Base station device, information processing method, and information processing program

Abstract

This base station device (100), in which both a radio access network (RAN) control function for controlling an RAN and an artificial intelligence (AI) function having AI are implemented on a server by software, comprises a determination unit (132), a resource control unit (133), and an execution control unit (134). When an execution request for a prescribed AI process among AI processes that are to be executed by the AI function is received, the determination unit (132) determines whether the prescribed AI process can be executed by using, among calculation resources included in the server, the remaining surplus resource that is not used to execute the RAN control function. When it is determined that the prescribed AI process cannot be executed by means of the surplus resource, the resource control unit (133) limits a radio resource of the RAN control function, thereby generating, for the calculation resources, an AI resource capable of executing the prescribed AI process. The execution control unit (134) executes the prescribed AI process by means of the AI resource.

Description

Base station device, information processing method, and information processing program 【0001】 The present invention relates to a base station device, an information processing method, and an information processing program. 【0002】 Conventionally, a method for efficiently utilizing the resources of a base station device implemented by software on a general-purpose server has been proposed. 【0003】 Japanese Patent Application Laid-Open No. 2019-92081 【0004】 However, the above prior art does not assume a base station device in which a RAN function and an AI function are realized on the same server. For this reason, in the above prior art, nothing is considered about operating the AI function while maintaining the operation of the RAN function of the base station device. Therefore, in the above prior art, it is not always possible to appropriately allocate computing resources between the RAN function and the AI function of the base station device. 【0005】 Thus, the base station device according to the present application is a base station device in which both a RAN control function for controlling a radio access network (RAN) and an AI function having an artificial intelligence (AI) are realized on a server by software, and among AI processes that are processes executed by the AI function, when a request for executing a predetermined AI process is received, a determination unit that determines whether the predetermined AI process can be executed with surplus resources remaining among the computing resources of the server and not used for executing the RAN control function, a resource control unit that generates, with respect to the computing resources, AI resources that enable execution of the predetermined AI process by restricting the radio resources of the RAN control function when it is determined that the predetermined AI process cannot be executed with the surplus resources, and an execution control unit that causes the predetermined AI process to be executed with the AI resources. 【0006】Figure 1 is a diagram showing an example of a vRAN system configuration. Figure 2 is a diagram showing an example of an AI-RAN structure. Figure 3 is an explanatory diagram of surplus resources. Figure 4 is a diagram showing an overview of the proposed method. Figure 5 is a diagram showing an example of the configuration of a base station device according to the embodiment. Figure 6 is a diagram showing an example of an AI processing-related information storage unit 121. Figure 7 is a diagram showing an example of the operation of the base station device. Figure 8 is a flowchart showing the information processing procedure executed by the base station device 100. Figure 9 is a hardware configuration diagram showing an example of a computer that realizes the functions of the base station device according to the embodiment. 【0007】 Embodiments of the present invention will be described in detail below with reference to the attached drawings. In this specification and the drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant descriptions will be omitted. 【0008】 The one or more embodiments (including examples, modifications, and applications) described below can each be implemented independently. On the other hand, at least some of the embodiments described below may be implemented in appropriate combination with at least some of the other embodiments. These embodiments may contain novel features that differ from each other. Therefore, these embodiments may contribute to solving different objectives or problems and may produce different effects. 【0009】 <Embodiment> [1. Introduction] (About vRAN) In the communication network to which the mobile communication terminals (User Equipment, UE) that we use every day are connected, i.e., the Radio Access Network (RAN), a vast number of radio base stations (gNBs) are operated to provide communication services over a wide area. The main roles of gNBs include transmitting and receiving radio waves for wireless communication with UEs and processing radio signals. Because gNBs require high throughput and very strict latency requirements, dedicated hardware optimized for gNBs has traditionally been used. 【0010】On the other hand, server virtualization and cloud computing have established mechanisms that allow users to access the necessary computing resources when and where they are needed. Specifically, by abstracting computing resources and separating hardware and software, it has become possible to reduce physical resources and rapidly deploy systems. In communication infrastructure, this implementation method is called NFV (Network Functions Virtualization). 【0011】 According to NFV, by implementing network functions that were previously handled by dedicated hardware as software functions, it became possible to run virtualized workloads on inexpensive general-purpose servers. Therefore, vRAN (virtual Radio Access Network) is a virtualization system that implements the gNB (graphic network bounding box) that constitutes RAN, with the gNB's functions configured as software on a general-purpose server. Furthermore, vRAN is a concrete example of a system requiring high throughput and low latency. 【0012】 Figure 1 shows an example of a vRAN system configuration. Performing all the complex communication processing on a gNB would be costly, so in practice, small facilities called "slave stations" that perform minimal communication processing are installed throughout the country, and advanced communication processing (radio signal processing) is performed at a "master station" located beyond the slave stations. 