Base station function allocation determination device, method, and program

Abstract

To provide a base station function arrangement determination device for determining a base station function arrangement of each cell of a wireless access network by means of a prediction model while taking total optimality into account.SOLUTION: A prediction model M for predicting resource usage and required quality achievement rate in a control term t+1 is previously constructed for each cell based on a base station function arrangement, required throughput and resource usage in a control term (t). A required throughput, resource usage and all base station function arrangements at any control time are inputted to a corresponding prediction model for each model, thereby predicting the resource usage and the required quality achievement rate in each base station arrangement. A combination of base station function arrangements between cells is optimized based on prediction results of the resource usage and the required quality achievement rate in each base station arrangement of each cell, and a base station function arrangement of the optimized combination is adopted.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a base station function allocation determination device, method, and program, and in particular to a base station function allocation determination device, method, and program that determines base station function allocation for each cell of a radio access network using a prediction model while taking into consideration overall optimization. [Background technology]

[0002] In 5G, base station functions (vDU, vCU) virtualized by RAN slicing technology can be adaptively allocated to antenna sites and accommodating stations, as disclosed in Patent Document 1. The O-RAN specifications formulate a method for optimizing slices in response to traffic fluctuations (Non-Patent Document 1), stipulate the collection of various resource information to achieve this (Non-Patent Document 2), and propose a method for changing the allocation of base station functions using this information (Patent Document 2).

[0003] The amount of various network resources consumed varies depending on the placement of base station functions, and some resources are shared between cells, so it is important to consider overall optimization when deciding on the placement of base station functions.

[0004] As a method for determining the allocation of base station functions, methods using reinforcement learning are disclosed in Patent Documents 3 to 6. In these methods, when learning the allocation of base station functions, trial and error is performed on the allocation of base station functions for RAN, and learning data is collected to update the learning model. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2020-136787 [Patent Document 2] Japanese Patent Publication No. 2022-154666 [Patent Document 3] Japanese Patent Publication No. 2022-165659 [Patent Document 4] Japanese Patent Publication No. 2022-165660 [Patent Document 5] Japanese Patent Publication No. 2022-165661 [Patent Document 6] Japanese Patent Publication No. 2022-171145 [Non-patent literature]

[0006] [Non-Patent Document 1] O-RAN WG1,"Slicing architecture",v03.00,3.2.3 Use Case 3: NSSI Resource Allocation Optimization [Non-patent document 2] 3GPP (registered trademark), "TS 28.552", V17.1.0, 2020-12 [Non-patent document 3] Yu Tsukamoto, Haruhisa Hirayama, Seung Il Moon, Hiroyuki Shinbo, "Adaptive Function Placement with Distributed Deep Reinforcement Learning in RAN Slicing," The 2022 IEEE 95th Vehicular Technology Conference VTC2022-Spring) [Non-patent document 4] Yu Tsukamoto, Haruhisa Hirayama, Seung Il Moon, Shinobu Nanba, and Hiroyuki Shinbo, "Feedback control for adaptive function placement in uncertain traffic changes on an advanced 5G system," 2021 IEEE 18th Annual Consumer Communications and Networking Conference (CCNC 2021) Summary of the Invention [Problem to be solved by the invention]

[0007] However, trial and error in the placement of base station functions in a production RAN environment carries risks such as degrading communication quality due to the selection of an inappropriate placement of base station functions. Furthermore, trial and error must be continued until the accuracy of the learning model is improved, which takes time. Furthermore, while using a simulation environment could be considered to reduce risks in the production environment, modeling errors that arise between the production environment and the simulation environment degrade the accuracy of the learning model.

