System, program, and control method

Abstract

Provided is a system comprising: a resource allocation unit that, on the basis of the result of predicting the demand for a RAN, determines the amount of computing resources of one execution platform to be allocated to a RAN control function for controlling the RAN, and the amount of computing resources of the one execution platform to be allocated to an AI processing function for performing AI processing (RAN control AI processing such as a RAN intelligent controller (RIC), a non-RAN control AI in MEC, or the like), and allocates the amounts of computing resources to the RAN control function and the AI processing function; and an execution unit that executes the RAN control function and the AI processing function on the one execution platform according to the amounts of computing resources allocated by the resource allocation unit.

Description

System, program, and control method

[0001] The present invention relates to a system, a program, and a control method.

[0002] Patent Document 1 describes a radio access network control device that includes a processor configured by a Non-RT RIC (Non-Real Time RAN (Radio Access Network) Intelligent Controller), which acquires the operation status of a plurality of radio access network nodes from a virtualization platform that virtually manages a set of the plurality of radio access network nodes, acquires measured operation data from each radio access network node, and issues operation guidelines regarding the operation of each radio access network node to at least one of the virtualization platform and each radio access network node based on the operation status and operation data. [Prior Art Documents] [Patent Documents] [Patent Document 1] International Publication No. 2023-100385

[0003] Previously, there were technologies that allowed individual RAN services or AI (Artificial Intelligence) applications to run on an execution platform. Mobile networks are now approaching infrastructures such as water and electricity, and service outages would have a significant impact on people's lives. Therefore, conventional RAN base stations must achieve very high service level agreements (SLAs) and must create mechanisms, such as redundancy, to prevent service outages. As a result, only a few percent of the computing resources on the execution platform are being used. In other words, many computing resources are unused. Furthermore, at night, despite a decrease in user demand, computing resources are wasted in an attempt to maintain the same level of service as during the daytime. In other words, the system is designed to accommodate peak demand. In response to this, some advanced technologies have been implemented, such as powering down some computer resources at night to save power, but these systems still waste a significant amount of machine power.

[0004] In the system according to this embodiment, for example, by running RAN functions on a high-performance GPU (Graphics Processing Unit) server rather than on a general-purpose server, the surplus computing resources can be utilized for AI processing. Types of AI processing include AI processing related to RAN control (sometimes referred to as RAN control AI processing) and AI processing not related to RAN control (sometimes referred to as non-RAN control AI processing).

[0005] An example of RAN control AI processing is RIC. RIC is a technology that uses AI to optimize RAN radio resources and automate RAN operations. RIC includes Non-RT RIC and Near-RT RIC (Near-Real Time RIC). Non-RT RIC is sometimes called Centralized RIC. Non-RT RIC is located inside SMO (Service Management and Orchestration), which manages and orchestrates the RAN. Non-RT RIC generates and notifies policies related to RAN control and transmits information to Near-RT RIC. For example, the Non-RT RIC performs machine learning using data collected from the RAN to generate a trained model for RAN control and transmits it to the Near-RT RIC. The Near-RT RIC is sometimes called a Distributed RIC. Compared to the Non-RT RIC, the Near-RT RIC is located closer to the RAN nodes (RU (Radio Unit), DU (Distributed Unit), CU (Central Unit)) and controls the RAN nodes, resources, etc. The Near-RT RIC performs processing with higher real-time performance than the Non-RT RIC. The Near-RT RIC performs inference processing related to RAN control using, for example, a trained model acquired from the Non-RT RIC. RAN control AI processing is not limited to the RIC.

[0006] The non-RAN control AI processing may correspond to a so-called MEC (Multi-access Edge Computing) application. Examples of the non-RAN control AI processing include a monitoring AI execution process that determines the situation within the imaging range of an input captured image, and a response AI execution process that outputs a response to an input user inquiry. However, this is not limited to these.

[0007] Even when surplus computing resources can be utilized for AI processing, the concept of SLA as described above still exists. For example, mobile network services have very high SLAs, such as 99.9999. However, MEC applications differ from mobile network services in that, considering user fees and service improvement cycles, low execution costs and fast release cycles are more important than high SLAs. The system according to this embodiment contributes to providing a mechanism that enables efficient operation of services with different SLAs and usage policies on the same platform.

[0008] According to one embodiment of the present invention, there is provided a system. The system may include a resource allocation unit that determines, based on a prediction result of RAN demand, an amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN and an amount of computational resources of the execution platform to be allocated to an AI processing function that performs AI processing, and allocates the resources to the RAN control function and the AI ​​processing function. The system may include an execution unit that executes the RAN control function and the AI ​​processing function on the execution platform in accordance with the amount of computational resources allocated by the resource allocation unit.