【0013】 In vRAN, virtualization is performed on the master station portion. Traditionally, communication processing at the master station was performed using dedicated hardware. By virtualizing this function within a general-purpose server, advanced communication processing can be performed without the need for dedicated hardware. 【0014】 As shown in Figure 1, vRAN is based on the idea of ​​dividing the gNB, which was previously composed of dedicated hardware, into three components: CU (Central Unit), DU (Distributed Unit), and RU (Radio Unit). In this example, the RU corresponds to a slave station, and the DU and CU correspond to master stations. 【0015】Therefore, the functional parts of the DU and CU, which are responsible for wireless signal processing, are implemented in software on the general-purpose server. The vRAN transmits and receives wireless signals with the UE, but the wireless resources are multiplexed in the frequency and time directions, and radio waves are exchanged in time slots with short intervals on the order of microseconds. The DU and CU must process the wireless signals without delay in these time slots, requiring strict delay requirements on the order of microseconds. 【0016】 Furthermore, as shown in the example in Figure 1, the RU is a communication unit (communication antenna) that performs wireless communication with the UE. The UE is a general term for mobile communication terminals (for example, smartphones, tablets, and other devices that users use on a daily basis). The UE connects to a wireless access network using radio waves and transmits and receives data and voice. 【0017】 The Core Network (CN) is a collective term for a group of Network Functions that handle the transfer of data sent and received by the User Environment (UE) and the processing of the UE's communication protocols. The CN is connected to multiple Ranging Areas (RANs) and plays a role in data routing and connecting to other networks (NW) such as the Internet and telephone networks (Figure 2). 【0018】 Here, we will also explain specific examples of the processing performed by RU, DU, and CU. For example, in the 5th generation (5G) cellular standard of 3GPP (3rd Generation Partnership Project), gNB can be divided into L1 layer (physical layer), L2 layer (data link layer), and L3 layer (network layer) as the 5G protocol layer. 【0019】 The RU processes RF (radio frequency) and the lower part of the L1 layer (Low-PHY). 【0020】 The DU and CU perform radio signal processing. Specifically, the DU handles tasks in the upper part of the L1 layer (High-PHY). These tasks include, for example, channel estimation, equalization, modulation and demodulation, and coding (error correction). 【0021】Furthermore, the DU handles L2 layer tasks. These tasks include, for example, media access control (MAC) tasks (e.g., wireless resource scheduling) and wireless link control (RLC) tasks (e.g., packet segmentation and reconstruction). 【0022】 The CU manages PDCP (Packet Data Convergence Protocol), SDAP (Service Data Adaptation Protocol), and RRC (Radio Resource Control) protocol entities. 【0023】 Based on the above, DU performs L1 and L2 processing. CU performs L3 processing. 【0024】 (About AI-RAN) AI-RAN is a technology that enables vRAN applications and AI applications to run on the same virtualization platform. As explained in Figure 1, as RAN is increasingly software-based, general-purpose servers can utilize not only software that implements RAN control functions but also other software. For example, the amount of wireless communication traffic by UE fluctuates greatly depending on the time of day, so there is surplus resources (computing power) on general-purpose servers during off-peak hours such as late at night and early in the morning. The AI-RAN concept utilizes these surplus resources for AI applications. 【0025】 One way to implement AI-RAN is to use a GPU (graphics processing unit) as an accelerator (a function that accelerates application processing) on ​​a general-purpose server. For example, if the computing resources of a general-purpose server (a server equipped with a GPU) that has RAN software installed, which are not normally used, can be utilized for AI processing (for example, the computing processing that AI uses to derive conclusions from new data: AI inference), then this single computing resource (GPU) on the general-purpose server can be given a second revenue source in the form of AI services. 【0026】Figure 2 shows an example of an AI-RAN structure. Figure 2 shows an example where a vRAN application and an AI application are implemented on the same virtualization platform. 【0027】 Furthermore, Figure 2 shows a base station device 100 as an example of a general-purpose server. According to the base station device 100, it is possible to provide a large-capacity, high-performance, and high-quality vRAN at carrier grade on a virtualization platform equipped with GPU (Graphics Processing Unit) computing, and to provide various AI applications such as generative AI on top of it. 【0028】 More specifically, the base station device 100 has a virtualization platform equipped with computing resources such as a GPU, and a RAN control function F1, which corresponds to a vRAN application, and an AI function F2, which corresponds to an AI application, are implemented on the virtualization platform. 【0029】 In the example shown in Figure 2, the RAN control function F1 executes the L1, L2, and L3 processes described in Figure 1. Furthermore, the L1 process may be implemented by L1 software, and the L2 and L3 processes may be implemented by L2 / L3 software. 【0030】 Furthermore, AI processing using AI function F2 includes image diagnosis, voice analysis, specialized LLM, confidential LLM, LLM robots, and LLM autonomous driving. Image diagnosis, voice analysis, specialized LLM, confidential LLM, LLM robots, and LLM autonomous driving may be provided to external users as AI services. 【0031】 As shown in Figure 2, by providing a second revenue stream, such as AI services, to the same GPU, the maintenance and management costs of RAN can be covered by the sales of the AI ​​services, and it may even be possible to generate further profits. 【0032】Furthermore, implementing both the vRAN application and the AI ​​application on the same virtualization platform means implementing the AI ​​on the same platform as the DU (Data Unit). Such AI is also called edge AI. Edge AI has advantages such as being close to the UE (User Environment) that actually receives the signal, resulting in less latency, and being much more responsive than when running the AI ​​on a geographically distant server (for example, a cloud server), because it is located very close to the base station (RU). 【0033】 (Regarding surplus resources) According to AI-RAN, not only can a specific operator (for example, the operator of base station equipment 100) provide the above-mentioned AI services developed by that operator to users, but surplus resources can also be provided to users. As a result, users (for example, corporations) can utilize surplus resources to develop their own AI. 