[0008] The object of the present invention is to solve the above technical problems and to provide a base station function allocation determination device, method, and program that determines the base station function allocation by optimizing the combination of base station function allocations using a prediction model that has learned, through supervised learning, the relationship between the base station function allocation and the usage of various resources (transmission paths, computers, etc.) and the rate of achievement of required quality (throughput, communication delay, etc.). [Means for solving the problem]

[0009] In order to achieve the above object, the present invention provides a base station function allocation determination device that determines the allocation of base station functions in each cell of a radio access network, comprising: a prediction model that predicts resource usage and a required quality achievement rate in control period t+1 based on the base station function allocation, required throughput, and resource usage in control period t for each cell; means for inputting the required throughput and resource usage and all base station function allocations at an arbitrary control time for each cell into the corresponding prediction model to predict the resource usage and required quality achievement rate for each base station allocation; and means for optimizing combinations of base station function allocations between cells based on the prediction results of the resource usage and required quality achievement rate for each base station allocation for each cell, and adopting the optimized combination of base station function allocations. [Effects of the Invention]

[0010] According to the present invention, it is possible to manually control the RAN or to pre-train a learning model using actual data from feedback control with relatively low risk, thereby eliminating the risk and time associated with trial and error, which are issues with methods using reinforcement learning, and eliminating the need for a simulation environment. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a block diagram showing an example of the configuration of a radio access network (RAN) according to an embodiment of the present invention. [Figure 2] FIG. 1 is an explanatory diagram showing an example of a base station function arrangement in a RAN. [Figure 3] FIG. 2 is a block diagram showing a configuration example of a base station function arrangement control unit. [Figure 4] 1 is a flowchart showing an outline of a base station function allocation control method. [Figure 5] FIG. 10 is a sequence diagram showing an example of detailed procedures of a base station function allocation control method. [Figure 6] FIG. 1 is a diagram illustrating a method for constructing a learning model. [Figure 7] 10 is a flowchart showing an example of an optimization procedure using a hill climb method. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Fig. 1 is a functional block diagram showing an example of the configuration of a radio access network (RAN) 1 according to an embodiment of the present invention. Here, the description will be given taking as an example a RAN to which RAN slicing technology and O-RAN specifications are applied, but the application of the O-RAN specifications is not limited to this.

[0013] RAN 1 mainly comprises SMO Functions (Service and Management Orchestration function unit) 11, Non-RT RIC (Non-real-time RAN intelligent controller) 12, and base station function group 2, and Non-RT RIC includes base station function placement control unit 20.

[0014] In this embodiment, the base station function allocation control unit 20 is realized using the Non-RT RIC 12. The SMO function unit 11 and the Non-RT RIC 12 are realized using the SMO framework 10. Of the base station functions, the DU, which is responsible for processing radio signals, and the CU, which is responsible for encryption and quality control, are virtualized, and the base station function group 2 is composed of an RU (Radio Unit), a vCU (Virtual Central Unit), and a vDU (Virtual Distributed Unit), as exemplified in Fig. 2. The vDU and vCU are adaptively allocated to either an antenna site or an accommodating station.

[0015] 2 is a diagram showing an example of base station function allocation in RAN1, in which a total of five sets of vCUs and vDUs are connected to RUs that form a single cell. There are three base station function allocation patterns depending on how the vDUs and vCUs are allocated to the accommodating stations and antenna sites, and the network characteristics differ depending on the allocation pattern. In order to meet the quality requirements of various services (emphasis on stability, emphasis on low latency, etc.), in this embodiment, the base station function allocation is appropriately determined for each cell and service.

[0016] The first example, "O-RAN slice subnet #1," is a C-RAN (Centralized-RAN) arrangement in which both the vCU and vDU are located in the accommodating station. While the C-RAN arrangement allows for inter-cell cooperation, it is characterized by the large bandwidth required for the transmission path.

[0017] The second example, "O-RAN slice subnet #2," and the third example, "O-RAN slice subnet #3," are S-RAN (Split-RAN) deployments, in which a vDU is deployed at the antenna site and a vCU is deployed at the accommodating station. S-RAN deployments do not allow inter-cell cooperation, but have the advantage of requiring a small bandwidth for the transmission path.

[0018] The fourth example, "O-RAN slice subnet #4," and the fifth example, "O-RAN slice subnet #5," are D-RAN (Distributed RAN) deployments, in which both vCU and vDU are deployed at the antenna site. D-RAN deployments do not allow inter-cell cooperation, but are characterized by low communication latency.

[0019] As shown in Figure 2, by connecting multiple sets of vCU and vDU to an RU that forms a single cell in RAN1, it is possible to realize slices of base station function allocation with multiple different characteristics for a single cell. This makes it possible to provide wireless communication services with multiple different characteristics in a single cell in RAN1, and to adapt to the diverse communication qualities required by a wide variety of services.