[0009] The system may further include a demand prediction unit that estimates people flow for each of a plurality of coverage areas constituting a RAN using a people flow estimation model that estimates people flow within the coverage area and predicts demand for the RAN based on the estimation result, and the resource allocation unit may determine an amount of computational resources of the system to be allocated to the RAN control function and an amount of computational resources of the system to be allocated to the AI ​​processing function based on the prediction result of the demand for the RAN by the demand prediction unit. The system may further include a model generation unit that generates the people flow estimation model by performing machine learning using people flow data indicating people flow within the coverage area of ​​a radio base station. The model generation unit may use normal time people flow data indicating people flow by date and time, event time people flow data indicating people flow when an event occurs within the coverage area of ​​the radio base station, and disaster time people flow data indicating people flow when a disaster occurs within the coverage area of ​​the radio base station to generate the people flow estimation model, with date and time, events, and disasters as inputs and people flows as outputs.

[0010] Any of the above systems may further include a classifier that classifies a plurality of wireless base stations into a plurality of clusters, and the model generator may generate the people flow estimation model for each of the plurality of clusters by performing machine learning using the people flow data within the coverage areas of a plurality of wireless base stations included in the cluster. The classifier may collect, for each of the plurality of wireless base stations, posted data posted by user terminals present in the coverage area of ​​the wireless base station, generate base station vector data by embedding the collected plurality of posted data, and classify the plurality of wireless base stations into the plurality of clusters based on the base station vector data of the plurality of wireless base stations.

[0011]

[0013] Any of the systems may further include a resource management unit that divides and manages the computational resources on the one execution platform into a RAN occupied area available only to the RAN control function, an AI occupied area available only to the AI ​​processing function, and a spot area available to both the RAN control function and the AI ​​processing function, and the resource allocation unit may preferentially allocate the spot area to the RAN control function when the amount of computational resources required by the RAN control function is greater than the RAN occupied area. When at least a portion of the spot area is not allocated to the RAN control function, the resource allocation unit may determine, according to demand for the AI ​​processing, whether to allocate the portion of the spot area not allocated to the RAN control function to the AI ​​processing function or transition hardware corresponding to the portion of the spot area not allocated to the RAN control function to a power saving state. The system may further include a state control unit that, when it is determined to transition hardware corresponding to a portion of the spot area not assigned to the RAN control function to a power saving state, transitions the hardware corresponding to the portion of the spot area not assigned to the RAN control function to a different power saving state depending on the allocation rate of the RAN occupied area. The state control unit may control to turn off the power of the hardware corresponding to the portion of the spot area not assigned to the RAN control function when the allocation rate of the RAN occupied area is lower than a predetermined first threshold, and may control to transition the hardware corresponding to the portion of the spot area not assigned to the RAN control function to a sleep state when the allocation rate of the RAN occupied area is higher than the first threshold and lower than a second threshold higher than the first threshold.

[0012] According to one embodiment of the present invention, a program is provided for causing a computer to execute a determination step of determining, based on the results of predicting RAN demand, the amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN and the amount of computational resources of the execution platform to be allocated to an AI processing function that performs AI processing, and an execution step of executing the RAN control function and the AI ​​processing function on the execution platform in accordance with the amount of computational resources determined in the determination step.

[0013] According to one embodiment of the present invention, there is provided a control method executed by a computer. The control method may include a determination step of determining, based on a result of predicting RAN demand, an amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN and an amount of computational resources of the execution platform to be allocated to an AI processing function that performs AI processing. The control method may include an execution step of executing the RAN control function and the AI ​​processing function on the execution platform in accordance with the amount of computational resources determined in the determination step.

[0014] The above summary of the invention does not list all of the necessary features of the present invention, and subcombinations of these features may also constitute inventions.

[0015] FIG. 1 is a schematic diagram illustrating an example of a system 10. FIG. 1 is a schematic diagram illustrating an example of a configuration of a distributed infrastructure 200. FIG. 2 is an explanatory diagram illustrating computational resources 210 used by the distributed infrastructure 200 for the vRAN 203 and the AI ​​processing 204. FIG. 3 is an explanatory diagram illustrating an example of generation of a people flow estimation model 330. FIG. 4 is an explanatory diagram illustrating an example of generation of a people flow estimation model 330. FIG. 5 is an explanatory diagram illustrating management of computational resources 210 of the distributed infrastructure 200. FIG. 6 is an explanatory diagram illustrating allocation of computational resources 210 of the distributed infrastructure 200. FIG. 7 is an explanatory diagram illustrating allocation of computational resources 210 of the distributed infrastructure 200. FIG. 8 is an explanatory diagram illustrating a power saving function of the distributed infrastructure 200. FIG. 9 is an explanatory diagram illustrating the power saving function of the distributed infrastructure 200. FIG. 10 is an explanatory diagram illustrating an example of a functional configuration of the distributed infrastructure 200. FIG. 11 is an explanatory diagram illustrating an architecture 400 in the system 10. 1 shows an example of a hardware configuration of a computer 1200 that functions as the management infrastructure 100 or the distribution infrastructure 200.

[0016] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention as claimed. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0017] 1 schematically illustrates an example of a system 10. The system 10 includes a distributed infrastructure 200. The system 10 may include multiple distributed infrastructures 200. The system 10 may include a management infrastructure 100 that manages the multiple distributed infrastructures 200. In the system 10 according to this embodiment, for example, the management infrastructure 100 and the multiple distributed infrastructures 200 may cooperate to control the RAN 310 and perform AI processing.