【0034】 Here, we will explain surplus resources using Figure 3. Figure 3 is an explanatory diagram of surplus resources. As mentioned above, wireless communication traffic by UE fluctuates greatly depending on the time of day. For this reason, for example, during times when there are few users, such as from late night to early morning, there is a surplus of computing resources CR (computing capacity) of the base station equipment 100. This surplus relative to the total computing resources CR is called surplus resources SR. By utilizing surplus resources SR, for example, as an AI service by the AI ​​function F2 (i.e., by monetizing it), it becomes possible to cover the costs of the RAN control function F1. 【0035】 Figure 3(a) conceptually illustrates an example in which vRAN applications and AI applications are dynamically deployed on the computing resource CR in response to the user's demand for wireless communication and the user's demand for AI services. The base station device 100 may have the dynamic deployment control function shown in Figure 3(a). 【0036】On the other hand, Figure 3(b) shows an example of dynamic allocation control. According to the example in Figure 3(b), surplus resources SR correspond to the surplus portion of computing resources CR that is not used in vRAN applications, and this surplus can be utilized for AI applications. 【0037】 Specifically, as shown in Figure 3(b), the utilization rate of computing resource CR fluctuates depending on the time of day. For example, traffic decreases during late night and early morning hours when there are fewer users, and therefore the utilization rate of computing resource CR tends to be lower. From this, the portion of computing resource CR used for vRAN applications can be called "RAN resource R1". On the other hand, the surplus portion of computing resource CR that is not used for vRAN applications, known as surplus resource SR, can be utilized as "AI processing resource R2". 【0038】 [2. Proposed Method] Based on the above explanation, the proposed method of this application will now be explained. First, in the base station device 100 (AI-RAN), the RAN control function F1 is the most important function and cannot be stopped. If the RAN control function F1 is stopped, we users will not be able to perform wireless communication via the UE. Thus, wireless signal processing by the RAN control function F1 is basically prioritized over AI processing by the AI ​​function F2, but there are some AI processes that have a high priority and should be completed as early as possible, even if it means reducing the load (utilization rate) on wireless signal processing (even if it means relatively increasing the load on AI processing). 【0039】 However, when a request is received to execute a high-priority AI process, there may not be enough surplus resources SR to perform that AI process. In other words, when a request is received to execute a high-priority AI process, a large portion of the computing resources CR may be used for wireless signal processing by the RAN control function F1 (for example, with a utilization rate of 80% or more), and the surplus resources SR, which correspond to the remaining free area (20% free), may not be able to perform the requested AI process. 【0040】Therefore, the proposed method generates an AI resource AR capable of executing the requested AI processing for the computing resource CS by limiting the wireless resource WR of the RAN control function F1 when it is determined that the requested AI processing cannot be executed. More specifically, the proposed method is characterized by temporarily and forcibly generating AI resources by setting an upper limit on the allocation process so that the load (utilization rate) of the computing resource CS involved in the allocation process of assigning a group of resource blocks RS, each consisting of at least one or more consecutive resource blocks RB on the frequency axis, to the UE does not rise above a certain level (for example, not exceeding 60% utilization rate). 【0041】 For example, L1 processing (channel estimation, equalization, modulation / demodulation, coding, etc.) involves complex mathematical operations and requires high computational density. Furthermore, since L1 processing is a crucial requirement for achieving high throughput and low latency, it is not advisable to restrict L1 processing. On the other hand, in L2 processing, for example, in the allocation control of wireless resources, methods are sometimes used to restrict the allocation of wireless resources to UEs based on the utilization rate of resource block RB or the number of active UEs in a cell, in order to maintain fairness among UEs. For these reasons, there is still room to restrict L2 processing (especially wireless resource allocation control) compared to L1 processing, and the proposed method focuses on this point. 【0042】 According to this proposed method, even when a request to execute a predetermined AI process (for example, a high-priority AI process) is received, the predetermined AI process can be appropriately interrupted without stopping the wireless signal processing by the RAN control function F1. Therefore, according to this proposed method, computing resources CR can be appropriately allocated between the RAN control function F1 and the AI ​​function F2. 【0043】Further, the proposed method is executed by the base station device 100. The predetermined AI processing may be AI processing to which a priority (priority level) higher than a certain value is assigned according to predetermined conditions (for example, importance or urgency as an AI service). The priority may be dynamically assigned by the base station device 100, or may be assigned by the provider of the AI service according to its own rules of thumb. 【0044】 Further, as described above, all or at least a part of the AI processing is sold as an AI service that can be used by a user (an external user such as a corporation), and the surplus resource SR may be released to the user as a service resource for use in the AI service. Therefore, the predetermined AI processing may be AI processing as an AI service used by a billing user who has been billed more than a predetermined amount among users, or AI processing designated by the billing user among the AI processing as an AI service used by the billing user. 【0045】 Further, the proposed method can realize improvement of communication quality and efficiency of operation by utilizing AI, and becomes an innovative technical foundation in the communication business, so it can contribute to the achievement of Goal 9, "Build the infrastructure for industry and technological innovation," of the Sustainable Development Goals (SDGs). 