[0020] Returning to Figure 1, the SMO framework 10 provides the O1 interface defined in the O-RAN specifications. The Non-RT RIC 12 acquires KPIs (key performance indicators) from the base station function group 2 via the O1 interface. The SMO function unit 11 and the Non-RT RIC 12 exchange messages via the R1 interface.

[0021] In this embodiment, a new message A is added to the R1 interface between the SMO function unit 11 and the Non-RT RIC 12, which allows the Non-RT RIC 12 to obtain base station function layout information from the SMO function unit 11. The base station function layout information is information that indicates the current base station function layout in the RAN 1. The SMO function unit 11 transmits message A including the base station function layout information to the Non-RT RIC 12 via the R1 interface. The base station function layout control unit 20 obtains the base station function layout information from message A transmitted from the SMO function unit 11.

[0022] Furthermore, in this embodiment, new base station function allocation setting information is included in the existing message B that the Non-RT RIC 12 transmits to the SMO function unit 11 via the R1 interface. The base station function allocation setting information is information that indicates the contents of the change to the base station function allocation for changing the base station function allocation in the RAN1.

[0023] The base station function allocation control unit 20 includes base station function allocation setting information in message B transmitted from the Non-RT RIC 12. The SMO function unit 11 receives message B including the base station function allocation setting information via the R1 interface. The SMO function unit 11 acquires the base station function allocation setting information from message B received via the R1 interface.

[0024] 3 is a functional block diagram showing an example of the configuration of the base station function layout control unit 20 according to this embodiment. The base station function layout control unit 20 mainly includes a base station function layout information acquisition unit 201, a base station function layout information storage unit 202, a KPI acquisition unit 203, a base station function layout setting information transmission unit 204, and a control unit 205.

[0025] The base station function layout information acquisition unit 201 acquires base station function layout information via an interface with the SMO function unit 11. In this embodiment, this interface is the R1 interface. More specifically, the base station function layout information acquisition unit 201 acquires the base station function layout information from message A transmitted from the SMO function unit 11 via the R1 interface.

[0026] The base station function layout information storage unit 202 stores the base station function layout information acquired by the base station function layout information acquisition unit 201. The KPI acquisition unit 203 acquires KPIs from the base station function group 2 via the O1 interface.

[0027] The base station function allocation setting information transmission unit 204 transmits the base station function allocation setting information via an interface with the SMO function unit 11. In one example of this embodiment, this interface is the R1 interface. More specifically, the base station function allocation setting information transmission unit 204 transmits a message B including the base station function allocation setting information via the R1 interface. The control unit 205 determines the next base station function allocation based on the KPI acquired by the KPI acquisition unit 203.

[0028] Such a base station function allocation control unit 20 can be configured by installing applications (programs) that realize the functions detailed below on at least one general-purpose computer or server equipped with a CPU, ROM, RAM, bus, interface, etc. Alternatively, it can be configured as a dedicated machine or a single-function machine in which part of the application is implemented as hardware or software.

[0029] FIG. 4 is a flowchart showing the outline of the procedure of the base station function arrangement control method according to this embodiment. In step S1, the SMO function unit 11 performs initial settings on the base station function group 2 after the base station function group 2 has been started up.

[0030] In step S2, after the initial setting of the base station function group 2 is completed, the base station function layout control unit 20 (base station function layout information acquisition unit 201) acquires base station function layout information from message A transmitted from the SMO function unit 11. The base station function layout control unit 20 (base station function layout information storage unit 202) stores the base station function layout information.

[0031] In step S3, the base station function allocation control unit 20 (KPI acquisition unit 203) acquires the KPI from the base station function group 2 via the O1 interface. The base station function allocation control unit 20 (control unit 205) determines the next base station function allocation based on the KPI acquired by the KPI acquisition unit 203.

[0032] In step S4, the base station function allocation control unit 20 (control unit 205) generates base station function allocation setting information for the determined next base station function allocation. The base station function allocation setting information is information indicating the changes to be made to the base station function allocation in RAN1 to the next base station function allocation.

