitting unit 904 does not accept the request, it transmits a response message rejecting the request to the control device 102. The response message rejecting the request may include a reason for the rejection. The processing information sharing unit 905 can receive information on processing to be used by the processing unit 906 from the control device 102. When executing the processing notified by the control device 102, the processing information sharing unit 905 may transmit a message to the control device 102 indicating that the processing is accepted.Furthermore, when the processing information sharing unit 905 refuses to execute the processing notified by the control device 102, it may transmit a message indicating the refusal to the control device 102. Furthermore, the processing information sharing unit 905 may identify a processing process that the device itself can execute and transmit a message to the control device 102 requesting that the processing process be changed to the identified processing process. Note that the processing information sharing unit 905 may identify and notify the processing process before the control device 102. The processing processing unit 906 executes processing of the measurement information, etc. stored in the storage unit 902. The measurement information, etc. processed by the processing processing unit 906 may be used for analysis by the RAN control unit 908 and may be provided to the control device 102. The measurement information provided to the control device 102 may be used as input for machine learning in the control device 102. The model acquisition unit 907 acquires a trained model from the control device 102. The RAN control unit 908 performs analysis on the RAN using the measurement information and controls the RAN using the obtained analysis results. The RAN control unit 908 may perform future predictions of the communication quality of the RAN, for example, using a trained model provided from the control device 102. The RAN control unit 908 may acquire the trained model from a source other than the control device 102. The RAN control unit 908 may also analyze the RAN using a method other than machine learning. For example, future predictions of the communication quality of the RAN may be performed by applying a predetermined analysis algorithm to measurement information acquired from the E2 node 104. The RAN control unit 908 also configures and operates the RAN by controlling the E2 node 104. The RAN control unit 908 may control the E2 node 104 in accordance with the A1 policy received from the control device 102. The RAN control unit 908 may also generate parameters for controlling the E2 node 104 using the communication quality of the RAN predicted by the analysis.

[0037] In the above description, an embodiment has been described in which the processing device 103 collects information measured by E2 nodes and the like in the RAN and provides the information to the control device 102. However, the information provided to the control device 102 is not limited to such measurement information. The measurement information in this embodiment may also include, for example, information obtained by processing or statistical processing of information measured by devices in the RAN. The measurement information may also broadly include information generated by analysis or the like in each device other than the above, and information such as the resource usage status and load of each device.

[0038] As described above, according to the present embodiment, the control device 102 requests the provision of measurement information held by the processing device 103, and the processing device 103 provides the measurement information to the control device 102 in accordance with the request. This measurement information can be used to update the A1 policy in the control device 102 or to generate a trained model through machine learning. The control device 102 provides the processing device 103 with an A1 policy updated based on the measurement information or a trained model generated using fine-grained measurement information, and the processing device 103 uses these to control the E2 node 104, thereby improving the accuracy of control related to RAN configuration and operation. Furthermore, according to the present embodiment, when the processing device 103 performs inference using the trained model generated by the control device 102 to control the RAN, information specifying the processing required in each device is shared. This makes it possible to use machine learning in RAN control, using measurement information collected from devices constituting the RAN, such as the E2 node, as input. This will enable us to contribute to Goal 9 of the United Nations' Sustainable Development Goals (SDGs), which is to "build resilient infrastructure, promote sustainable industrialization and foster innovation."

[0039] The invention is not limited to the above-described embodiment, and various modifications and variations are possible within the scope of the gist of the invention.

[0040] This application claims priority based on Japanese Patent Application No. 2023-161692, filed September 25, 2023, the entire contents of which are incorporated herein by reference.

Claims

1. A control device that functions as a Non-Real-Time RAN Intelligent Controller (non-real-time RIC) in an Open-Radio Access Network (O-RAN), comprising: an identification means for identifying a processing process to be performed on second information acquired by a near-real-time RIC in order to acquire first information to be used as an input for machine learning and a trained model obtained by the machine learning; a generation means for generating the trained model by machine learning using as input the first information acquired by the identified processing process; a communication means for communicating third information indicating the identified processing process with the near-real-time RIC; and a provision means for providing the trained model to the near-real-time RIC.