【0046】 [3. Outline of the proposed method] Next, the outline of the proposed method will be described using FIG. 4. FIG. 4 shows an embodiment of the information processing according to the proposed method. In FIG. 4, the information processing according to the embodiment will be described from the relationship between time and usage rate in the computing resource CR. 【0047】 In FIG. 4(a), the actual usage rate fluctuation situation according to the passage of time on a predetermined day, up to the timing when a request to execute the predetermined AI processing is received (in the example of FIG. 4(a), 10:00), is shown by a solid line. Also, in FIG. 4, the predicted usage rate fluctuation situation according to the passage of time after the timing when the execution request is received is shown by a dotted line. The predicted fluctuation situation may be obtained, for example, based on a machine learning model that uses the fluctuation situation in the past as实绩 than the current predetermined day. 【0048】In the information processing according to the embodiment, in the case where the usage rate of the computing resource CR exceeds 80% (threshold value 80%) at the timing when the execution request is received, the base station apparatus 100 determines that a predetermined AI process (hereinafter abbreviated as "target AI process") targeted by the execution request cannot be executed with the surplus resource SR. Specifically, when 80% or more of the computing resource CR is being used as the RAN resource R1 and the remaining capacity as the AI process resource R2 is less than 20%, the base station apparatus 100 determines that the target AI process cannot be executed with the current 20% of the surplus resource SR. Note that the method and criteria for determining whether the target AI process can be executed with the surplus resource SR are not limited to this example. 【0049】 Thus, when the base station apparatus 100 determines that the target AI process cannot be executed with the surplus resource SR, it executes a restriction process of temporarily generating an AI process resource R2 (an example of an AI resource) capable of executing a predetermined AI process by restricting the physical resource used in the L2 process (radio resource allocation control) by the RAN control function F1, that is, the radio resource WR. In short, the restriction process is to temporarily create (secure) a surplus resource SR as a remaining capacity (free area) estimated to be capable of executing a predetermined AI process. 【0050】 A specific example of the restriction process is shown in FIG. 4(b). For example, the base station apparatus 100 may define a predetermined period T according to the timing when the execution request is received. The base station apparatus 100 may define the predetermined period T according to any conditions. For example, the base station apparatus 100 may define the predetermined period T based on the estimated required time from the start to the completion of the target AI process, or the predetermined period T may be predefined for the base station apparatus 100 to use a fixed time length. In the example of FIG. 4(b), the base station apparatus 100 defines the period from the timing (10:00) when the execution request is received to 5 hours later (15:00 in the example of FIG. 4(b)) as the predetermined period T. 【0051】Here, according to the example in Figure 4 (see in particular Figure 4(a)), the utilization rate of the computing resource CR is predicted to fluctuate between 60% and 98% during a predetermined period T. Therefore, the base station device 100 limits the radio resource WR to ensure that the load (utilization rate) of radio signal processing on the computing resource CR does not exceed 60% during the predetermined period T, based on the lowest utilization rate of 60% out of the range of 60% to 98%. Specifically, the base station device 100 temporarily generates the AI ​​processing resource R2 only during the predetermined period T by setting an upper limit on the resource block RB so that the utilization rate of the RAN resource R1 on the computing resource CR does not exceed 60%. That is, as shown in Figure 4(b), the base station device 100 generates 40% of the AI ​​processing resource R2 on the computing resource CR (securing 40% of the surplus resource SR). 【0052】 This type of limitation processing is based on the idea that the minimum RAN resource R1 required for wireless signal processing by the RAN control function F1 is 60% of the total computing resource CR during a predetermined period T, and that wireless signal processing can be performed even if the speed is somewhat slowed down as long as this 60% of the RAN resource R1 is available. 【0053】 [4. Configuration of the Base Station Device] The base station device 100 according to the embodiment will be described using Figure 5. Figure 5 is a diagram showing an example of the configuration of the base station device 100 according to the embodiment. As shown in Figure 5, the base station device 100 has a communication unit 110, a storage unit 120, and a control unit 130. As described above, the base station device 100 is an AI-RAN in which both the RAN control function F1 and the AI ​​function F2 are realized by software on the same virtualization platform on a general-purpose server. 【0054】 (Communication Unit 110) The communication unit 110 is implemented by, for example, a NIC (Network Interface Card). For example, it transmits and receives information with the outside world. 【0055】(Storage Unit 120) The storage unit 120 is implemented by, for example, a semiconductor memory element such as RAM (Random Access Memory) or flash memory, or a storage device such as a hard disk or optical disc. The storage unit 120 may store, for example, data and programs related to information processing according to the embodiment. Also, as shown in Figure 5, the storage unit 120 may have an AI processing-related information storage unit 121, a first model storage unit 122, and a second model storage unit 123. 【0056】 (AI Processing Related Information Storage Unit 121) Here, Figure 6 shows an example of the AI ​​processing related information storage unit 121. In the example in Figure 6, the AI ​​processing related information storage unit 121 may be divided into a storage unit 121-1 that stores information about the AI ​​service and a storage unit 121-2 that stores information about the user who uses the AI ​​service. 【0057】 (Memory Unit 121-1) Memory unit 121-1 contains items such as "service details," "AI model," and "priority." 【0058】 "Service Details" describes the specific content of the AI ​​service implemented through AI processing using the "AI Model." The "AI Model" is a machine learning model used for AI processing as an AI service. 