[0033] In step S5, the base station function location control unit 20 (base station function location setting information transmission unit 204) transmits a message B including the base station function location setting information generated by the control unit 205 to the SMO function unit 11.

[0034] In step S6, the SMO function unit 11 executes the setting to change the base station function allocation of the base station function group 2 based on the base station function allocation setting information included in the message B.

[0035] In step S7, after the setting of the change in base station function layout of base station function group 2 is completed, base station function layout control unit 20 (base station function layout information acquisition unit 201) acquires base station function layout information from message A transmitted from SMO function unit 11. Base station function layout control unit 20 (base station function layout information storage unit 202) stores the base station function layout information as the latest base station function layout information.

[0036] Next, the detailed procedure of the base station function allocation control method according to this embodiment will be described with reference to the sequence diagram of FIG.

[0037] In steps S101 to S105, a base station function startup sequence is executed. This base station function startup sequence is based on the O-RAN specification "Instantiate Network Function on O-Cloud (O-RAN.WG6.ORCH-USE-CASES-v01.00)."

[0038] In step S101, the "Network Function install project Mgr" sends a "Service Request" message to the "O-Cloud M&O".

[0039] In step S102, the "O-Cloud M&O" sends a "Create workload" message to the "DMS" via the O2 interface.

[0040] In step S103, the "DMS" transmits a "Deploy NF" message to each node in the base station function group 2 (O-CU, O-DU1, O-DUN).

[0041] In step S104, the "DMS" sends a "Notify workload created" message to the "O-Cloud M&O" via the O2 interface.

[0042] In step S105, the SMO function unit 11 transmits a "Configure NF" message to each node (O-CU, O-DU1, O-DUN) in the base station function group 2 via the O1 interface. This completes the base station function startup sequence.

[0043] In step S106, the SMO function unit 11 transmits a message A including base station function layout information via the R1 interface to the Non-RT RIC 12. The base station function layout control unit 20 acquires the base station function layout information from the message A.

[0044] In step S107, the base station function allocation control unit 20 receives a "Performance measurements" message via the O1 interface from each node (O-CU, O-DU1, O-DUN) in the base station function group 2. The base station function allocation control unit 20 acquires the KPI from the "Performance measurements" message.

[0045] In step S108, the base station function allocation control unit 20 determines the next base station function allocation based on the KPI acquired from the "Performance measurements" message. The base station function allocation control unit 20 generates base station function allocation setting information related to the determined next base station function allocation.

[0046] A sequence for changing the settings of the base station functions is executed from step S109 to step S110. This sequence is based on the O-RAN specification "Reconfiguration of O-RAN Virtual Network Function(s) (O-RAN.WG6.ORCH-USE-CASES-v01.00)."

[0047] In step S109, the base station function allocation control unit 20 transmits a "Reconfig NF set" message including the base station function allocation setting information generated in step S108 to the SMO function unit 11 via the R1 interface. The "Reconfig NF set" message corresponds to the existing message B. The SMO function unit 11 acquires the base station function allocation setting information from the "Reconfig NF set" message received via the R1 interface.

[0048] In step S110, the SMO function unit 11 transmits a "Configure NF" message via the O1 interface to each node (O-CU, O-DU1, O-DUN) in the base station function group 2. This "Configure NF" message contains setting information for changing to the next base station function configuration indicated in the base station function configuration setting information acquired in step S109. As a result, the base station function configuration of the base station function group 2 is changed to the next base station function configuration.

[0049] Next, we will explain how the base station function allocation control unit 20 determines the base station function allocation based on the prediction model M in step S108. The base station function allocation determination method according to this embodiment, including prior preparation, includes the following steps: (1) collecting data necessary for learning the prediction model M; (2) learning the prediction model M; (3) using the prediction model M to obtain predicted values ​​for the base station function allocation, resource usage, and required quality achievement rate for all services, for each cell; and (4) optimizing the combination of base station function allocations between cells using the list of combinations of the predicted values ​​obtained for each cell.

[0050] (1) Collecting data necessary for training the predictive model M The data items required for learning the prediction model M are information indicating the state of the RAN at a certain point t in the control period (required throughput z for each cell / service, various resource usage ω, and base station function allocation u for each cell / service), and the resource usage ω and required quality achievement rate q at the next control period, point t+1.

