Control device for wireless access network

Abstract

A control unit (Non-RT RIC) for a non-real-time system and a control unit (Near-RT RIC) for a near-real-time system are layered, and learning reasoning units (11, 12, 13, 16, 17, 18) for controlling a wireless access network on the basis of the result of reasoning by applying latest data to a learning model generated on the basis of data collected from an O-RAN base station device (10) are disposed in the control unit for the near-real-time system. A re-learning unit (14, 15, 19, 20) is disposed in a control unit of a non-real-time system, said re-learning unit detecting, on the basis of the history of the collected data, whether or not a concept drift occurs, and if the occurrence of the concept drift is detected, causing the learning reasoning unit to re-learn the learning model.

Description

Technical Field

[0001] The present invention relates to a control device for a wireless access network, and more particularly to a control device for a wireless access network having a function of relearning a learning model generated by learning data collected from the wireless access network. Background Art

[0002] In the Radio Access Network (RAN), the functions of the previously integrated base station are divided into the CU (Centralized Unit) for session processing, the DU (Distributed Unit) for baseband processing, and the RU (Radio Unit) for wireless processing. The O-RAN Alliance (Open Radio Access Network Alliance) is promoting research on specifications for opening up the interface specifications between the units.

[0003] In the Beyond 5G (beyond the fifth generation communication technology) system, it is expected that the performance such as throughput, communication latency, and number of connections will be further improved, and a variety of services (such as robot control, connected cars, AR / VR, etc.) will be provided. AI (artificial intelligence) / ML (machine learning) is attracting attention as a key technology for realizing these.

[0004] In non-patent literature 1 and 2, the following situation is studied: in order to maximize network performance within the limited network resources in RAN, AI / ML is applied to various purposes such as beamforming control, wireless resource allocation, traffic prediction, base station function configuration, etc.

[0005] Non-patent document 3 discloses a technology of generating a learning model by learning based on data collected from the RAN, performing inference using the data collected from the RAN and the learning model, and controlling the RAN based on the inference result.

[0006] However, as time passes and the environment changes, the characteristics of data used for inference may change from the data at the time of learning (concept drift), which may reduce the inference performance of the model.

[0007] In response to the technical problems described above, the inventors of the present invention proposed an AI system that accumulates and monitors data related to AI / ML learning and reasoning based on an O-RAN base station device, detects concept drift, and performs relearning, and filed a patent application (Patent Document 1).

[0008] Figure 6 This is a functional block diagram showing the conventional structure of an AI system that detects concept drift and performs relearning.

[0009] The data collection unit 11 repeatedly collects the latest data from the O-RAN base station device 10, provides the collected latest data (collected data) to the AI / ML learning unit 12 and the AI / ML reasoning unit 13, and accumulates the collected latest data (collected data) in the data accumulation unit 14. The collected data accumulated in the data accumulation unit 14 is managed by the AI / ML database 15. The AI / ML learning unit 12 learns the collected data and generates a learning model for controlling the O-RAN base station device 10.

[0010] The AI / ML model management unit 16 manages the learning model generated in the past by the AI / ML learning unit 12. The AI / ML reasoning unit 13 performs reasoning based on the newly collected data and the learning model collected by the data collection unit 11, and outputs the reasoning result to the control unit 17 and the reasoning performance measurement unit 18. The control unit 17 controls the O-RAN base station device 10 based on the reasoning result.

[0011] The inference performance measurement unit 18 determines the inference performance based on the latest data collected after the control unit 17 controls the O-RAN base station device 10 based on the inference result and the inference result, and stores inference performance data indicating the determined inference performance in the AI / ML database 15 .

[0012] The concept drift detection unit 19 periodically acquires at least one of the collected data and the inference performance data from the AI / ML database 15 to determine whether a concept drift has occurred. If the occurrence of a concept drift is detected, the concept drift detection unit 19 instructs the relearning control unit 20 to generate a new learning model (relearning). The relearning control unit 20 provides the AI / ML learning unit 12 with data for relearning and instructs relearning.