【0059】 "Priority" is assigned to the AI ​​processing implemented in the "AI model" and is a trigger that serves as a criterion for deciding whether or not to perform the control processing described above. When the base station device 100 receives a request to execute an AI processing, if the priority assigned to that AI processing is "1", it may determine whether or not the AI ​​processing can be executed with the current surplus resource SR. 【0060】 (Storage Unit 121-2) Storage unit 121-2 has items such as "User ID", "Amount Charged", and "Services Used". 【0061】 "User ID" is identification information that identifies users of the AI ​​service. "Billing Amount" indicates the amount that users of the AI ​​service pay to use that service (for example, the monthly billing amount). 【0062】"Services Used" refers to the content of the AI ​​service that a user of the AI ​​service is using. 【0063】 (Control Unit 130) Returning to Figure 5, the control unit 130 is realized by a CPU (Central Processing Unit) or MPU (Micro Processing Unit), etc., executing various programs (for example, the information processing program according to the embodiment) stored in the memory device inside the base station device 100 using RAM as the working area. The control unit 130 is also realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array). 【0064】 As shown in Figure 5, the control unit 130 includes a reception unit 131, a determination unit 132, a resource control unit 133, an execution control unit 134, a prediction unit 135, a generation unit 136, and an estimation unit 137, and realizes or executes the information processing functions and operations described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in Figure 5, and other configurations are also possible as long as they perform the information processing described later. Also, the connection relationships of the various processing units in the control unit 130 are not limited to the connection relationships shown in Figure 5, and other connection relationships are also possible. 【0065】 (Reception Unit 131) The reception unit 131 receives requests to execute AI processing. For example, when the reception unit 131 receives a request to execute AI processing, it may determine whether or not the AI ​​processing has a trigger for starting control processing. 【0066】For example, the reception unit 131 may determine, based on the storage unit 121-1, whether the AI ​​process for which execution has been requested is an AI process that has been assigned a priority of "1" according to predetermined conditions. The reception unit 131 may also determine, based on the storage unit 121-2, whether the AI ​​process for which execution has been requested is an AI process as an AI service used by a paying user who has been charged more than a predetermined amount, or an AI process designated by a paying user. If the AI ​​process has a trigger for starting control processing, the reception unit 131 formally accepts the execution request. 【0067】 (Determination Unit 132) When the Reception Unit 131 receives an execution request, the Determination Unit 132 determines whether the target AI process for which the execution request has been formally received (the AI ​​process that has a trigger for the start of control processing) can be executed with the current surplus resources SR. 【0068】 For example, the determination unit 132 may determine whether the target AI processing can be executed with the current surplus resource SR based on whether the utilization rate of the computing resource CR exceeds 80% (threshold 80%) at the time the execution request is received. For example, if the utilization rate of the computing resource CR exceeds 80%, the determination unit 132 may determine that the target AI processing cannot be executed with the current surplus resource SR and proceed to control processing. On the other hand, if the utilization rate of the computing resource CR does not exceed 80%, the determination unit 132 may determine that the target AI processing can be executed with the current surplus resource SR and perform the normal processing of executing the target AI processing with the current surplus resource SR. 【0069】 (Resource Control Unit 133) If the Resource Control Unit 133 determines that it is not possible to execute the target AI processing with surplus resources, it limits the wireless resource WR of the RAN control function F1, thereby generating an AI processing resource R2 (AI resource) capable of executing the target AI processing for the computing resource CR. 【0070】For example, the resource control unit 133 generates an AI processing resource R2 capable of executing the target AI processing for the computing resource CR by limiting the wireless resources used in the F2 processing of the wireless signal processing. Specifically, the resource control unit 133 generates the AI ​​processing resource R2 by setting an upper limit on the allocation process (e.g., the number of allocation processes) so that the load on computing resources for the allocation process that assigns a group of resource blocks RB, each consisting of at least one or more consecutive resource blocks RB on the frequency axis, to the UE does not exceed a certain level. For example, the resource control unit 133 temporarily generates the AI ​​processing resource R2 for a predetermined period T by setting an upper limit on resource blocks RB so that it does not exceed a predetermined proportion of the computing resource CR occupied by the RAN resource R1. 【0071】 The resource control unit 133 may also perform a process to determine a predetermined period T according to the timing at which the execution request is received. 【0072】 (Execution Control Unit 134) The execution control unit 134 causes the target AI processing to be executed using the AI ​​processing resource R2. The execution control unit 134 may also release the restriction on the wireless resource WR once the target AI processing executed using the AI ​​processing resource R2 has finished. For example, the execution control unit 134 may release the generated AI processing resource R2 as an unused computing resource CR, thereby controlling the computing resource CR to be available for use as the RAN resource R1. 【0073】 (Prediction Unit 135) The base station device 100 may have a first model M1 that predicts traffic in the cells covered by the radio unit, i.e., the RU. The first model M1 may be implemented as a human flow prediction model that has been learned based on human flow statistics data in the cells covered by the RU, weather data in the cells covered by the RU, event information in the cells covered by the RU, etc. 【0074】Therefore, the prediction unit 135 may use the first model M1 to predict how much traffic will occur in the cells covered by the RU during a predetermined period T corresponding to the timing when the execution request is received. For example, the prediction unit 135 may take timing information (e.g., a predetermined period T) as input and predict the traffic that may occur in the cells covered by the RU based on the output result output to the first model M1. 