[0051] As a method of collection, manual control may be performed on the RAN, or a relatively low-risk feedback control such as that described in Non-Patent Document 4 may be performed to collect various information based on the O-RAN specifications.

[0052] (2) Training the predictive model M The prediction model M is a resource usage prediction model M Ω and the required quality achievement rate prediction model M QThese may be prepared separately, and each may be configured with a multilayer perceptron. Note that the base station function allocation may be vectorized in an embedding layer, on the premise that the combinations of base station function allocations for all services are input in correspondence with positive integers.

[0053] The prediction model M is trained by using the required throughput Z, the amount of various resources used ω, and the base station arrangement u at a certain point in the control cycle for each cell as input information, predicting the amount of resource used ω and the required quality achievement rate q at a certain point in the next control cycle, t+1, measuring the prediction error using RMSE, and updating the prediction model M using the error backpropagation method.

[0054] FIG. 6 is a diagram showing an example of a learning method for the prediction model M. Here, the resource usage prediction model M Ω and the required quality achievement rate prediction model M Q The following describes an example in which the above two methods are learned separately.

[0055] The prediction model M may be different for each cell, or a single prediction model M may be learned between cells, ignoring the differences between cells. However, the amount of computer resource usage ω n CC and the transmission path resource usage ω n TL Since the amount of data is shared between cells, it is desirable to convert it into the amount of data used for each cell in advance.

[0056] Resource usage per cell of the transmission path ω n TL can be measured by monitoring which cell the packet originates from. n CC As described in Non-Patent Document 3, can be calculated by assuming that the usage of computer resources is proportional to the required throughput.

[0057] In this embodiment, the input information is a throughput Z n,s,t, base station placement u n,s,t and resource usage ω t The resource usage ω t The resource usage of the transmission line in the control period t is n TL ,Computer resources at the antenna site ω n CA and the computer resource ω of the accommodating station n CC Base station layout u n,s,t are expressed as ternary numbers {0,1,2}, which correspond to the C-RAM / S-RAM / D-RAM arrangements, respectively.

[0058] Base station placement u n,s,t is converted to a decimal number and then input to the first and second embedding layers, where the integers are converted into a dense vector that can be learned. Ω is the output of the first embedding layer, and the throughput Z n,s,t and resource usage ω t is input. The quality achievement prediction model M Q is the output of the second embedding layer, and the throughput Z n,s,t and resource usage ω t is entered.

[0059] Resource usage prediction model M Ω is the resource usage in control period t+1, ω n,t+1 Quality achievement prediction model M Q is the required quality achievement rate q in control period t+1 n,s,t+1 The required quality achievement rate q is the throughput achievement rate q n,s tp and the communication delay achievement rate q n,s la Includes:

[0060] The correct answer is the resource usage ω n,t+1 and required quality achievement rate q n,s,t+1The RMSE (Root Mean Squared Error) of the resource usage and the required quality achievement rate are calculated as the objective functions.

[0061] (3) Using the prediction model M, we obtain predicted values ​​for base station function allocation, resource usage, and required quality achievement rate for all services for each cell. In the prediction for each cell, information indicating the most recent state of RAN1 other than the base station function placement for each cell is input to the prediction model M, and possible patterns of base station function placement for all services are input, thereby obtaining predicted values ​​for resource usage and the required quality achievement rate when a certain base station function placement is performed for all services in a specific cell.

[0062] For example, if there are six services, there are three possible base station function placement patterns, so the total number of possible base station function placement patterns for all services is 3^6 = 729. Therefore, for each cell, 729 combinations of base station function placement candidates, predicted resource usage values, and predicted desired quality achievement rates can be obtained.

[0063] The following explains how to calculate the average value of the expected quality achievement rate. tp The average value of Q tp,n,u can be calculated, for example, by the following equation (1):

[0064]

number

[0065] Predicted achievement rate of communication delay la for each service s Q la The average value of Q la,n,u can be calculated, for example, by the following equation (2):

[0066]

number

[0067] The predicted value Q of the required quality achievement rate of placement u in each cell n n,u is calculated using the following equation (3).