[0013] If re-learning is instructed, the AI / ML learning unit 12 generates a new learning model based on the newly collected data by the data collection unit 11, and outputs the new learning model to the AI / ML model management unit 16. The AI / ML model management unit 16 compares the current learning model used by the AI / ML reasoning unit 13 with the new learning model, and outputs the new learning model to the AI / ML reasoning unit 13 if the reasoning performance based on the new learning model is higher than the reasoning performance based on the current learning model.

[0014] Thereafter, the AI / ML reasoning unit 13 performs reasoning using the new learning model. If the reasoning performance based on the new learning model is lower than the reasoning performance based on the current learning model, the AI / ML model management unit 16 can instruct the AI / ML learning unit 12 to relearn.

[0015] Prior art literature

[0016] Patent Literature

[0017] Patent Document 1: Japanese Patent Application No. 2022-046347

[0018] Non-patent literature

[0019] Non-patent document 1: MEMorocho-Cayamcela, H.Lee and W.Lim, "Machine Learning for 5G / B5G Mobile and Wireless Communications: Potential, Limitations, and Future Directions," in IEEE Access, vol.7, pp.137184-137206, 2019

[0020] Non-patent literature 2: J.Kaur, MAKhan, M.Iftikhar, M.Imran and Q.Emad Ul Haq, "Machine Learning Techniques for 5G and Beyond," in IEEE Access, vol.9, pp.23472-23488, 2021

[0021] Non-patent document 3: O-RAN Alliance, “AI / ML workflow description and requirements,” O-RAN.WG2.AIML-vo1.03, Jul.2021 Summary of the invention

[0022] Problems to be solved by the invention

[0023] like Figure 7 As shown, the RAN intelligent controller (RIC) responsible for controlling and optimizing RAN functions has a hierarchical structure of "Non-RT (Real Time) RIC" which is a non-real-time system component with different control cycles and "Near-RT RIC" which is a near-real-time system component.

[0024] Here, there are different characteristics as follows: Non-RT RIC has a large area that becomes the control object when the control period is more than 1 sec. On the contrary, Near-RT RIC has a small area that becomes the control object when the control period is 10msec to 1sec. Therefore, the optimal configuration of each functional block related to AI / ML of Near-RT RIC and Non-RT RIC has been studied in the past.

[0025] For example, it is assumed that the Non-RT RIC is set up in the machine room and the Near-RT RIC is configured at the edge site such as the roof of the building. In this case, the following technical problems may arise: Prioritizing real-time performance, not only the functions involved in AI / ML learning, but also all the functions involved in AI / ML re-learning are configured in the Near-RT RIC.

[0026] First, the processing load of Near-RT RIC increases. That is, edge sites have power and space limitations, so they cannot be equipped with sufficient computers.

[0027] Second, only the information below the Near-RT RIC level can be used for concept drift detection. That is, since the information of neighboring regions cannot be used, the detection of concept drift is delayed.

[0028] For example, if road construction occurs in a certain area and the volume of connected car traffic changes, the change in vehicle traffic will also affect adjacent areas. In this case, if only the information of the own area is monitored, the environmental changes caused by the road construction cannot be detected immediately, so the follow-up of re-learning of environmental changes becomes low.

[0029] Means used to solve problems

[0030] The object of the present invention is to provide a control device for a wireless access network as follows: to solve the above-mentioned technical problems and optimize the configuration of Near-RT RIC and Non-RT RIC of functional blocks related to AI / ML.

[0031] In order to achieve the above-mentioned purpose, the control device of the wireless access network formed by hierarchical control units of a non-real-time system and a near-real-time system of the present invention generates a learning model based on data collected from the wireless access network, and configures a learning reasoning unit that controls the wireless access network based on a result inferred by applying the latest data collected in the learning model to the control unit of the near-real-time system, and configures a re-learning unit that detects whether a concept drift occurs based on the history of the collected data and causes the learning reasoning unit to re-learn the learning model if the occurrence of concept drift is detected to the control unit of the non-real-time system.