【0075】 (Generation Unit 136) The generation unit 136 generates a second model M2. Specifically, the generation unit 136 generates a second model M2 that estimates the fluctuation status of the allocation process at a predetermined timing from the traffic prediction result at a predetermined timing, based on the relationship between the traffic history and the allocation process history in the cells covered by the RU. For example, the generation unit 136 uses the combination of the traffic history and the allocation process history as training data DA to generate a second model M2 that has been trained to output the fluctuation status of the allocation process at a predetermined timing when the traffic prediction result at a predetermined timing is input. 【0076】 (Estimation Unit 137) The estimation unit 137 estimates the fluctuation status of the allocation process over a predetermined period T based on the traffic prediction results from the prediction unit 135. For example, the estimation unit 137 estimates the fluctuation status of the allocation process over a predetermined period T based on the traffic prediction results for the predetermined period T and the second model M2. 【0077】 [5. Example of Base Station Equipment Operation] Next, an example of the operation of the base station equipment 100 will be explained using Figure 7. Figure 7 is a diagram showing an example of the operation of the base station equipment 100. 【0078】 In the example shown in Figure 7, the base station device 100 has the first model M1 in the first model storage unit 122. In the example shown in Figure 7, the base station device 100 has already generated the second model M2 during the learning phase. 【0079】As shown in the example in Figure 7, the generation unit 136 may generate a second model M2 by using a pair of traffic data history and allocation processing data history, which are linked to date and time data, as training data DA, and training the model on the relationship between traffic data and allocation processing data. 【0080】 Here, traffic data refers to actual traffic data generated in the cells covered by the RU at a given date and time. 【0081】 Allocation processing data refers to scheduling that was actually performed in accordance with the traffic conditions indicated by the actual traffic data (for example, the allocation processing that allocates downlink data channels to UEs, which may be the utilization rate of the computing resources CR actually used). Thus, allocation processing data may be actual data on the utilization rate of computing resources CR, but it may also include actual data on the number of allocation processes performed. 【0082】 Therefore, according to the example in Figure 7, the second model M2 may be a model that outputs the utilization rate of the computing resource CR used for the allocation process to process the traffic at a predetermined timing when traffic information is input at that predetermined timing. 【0083】 Based on the above, an example of the operation of the base station device 100 will be described. Figure 7 shows a scene in which the control processing according to the embodiment is executed because it has been determined that the target AI processing cannot be performed with the current surplus resources SR. 【0084】 First, the prediction unit 135 inputs a predetermined period T as timing information to the first model M1 (step S71), and from the output result of the first model M1, predicts the traffic that may occur in the cells covered by the RU during the predetermined period T (step S72). Figure 7 shows an example in which, following the example in Figure 4, the prediction unit 135 inputs "10:00 to 15:00" as the predetermined period T, and predicts "TR#11" as the traffic during the predetermined period T based on the output result of the first model M1. 【0085】Next, the estimation unit 137 inputs the prediction results information from the prediction unit 135 into the second model M2 (step S73). Specifically, the estimation unit 137 inputs the predetermined period T "10:00 to 15:00" and the traffic "TR#11" during the predetermined period T into the second model M2 as prediction results information from the prediction unit 135. 【0086】 Then, the estimation unit 137 estimates the fluctuation status of the allocation process over a predetermined period T from the output result of the first model M1 (step S74). Specifically, the estimation unit 137 estimates how the utilization rate of the computing resource CR used for the allocation process will fluctuate for the traffic "TR#11" predicted in the cell during the predetermined period T "10:00 to 15:00". In other words, the estimation unit 137 may estimate the fluctuation status of the utilization rate of the computing resource CR used for the allocation process as the fluctuation status of the allocation process. To put it another way, the estimation unit 137 may estimate the fluctuation status of the ratio that the RAN resource R1 occupies to the total computing resource CR. Figure 7 shows an example in which the estimation unit 137 estimates "60% to 98%" as the fluctuation status of the utilization rate, following the example in Figure 4. 【0087】 In this state, the resource control unit 133, based on the usage rate fluctuation status of "60% to 98%", executes a process to set an upper limit for the resource block RB so that the usage rate of the RAN resource R1 with respect to the computing resource CR does not exceed 60%. 【0088】 For example, the resource control unit 133 refers to the conversion table TB and converts the usage rate of the RAN resource R1 with respect to the computing resource CR to 60% into the number of allocation processes (step S75). Specifically, the resource control unit 133 refers to the conversion table TB and performs a conversion process to determine how to control the number of allocation processes in order to keep the usage rate of the RAN resource R1 with respect to the computing resource CR below 60%. The conversion table TB may be generated in advance from statistical data on the relationship between usage rate and allocation processes. 【0089】Controlling the number of allocation processes is equivalent to controlling the resource block RB. Therefore, the resource control unit 133 may set an upper limit on the resource block RB so that the number of allocation processes is the number of processes specified in step S75 (step S76). Setting an upper limit on the resource block RB may, for example, mean setting an upper limit on the number of scheduling information, which is information that allocates the resource block RB to each of the following UEs, or on the number of control signals PDCCH (Physical Downlink Control Channel) that notify the UEs of this scheduling information (or the number of bits). As another example, the resource control unit 133 may also exclude priority bandwidth from allocation targets or reduce transmission power. 