[0068]

number

[0069] By performing the above calculation for each cell, 729 predicted values ​​of the required quality achievement rate for each cell are obtained as shown in the following formula (4).

[0070]

number

[0071] (4) Optimize the combination of base station function placement between cells using the list of predicted values ​​obtained for each cell. In inter-cell optimization, a list of combinations of base station function placement candidates, predicted resource usage amounts, and predicted required quality achievement rates obtained for each cell is used to optimize the combination of base station function placements between cells, taking into account various resource constraints.

[0072] The hill-climbing method, which uses cells as a dimension, can be applied to optimize the combinations. The specific steps for optimizing the combinations using the hill-climbing method are shown in Figure 7. Here, we will explain the assumption that the predicted values ​​of resource usage and the required quality achievement rate when a certain base station function allocation is performed for all services in a specific cell at a certain point in time have been obtained in advance using the prediction model M.

[0073] In step S301, the required quality achievement rate Q n,u The base station function allocation u that maximizes the predicted value of is selected as the initial solution.

[0074] In step S302, the value of the objective function based on the initial solution is calculated. The objective function Obj is given, for example, by the following equation (5): where λ is a balancing parameter that indicates how much weight is given to the constraint violation L in the objective function.

[0075]

number

[0076] Constraint violation of the computer resources of the accommodating station L CC is given by the following equation (6): The amount of computer resource usage of the accommodating station ω n CC Since the resource is shared between cells, the constraint violation Lcc of the accommodation station is calculated by summing all the cells and then calculating the allowable capacity Ω CC It can be found by subtracting

[0077]

number

[0078] Violation of computer resource constraints at the antenna site L CA is given by the following equation (7): The amount of computer resource usage at the antenna site ω n CA Since resources are shared within the cell, the antenna constraint violation L CA is calculated by adding up all services for each cell and then calculating the capacity Ω CA It can be found by subtracting

[0079]

number

[0080] Violation of resource constraints on the transfer path L TL is given by the following equation (8): The amount of transmission path resource usage ω n TL Since the resource is shared between cells, the total is calculated for all cells and then the capacity Ω TL Subtract the value to see if there are any violations.

[0081]

number

[0082] Therefore, the constraint violation penalty is calculated as the sum of all constraint violations using the following equation (9).

[0083]

number

[0084] Since this initial solution may significantly violate resource constraints, we next calculate the objective function for all placements when the placement of each cell is changed, identify the placement that maximizes the objective function after the change, and record the change in the objective function at that time.

[0085] The processes in steps S303 and S304 are executed for each cell. In step S303, the objective function for changing the placement in one specific cell is calculated for all placements. To calculate the objective function after the change, for example, the following equation (10) can be used for the violation of the computer resource constraints of the accommodating station.

[0086]

number

[0087] In step S304, the layout that maximizes the objective function after the change is identified, and the amount of change in the objective function at that time is recorded.

[0088] Once the change in the objective function for all cells has been recorded, the process proceeds to step S305, where the cell with the largest change in the objective function is identified. In step S306, it is determined whether the largest change is equal to or less than zero. If it is not equal to or less than zero, the process proceeds to step S307, where the solution is updated to the layout for that cell with the largest post-change objective function, and the process returns to step S302, where the above processes are repeated.

[0089] After that, the cell with the largest change in the objective function is identified, and if the change is 0 or less, there is no improvement and the optimization is terminated, but if the change is positive, there is an improvement and the solution is updated to the arrangement in that cell where the changed objective function is largest, and the calculation of the objective function value is resumed.

[0090] In addition to the hill climbing method described above, other general combinatorial optimization methods can be applied to combinatorial optimization. For example, simulated annealing may be applied.

[0091] According to this embodiment, it is possible to obtain an effect that it is possible to change the arrangement of base station functions in a radio access network (RAN) conforming to the O-RAN specifications.