[0032] Effects of the Invention

[0033] According to the present invention, the following effects are achieved.

[0034] (1) The functions related to learning and reasoning are arranged in the control unit of the near real-time system, and the functions related to re-learning are arranged in the control unit of the non-real-time system. Therefore, the processing load of the control unit of the near real-time system can be reduced.

[0035] (2) Since the functions related to relearning are arranged in the control unit of the non-real-time system, even if the computer resources of the edge site are limited, the learning model can be relearned.

[0036] (3) Since the information of the lower level of the control unit of the non-real-time system can be used for the detection of concept drift, the concept drift can be detected immediately and the ability to follow environmental changes can be improved.

[0037] Other features and advantages of the present invention will become apparent from the following description with reference to the accompanying drawings. In the accompanying drawings, the same or similar components are denoted by the same reference numerals. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present invention and, together with the description, serve to explain the principles of the present invention.

[0039] Figure 1 This is a functional block diagram showing the configuration of a main part of an O-RAN control device according to an embodiment of the present invention.

[0040] Figure 2A This diagram shows an example in which a method of specifying an object is added to the A1 interface.

[0041] Figure 2B This diagram shows an example in which a method of specifying an object is added to the A1 interface.

[0042] Figure 3A This is a diagram showing an example in which a policy instruction method is added to the A1 interface.

[0043] Figure 3B This is a diagram showing an example in which a policy instruction method is added to the A1 interface.

[0044] Figure 4A This diagram shows an example of adding data for re-learning to the A1 interface.

[0045] Figure 4B This diagram shows an example of adding data for re-learning to the A1 interface.

[0046] Figure 5 It is a sequential flow chart showing the actions of the present invention.

[0047] Figure 6 This is a functional block diagram showing the conventional structure of an AI system that detects concept drift and performs relearning.

[0048] Figure 7 It is a functional block diagram of the RAN Intelligent Controller (RIC). DETAILED DESCRIPTION

[0049] Hereinafter, the embodiments are described in detail with reference to the accompanying drawings. In addition, the following embodiments do not limit the invention involved in the technical solution. In addition, the combination of features described in the embodiments may not all be necessary for the invention. Two or more of the multiple features described in the embodiments may also be combined arbitrarily. In addition, the same or similar configurations are marked with the same reference numerals, and repeated descriptions are omitted.

[0050] Figure 1 This is a functional block diagram showing the structure of the main part of the O-RAN control device involved in one embodiment of the present invention, and the illustration of the structure that is not necessary for the description of the present invention is omitted. In addition, the same reference numerals as those described above represent the same or equivalent parts. The feature of this embodiment is that the functions involved in AI / ML learning are configured in Near-RT RIC, and the functions involved in AI / ML re-learning are configured in Non-RT RIC.

[0051] The O-RAN control device is composed of O-CU / O-DU31, Near-RT RIC32 and Non-RT RIC33. Each function can notify each other through various interfaces including O1 interface, A1 interface and E2 interface specified by O-RAN Alliance. The O-RAN base station device 10 is configured in O-CU / O-DU31.

[0052] The Near-RT RIC 32 is mainly configured with a data collection unit 11, an AI / ML learning unit 12, an AI / ML reasoning unit 13, an AI / ML model management unit 16, a control unit 17, and a reasoning performance measurement unit 18 as functions related to AI / ML learning and reasoning. The Non-RT RIC 33 is mainly configured with a data storage unit 14, an AI / ML database 15, a concept drift detection unit 19, and a relearning control unit 20 as functions related to relearning of the learning model.

[0053] In the Near-RT RIC 32, the data storage unit 14 collects the latest data from the O-RAN base station device 10 and sends the latest data to the data storage unit 14 of the Non-RT RIC 33 via the O1 interface. The inference performance measurement unit 18 sends the inference performance data to the AI / ML database 15 of the Non-RT RIC 33 via the O1 interface. When concept drift is detected, the relearning control unit 20 of the Non-RT RIC 33 requests relearning to the AI / ML learning unit 12 of the Near-RT RIC 32 via the A1 interface.