【0090】 Then, the resource control unit 133 controls, for example, the RAN control function F1, so that it does not perform allocation processing beyond the upper limit set in step S76 during a predetermined period T. As a result, as shown in Figure 4(b), 40% of the computing resource CR is temporarily generated as AI processing resource R2. 【0091】 Although not shown in Figure 7, the execution control unit 134 executes the target AI processing using the AI ​​processing resource R2 when it is generated. Furthermore, when the target AI processing is completed, or when a predetermined period T has elapsed, the execution control unit 134 may release the generated AI processing resource R2 as unused computing resource CR. The released computing resource CR becomes available for use as RAN resource R1. 【0092】 [6. Information Processing Procedure] The information processing procedure according to the embodiment will be explained using Figure 8. Figure 8 is a flowchart of the information processing procedure performed by the base station device 100. 【0093】 The reception unit 131 determines whether or not it has received an execution request for AI processing (step S801). If the reception unit 131 has not received an execution request (step S801; No), it waits until it receives an execution request. 【0094】 On the other hand, if the reception unit 131 receives an execution request (step S801; Yes), it determines whether the AI ​​process has a trigger for starting the control process (step S802). 【0095】 If the determination unit 132 determines that the AI ​​process has a trigger for starting a control process (step S802; Yes), it determines whether the target AI process (the AI ​​process having a trigger for starting a control process) can be executed with the current surplus resources SR (step S803). 【0096】 If the determination unit 132 determines that the target AI processing can be executed with the current surplus resources SR (step S803; Yes), it executes the normal process of executing the target AI processing with the current surplus resources SR (step S804). 【0097】 On the other hand, if the determination unit 132 determines that it is not possible to execute the target AI processing with the current surplus resources SR (step S803; No), it causes the control processing according to the embodiment to be executed (step S805). The procedure for the control processing according to the embodiment is as described in Figure 7. 【0098】 Furthermore, if the execution control unit 134 determines that the AI ​​process does not have a trigger for starting the control process (step S802; No), it decides that there is no need to complete the AI ​​process early and executes the AI ​​process at a time when sufficient surplus resources SR can be secured (for example, during late-night or early-morning hours) (step S806). 【0099】 [7. Hardware Configuration] The base station device 100 according to the embodiment may be implemented by a computer 1000 having the configuration shown in Figure 9. Figure 9 is a hardware configuration diagram showing an example of a computer that implements the functions of the base station device 100 according to the embodiment. The computer 1000 has a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700. 【0100】 The CPU 1100 operates based on programs stored in the ROM 1300 or HDD 1400, and controls various parts. The ROM 1300 stores boot programs executed by the CPU 1100 when the computer 1000 starts up, as well as programs that depend on the computer 1000's hardware. 【0101】 The HDD 1400 stores programs executed by the CPU 1100, and data used by such programs. The communication interface 1500 receives data from other devices via a predetermined communication network and sends it to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined communication network. 【0102】 The CPU 1100 controls output devices such as displays and input devices such as keyboards via the input / output interface 1600. The CPU 1100 acquires data from input devices via the input / output interface 1600. The CPU 1100 also outputs the generated data to output devices via the input / output interface 1600. 【0103】 The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1200 via the media interface 1700 and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase Change Rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. 【0104】For example, when the computer 1000 functions as a base station device 100 according to the embodiment, the CPU 1100 of the computer 1000 realizes the functions of the control unit 130 by executing a program loaded on the RAM 1200. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, these programs may be obtained from other devices via a predetermined communication network. 【0105】 [8. Others] Furthermore, all or part of the processes described as being performed automatically in each of the above embodiments may be performed manually, or all or part of the processes described as being performed manually may be performed automatically by known methods. In addition, the processing procedures, specific names, and information including various data and parameters shown in the above documents and drawings may be changed at will unless otherwise specified. For example, the various information shown in each figure is not limited to the information shown. 【0106】 Furthermore, the components of each illustrated device are functionally conceptual and do not necessarily need to be physically configured as shown. In other words, the specific forms of distribution and integration of each device are not limited to those shown, and all or part of them can be functionally or physically distributed and integrated in any unit according to various loads and usage conditions. 【0107】 Furthermore, the above embodiments can be combined as appropriate, provided that the processing content is not contradictory. 【0108】 Although some embodiments of the present application have been described in detail above with reference to the drawings, these are illustrative examples, and the present invention can be implemented in various other forms with modifications and improvements based on the knowledge of those skilled in the art, including the embodiments described in the section on the present invention. 【0109】100 Base station device 120 Storage unit 121 AI processing related information storage unit 122 First model storage unit 123 Second model storage unit 130 Control unit 131 Reception unit 132 Determination unit 133 Resource control unit 134 Execution control unit 135 Prediction unit 136 Generation unit 137 Estimation unit