[0092] This will enable, for example, improvements to the overall service quality of wireless access networks, thereby contributing to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs), which is to "Build resilient infrastructure, promote sustainable industrialization and foster innovation." [Explanation of symbols]

[0093] 1...Radio Access Network (RAN), 2...Base Station Function Group, 10...SMO Framework, 11...SMO Function Unit, 12...Non-RT RIC, 20...Base Station Function Layout Control Unit, 201...Base Station Function Layout Information Acquisition Unit, 202...Base Station Function Layout Information Storage Unit, 203...KPI Acquisition Unit, 204...Base Station Function Layout Setting Information Transmission Unit, 205...Control Unit

Claims

1. 1. A base station functions allocation determination device that determines the allocation of base station functions in each cell of a radio access network, comprising: a prediction model for predicting resource usage and a required quality achievement rate in a control period t+1 based on the base station function allocation, required throughput, and resource usage in each control period t; a means for inputting the required throughput and resource usage amount at an arbitrary control time for each cell and the allocation of all base station functions into a corresponding prediction model to predict the resource usage amount and the required quality achievement rate for each base station allocation; means for optimizing a combination of base station function allocations between cells based on a predicted result of resource usage and a required quality achievement rate in each base station allocation in each cell; A base station function allocation determination device that adopts a base station function allocation that is an optimized combination for each cell.

2. 2. The base station function allocation determining device according to claim 1, wherein the prediction model predicts the required quality achievement rate for each service using the base station function allocation and required throughput as inputs for each service.

3. 3. The base station function allocation determination device according to claim 1, wherein the prediction models include a first prediction model that predicts resource usage in a control period t+1 based on base station allocation, required throughput, and resource usage in the control period t, and a second prediction model that predicts a required quality achievement rate.

4. 3. The base station function allocation determining device according to claim 1, wherein said optimizing means optimizes the combination of base station function allocation between cells using a hill climb method.

5. 3. The base station function allocation determination device according to claim 1, wherein the prediction model is constructed by supervised learning using, as an objective function, an error in resource usage and a required quality achievement rate predicted for each cell in a control period t+1 based on the base station function allocation, required throughput, and resource usage in a control period t.

6. 3. The base station function allocation determination device according to claim 1, wherein the base station function allocation is determined so that the virtualized DUs and CUs are distributed to at least one of the accommodating stations and antenna sites of each cell.

7. 7. The base station function allocation determination device according to claim 6, wherein the resource usage of each cell is the usage of computer resources of the accommodating station, computer resources of the antenna site, and resources of the transmission path connecting the accommodating station and the antenna site.

8. 3. The base station functions allocation determining apparatus according to claim 1, wherein the required quality achievement rate is an achievement rate of throughput and an achievement rate of communication delay.

9. A base station functions allocation method for determining the allocation of base station functions in each cell of a radio access network, comprising: a prediction model that predicts resource usage and a required quality achievement rate in a control period t+1 based on the base station function allocation, required throughput, and resource usage in the control period t for each cell, inputting the required throughput and resource usage of each cell and all base station function allocations, and predicting the resource usage and the required quality achievement rate for each base station allocation; optimizing a combination of base station function locations between cells based on the predicted results of resource usage and required quality achievement rate in each base station location in each cell; A base station function allocation determination method characterized by adopting a base station function allocation that is an optimized combination for each cell.

10. 10. The method for determining base station function allocation according to claim 9, wherein the prediction model is constructed by supervised learning using, as an objective function, an error between resource usage and a required quality achievement rate in a control period t+1 predicted for each cell based on the base station function allocation, required throughput, and resource usage in a control period t.

11. A base station functions allocation determination program for determining the allocation of base station functions in each cell of a wireless access network, comprising: a step of predicting resource usage and a required quality achievement rate for each base station arrangement by inputting the required throughput and resource usage of each cell and all base station function arrangements into a prediction model that predicts resource usage and a required quality achievement rate for a control period t+1 based on the base station function arrangement, required throughput, and resource usage for each cell in the control period t; optimizing a combination of base station function allocations between cells based on the predicted results of resource usage and required quality achievement rate in each base station allocation in each cell; A base station function allocation determination program characterized by adopting a base station function allocation that is an optimized combination for each cell.

12. 12. The base station function allocation determination program according to claim 11, further comprising a procedure for constructing the prediction model by supervised learning using, as an objective function, an error in resource usage and a required quality achievement rate in a control period t+1 predicted for each cell based on the base station function allocation, required throughput, and resource usage in a control period t.

Images