[0054] Thus, in this embodiment, since the functions involved in the learning reasoning of AI / ML and the functions involved in the re-learning of the learning model are respectively distributed in Near-RT RIC32 and Non-RT RIC33, the following three pieces of information are specially added to the A1 interface as information related to the re-learning requirements.

[0055] (1) Designation of the object

[0056] (2) Strategy Instructions

[0057] (3) Data used for relearning

[0058] In order to specify the object (learning model) described in (1), the ID of the learning model is added in this embodiment. In order to indicate the strategy described in (2), the instruction of re-learning is added in this embodiment. As data for the re-learning described in (3), in this embodiment, "experience information (state, next state, action, reward)" is used for reinforcement learning, and "input data and correct answer label" is used for supervised learning. The data format can be compressed in a table format.

[0059] Figure 2A , Figure 2B This is a diagram showing an example of a method for specifying an object (1). In O-RAN, policies and operations can be sent from Non-RT RIC 33 to Near-RT RIC 32 via the A1 interface. In the A1 interface, ScopeIdentifier (O-RAN.WG2.A1TD-v02.00) is defined as an identifier for specifying an object to which policies and operations are applied. In this embodiment, in order to specify an AI / ML model to be relearned, Figure 2A As shown, add the AI / ML model ID to the existing ScopeIdentifier, and further Figure 2B As shown, a table is added to define the details of the AI / ML model ID.

[0060] Figure 3A , Figure 3B 2 is a diagram showing an example of the method of indicating the above-mentioned (2) policy. In the A1 interface, policy objectives are defined as the sending policy from the Non-RT RIC 33 to the Near-RT RIC 32. In this embodiment, in order to instruct the AI / ML model to relearn, Figure 3A As shown, AimlObjectives (artificial intelligence markup language objects) are added to the existing policy objectives as strategies related to AI / ML.

[0061] Furthermore, if Figure 3B As shown, a table defining the details of AimlObjectives is added, and retrain indicating relearning is added.

[0062] Figure 4A , Figure 4B This is a diagram showing an example of data used for the above (3) relearning. In the case of reinforcement learning, for example, if considering base station function configuration control, an action (base station function configuration information) is determined from the state at a certain point in time (throughput, delay, resource utilization rate, etc.), and if the state is switched to the next state, a reward for evaluating the action (such as the achievement rate of the required quality) is obtained.

[0063] Therefore, these series of data are used as experience information for learning, thereby obtaining the optimal control that maximizes the reward. Figure 4A The table format shown sends m pieces of experience information (states s1 to s n 、Next state ns1~ns n , action a, reward r).

[0064] In the case of supervised learning, for example, if we consider traffic flow prediction, we learn the input data (time series information of traffic flow, etc.) and its correct answer label (correct answer value of traffic flow at the next moment) to infer the correct answer from the input data. Figure 4B Send m learning data (input data x1~x n 、Correct answer label y1~y n ).

[0065] Figure 5 This is a sequence flow chart showing the operation of this embodiment, and the description here focuses on the communication between the O-CU / O-DU, Near-RTRIC, and Non-RT RIC.

[0066] In this embodiment, communication between O-CU / O-DU and Near-RT RIC is performed via the E2 interface, communication from Near-RT RIC to Non-RT RIC is performed via the O1 interface, and communication from Non-RT RIC to Near-RT RIC is performed via the A1 interface.

[0067] The O-CU / O-DU repeatedly sends the latest data of the O-RAN base station device 10 to the Near-RT RIC at a predetermined period. In this embodiment, at time t1, if the O-CU / O-DU sends the latest data to the Near-RT RIC, the Near-RT RIC acquires the latest data through the data collection unit 11.

[0068] The Near-RT RIC sends the latest data to the Non-RT RIC at time t2 , and the AI / ML inference unit 13 applies the latest data to the current learning model to perform inference, and notifies the control unit 17 and the inference performance measurement unit 18 of the inference result.