2. The control device according to claim 1, wherein the third information includes information on a mathematical formula used in the processing.

3. The control device according to claim 1, wherein the third information includes information for identifying a function used in the processing, the function being shared in advance between the control device and the near real-time RIC.

4. The control device according to claim 1, wherein the third information includes information for identifying an application used in the processing.

5. A control device according to any one of claims 1 to 4, further comprising: an acquisition means for acquiring the second information from the quasi-real-time RIC; and a processing means for performing processing on the second information using the processing process.

6. A control device according to any one of claims 1 to 4, further comprising an acquisition means for acquiring the first information acquired by the quasi-real-time RIC executing the processing on the second information.

7. A processing device that functions as a Near-Real-Time RAN Intelligent Controller (Near-Real-Time RIC) in an Open-Radio Access Network (O-RAN), comprising: an acquisition means for acquiring second information including at least one of information collected from an E2 node or information generated in the processing device; a communication means for communicating with a non-real-time RIC third information indicating a processing process to be performed on the second information in order to acquire first information to be used as an input for machine learning and a trained model obtained by the machine learning; a processing means for performing the processing process on the second information to acquire first information; an acquisition means for acquiring a trained model from the non-real-time RIC; and a control means for controlling the RAN using an inference result obtained by inputting the first information into the trained model.

8. The processing device according to claim 7, wherein the third information includes information on a mathematical formula used in the processing.

9. The processing device according to claim 7, wherein the third information includes information for identifying a function used in the processing, the function being shared in advance between the processing device and the non-real-time RIC.

10. The processing device according to claim 7, wherein the third information includes information for identifying an application used for the processing.

11. The processing device according to any one of claims 7 to 10, further comprising: providing means for providing the second information to the non-real-time RIC.

12. The processing device according to any one of claims 7 to 10, further comprising: providing means for providing the first information to the non-real-time RIC.

13. A control method executed by a control device functioning as a Non-Real-Time RAN Intelligent Controller (non-real-time RIC) in an Open-Radio Access Network (O-RAN), comprising: identifying a processing process to be executed on second information acquired by a near-real-time RIC in order to acquire first information to be used as an input for machine learning and a trained model obtained by the machine learning; generating the trained model by machine learning using as input the first information acquired by the identified processing process; communicating third information indicating the identified processing process with the near-real-time RIC; and providing the trained model to the near-real-time RIC.

14. A processing method executed by a processing device functioning as a Near-Real-Time RAN Intelligent Controller (Near-Real-Time RIC) in an Open-Radio Access Network (O-RAN), comprising: acquiring second information including at least one of information collected from an E2 node or information generated in the processing device; communicating third information indicating a processing to be executed on the second information with a non-real-time RIC to acquire first information to be used as an input for machine learning and a trained model obtained by the machine learning; executing the processing on the second information to acquire first information; acquiring a trained model from the non-real-time RIC; and controlling a RAN using an inference result obtained by inputting the first information into the trained model.

15. A program for causing a computer included in a control device functioning as a Non-Real-Time RAN Intelligent Controller (non-real-time RIC) in an Open-Radio Access Network (O-RAN) to identify a processing process to be performed on second information acquired by a near-real-time RIC in order to obtain first information to be used as input for machine learning and a trained model obtained by the machine learning, generate the trained model by machine learning using as input the first information acquired by the identified processing process, communicate third information indicating the identified processing process with the near-real-time RIC, and provide the trained model to the near-real-time RIC.

16. A program for causing a computer provided by a processing device functioning as a Near-Real-Time RAN Intelligent Controller (Near-Real-Time RIC) in an Open-Radio Access Network (O-RAN) to acquire second information including at least one of information collected from an E2 node or information generated in the processing device, communicate third information indicating a processing to be performed on the second information with a non-real-time RIC to acquire first information to be used as input for machine learning and a trained model obtained by the machine learning, perform the processing on the second information to acquire first information, acquire a trained model from the non-real-time RIC, and control the RAN using the inference result obtained by inputting the first information into the trained model.

Images