Claims

1. A base station device in which both a RAN control function for controlling a wireless access network (RAN) and an AI function having artificial intelligence (AI) are implemented on a server by software, comprising: a determination unit that, when an execution request for a predetermined AI process is received, determines whether the predetermined AI process can be executed using surplus resources of the server's computing resources that are not used for the execution of the RAN control function; a resource control unit that, if it is determined that the predetermined AI process cannot be executed using the surplus resources, generates AI resources capable of executing the predetermined AI process for the computing resources by restricting the wireless resources of the RAN control function; and an execution control unit that causes the predetermined AI process to be executed using the AI ​​resources.

2. The base station device according to claim 1, wherein the predetermined AI processing is an AI processing to which priority is assigned according to predetermined conditions, and when the determination unit receives an execution request for the AI ​​processing to which priority is assigned as the predetermined AI processing, it determines whether or not the AI ​​processing to which priority is assigned can be executed with the surplus resources.

3. The base station device according to claim 1, wherein at least a portion of the AI ​​processing is sold as an AI service available to external users, the surplus resources are made available to the external users as resources for use in the AI ​​service, and the determination unit determines whether or not the AI ​​processing corresponding to the billing user can be executed with the surplus resources as the predetermined AI processing when the execution request is received from a billing user among the external users who has charged more than a predetermined amount.

4. The base station device according to claim 1, wherein the resource control unit generates the AI ​​resources by setting an upper limit on the allocation process so as not to exceed a certain level when allocating a group of resource blocks, each consisting of at least one resource block in a row on the frequency axis, to a terminal device.

5. The base station device is associated with a wireless unit that performs wireless communication with the terminal device, and comprises a first model for predicting traffic in the cell covered by the wireless unit, a prediction unit that uses the first model to predict the traffic for a predetermined period corresponding to the timing when the execution request is received, and an estimation unit that estimates the fluctuation status of the allocation process for the predetermined period based on the traffic prediction result, and the resource control unit sets an upper limit on the allocation process based on the estimated fluctuation status to prevent the allocation process from being executed beyond the upper limit, as described in claim 4.

6. The base station device according to claim 5, further comprising: a generation unit that generates a second model for estimating the fluctuation status of the allocation process at a predetermined timing from the traffic prediction result at a predetermined timing based on the relationship between the traffic history in the cell covered by the wireless unit and the allocation process history, wherein the estimation unit estimates the fluctuation status of the allocation process during the predetermined period based on the traffic prediction result during the predetermined period corresponding to the timing at which the execution request was received and the second model.

7. The base station device according to claim 6, wherein the generation unit generates a second model that has been trained to output the fluctuation status of the allocation process at a predetermined timing when the prediction result of the traffic at a predetermined timing is input, using the combination of the traffic history and the allocation process history as training data.

8. The base station device according to claim 1, wherein the execution control unit releases the restriction on the wireless resource when the AI ​​processing performed on the AI ​​resource is completed.

9. An information processing method to be performed by a base station device in which both a RAN control function for controlling a wireless access network (RAN) and an AI function having artificial intelligence (AI) are implemented on a server by software, the method comprising: a determination step of determining whether the predetermined AI process can be executed using surplus resources of the server's computing resources that are not used for the execution of the RAN control function when a request for execution of a predetermined AI process among the AI ​​processes executed by the AI ​​function is received; a resource control step of generating AI resources capable of executing the predetermined AI process for the computing resources by restricting the wireless resources of the RAN control function if it is determined that the predetermined AI process cannot be executed with the surplus resources; and an execution control step of causing the predetermined AI process to be executed using the AI ​​resources.

10. An information processing program executed by a base station device in which both a RAN control function for controlling a wireless access network (RAN) and an AI function having artificial intelligence (AI) are implemented on a server by software, the information processing program causing the base station device to execute: a determination procedure for determining whether the predetermined AI process can be executed using surplus resources of the server's computing resources that are not used for the execution of the RAN control function when a request for execution of a predetermined AI process among the AI ​​processes executed by the AI ​​function is received; a resource control procedure for generating AI resources capable of executing the predetermined AI process for the computing resources by restricting the wireless resources of the RAN control function if it is determined that the predetermined AI process cannot be executed with the surplus resources; and an execution control procedure for causing the predetermined AI process to be executed using the AI ​​resources.

Images