[0069] At time t3, the control unit 17 instructs the O-RAN base station of the O-CU / O-DU to control based on the inference result. The inference performance measurement unit 18 determines the inference performance based on the latest data collected after the control unit 17 controls the O-RAN base station device 10 based on the inference result and the inference result, and at time t4, transmits performance data indicating the inference performance to the Non-RT RIC.

[0070] In Non-RT RIC, the concept drift detection unit 19 monitors the latest data and performance data. If a concept drift is detected at time t5, the relearning control unit 20 instructs the learning model of the object specified by Near-RT RIC to relearn at time t6, and further reads the data for relearning from the AI / ML database 15 and sends it.

[0071] In the Near-RT RIC, if the re-learning instruction and the re-learning data are obtained, at time t7, the AI / ML learning unit 12 performs re-learning to generate a learning model, and updates the learning model and logs it to the AI / ML model management unit 16. Therefore, each time data is collected, control based on the re-learned learning model is performed.

[0072] According to this embodiment, the functions involved in learning reasoning are configured in the Near-RT RIC, and on the other hand, since the functions involved in relearning are configured in the Non-RT RIC, the processing load of the Near-RT RIC can be reduced. Therefore, even if there are restrictions on the computer resources of the edge site, if the concept drift does not occur frequently, the concept drift can be detected immediately, which can improve the tracking of environmental changes.

[0073] As a result, depending on the implementation method, it can contribute to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs) to "improve resilient infrastructure and promote sustainable industrialization" and Goal 11 to "make cities inclusive, safe, resilient, and sustainable."

[0074] The present invention is not limited to the above-described embodiments, and various changes and modifications can be made without departing from the spirit and scope of the present invention. Therefore, in order to disclose the scope of the present invention, the following claims are attached.

[0075] This application claims priority based on Japanese patent application No. 2022-149389 filed on September 20, 2022, the entire contents of which are incorporated herein by reference.

Claims

1. A control device for a wireless access network, wherein a control unit of a non-real-time system and a control unit of a near-real-time system are hierarchically connected, wherein: The control device of the wireless access network comprises: a learning inference unit that generates a learning model based on data collected from the wireless access network, and controls the wireless access network based on a result inferred by applying the latest collected data to the learning model; as well as a re-learning unit that detects whether a concept drift occurs based on the history of the collected data, and if the occurrence of a concept drift is detected, causes the learning and reasoning unit to re-learn the learning model, The learning inference unit is configured in a control unit of the near real-time system, and the re-learning unit is configured in a control unit of the non-real-time system.

2. The control device for a wireless access network according to claim 1, characterized in that: The learning reasoning unit comprises: a data collection unit that collects the latest data from the wireless access network; A learning unit, which generates a learning model based on the collected data; a control unit configured to apply the collected data to the learning model to perform inference, and control the wireless access network based on a result of the inference; as well as an inference performance measuring unit for measuring inference performance based on the collected data and the inference result, The data collection unit and the inference performance measurement unit transmit the latest data and the inference result, respectively, to the relearning unit via the O1 interface.

3. The control device for a wireless access network according to claim 2, characterized in that: The relearning unit comprises: A database storing the collected data and the inference performance; a concept drift detection unit that detects the occurrence of concept drift based on the data accumulated in the database and the reasoning performance; as well as a re-learning control unit, which, if the occurrence of the concept drift is detected, sends information for causing the learning inference unit to re-learn the learning model; The relearning control unit sends information for relearning to the learning inference unit via the A1 interface.

4. The control device for a wireless access network according to claim 3, characterized in that: The information for relearning includes an ID of a learning model that specifies an object, an instruction for relearning that specifies a strategy, and a data type and a data format that specifies data for relearning.

5. The control device for a wireless access network according to claim 4, characterized in that: The data types include state, next state, action, and reward as experience information in reinforcement learning.

6. The control device for a wireless access network according to claim 4 or 5, characterized in that: The data types include input data and correct answer labels in supervised learning.

Images