[0018] The RAN 310 may be a virtualized vRAN (Virtual RAN), and the system 10 may control the vRAN. The RAN 310 may be a physical RAN, and the system 10 may control the physical RAN. In this embodiment, a case where the RAN 310 is a vRAN will be mainly described as an example.

[0019] The AI ​​processing performed by the system 10 may include RAN-controlled AI processing (sometimes referred to as RAN_AI). The AI ​​processing performed by the system 10 may include non-RAN-controlled AI processing (sometimes referred to as non-RAN_AI).

[0020] The distributed infrastructure 200 may be a data center located in various locations. The distributed infrastructure 200 may be configured with multiple devices. The distributed infrastructure 200 may be realized on a virtualization platform made up of multiple devices. The distributed infrastructure 200 may also be realized by a single device. In other words, the distributed infrastructure 200 may be a distributed device.

[0021] The management infrastructure 100 may be a data center that manages multiple distributed infrastructures 200. The management infrastructure 100 may be configured with multiple devices. The management infrastructure 100 may be realized on a virtualization infrastructure made up of multiple devices. The management infrastructure 100 may also be realized by a single device. In other words, the management infrastructure 100 may be a management device.

[0022] The management infrastructure 100 may be referred to as a Core Brain, and the distributed infrastructure 200 may be referred to as a Regional Brain. While FIG. 1 illustrates an example in which a single-level management infrastructure 100 is arranged below the management infrastructure 100, this is not limiting. The distributed infrastructure 200 may have multiple levels. For example, when a two-level distributed infrastructure 200 is arranged below the management infrastructure 100, the management infrastructure 100 may be referred to as a Core Brain, the distributed infrastructure 200 at the level below that may be referred to as a Regional Brain, and the distributed infrastructure 200 at the level further below that may be referred to as a Sub-Regional Brain.

[0023] The distributed infrastructure 200 may be provided with one or more central processing units (CPUs). The distributed infrastructure 200 may be provided with one or more GPUs. The distributed infrastructure 200 may be provided with multiple super chips, each of which has a CPU and a GPU connected via an interconnect. The interconnect may have memory consistency and may be capable of achieving high bandwidth and low latency. In this way, the distributed infrastructure 200 may have CPU resources and GPU resources as computational resources.

[0024] 2 schematically illustrates an example of the configuration of a distributed infrastructure 200. The distributed infrastructure 200 controls a vRAN 203 and executes AI processing 204 on a virtualization infrastructure 202 implemented on HW (Hardware) 201. The HW 201 may include a CPU, a GPU, memory, a hard disk, an SSD, etc. The control of the vRAN 203 includes control of multiple radio base stations 300 constituting the RAN 310 and cooperation with other distributed infrastructures 200. The AI ​​processing 204 may include a RAN_AI 205 and a non-RAN_AI 206. The distributed infrastructure 200 executes the vRAN 203 and the AI ​​processing 204 using computational resources available for the vRAN 203 and the AI ​​processing 204.

[0025] 3 is an explanatory diagram for explaining the computational resources 210 that the distributed infrastructure 200 uses for the vRAN 203 and the AI ​​processing 204. The distributed infrastructure 200 according to this embodiment may predict demand for the RAN 310 that it manages and determine the amount of computational resources (sometimes referred to as the amount of computational resources for vRAN) to be used for the vRAN 203 among the computational resources 210. After determining the amount of computational resources for vRAN, if there is a surplus in the computational resources 210, the distributed infrastructure 200 may use the surplus for the AI ​​processing 204. This allows the service of the RAN 310 with a high SLA to be prioritized, and when there is still capacity available even after executing the control of the RAN 310, it is possible to further execute AI processing.

[0026] The distributed infrastructure 200 may predict the demand for the RAN 310 based on the usage history of the RAN 310. For example, if the usage amount of the RAN 310 is on the rise, the distributed infrastructure 200 predicts that the demand for the RAN 310 will increase, and if the usage fee for the RAN 310 is on the decline, the distributed infrastructure 200 predicts that the demand for the RAN 310 will decrease.

[0027] The distributed infrastructure 200 may predict demand for the RAN 310 by predicting demand for each of the multiple radio base stations 300 that make up the RAN 310 and integrating the predicted demands of the multiple base stations. Since changes in demand for the RAN 310 are strongly correlated with people who are moving and moving objects such as automobiles driven by people, it is possible to predict demand for the RAN by understanding people's movements.

[0028] Human movement is periodic. For example, in commuter towns, people leave home in the morning, go to work or school, and return home in the evening or at night. Even in areas other than commuter towns, each region has its own unique movement characteristics. Regarding the amount of movement, for example, in Japan, particularly as an island nation, where increases and decreases from airplanes and boats have only a minor impact, the population movement is relatively consistent, with population decreasing in some areas and increasing in others. By combining this with day of the week, season, and artificial trends, a human flow prediction model can be created. However, irregular human flow may occur due to events, disasters, etc. The distributed platform 200 creates a mechanism that can predict future demand for highly correlated RANs by constructing a human flow model using, for example, periodic human flow data based on time-series data on day of the week, hourly, and population trends, human flow data at surrounding event venues and events such as fireworks displays, and change data in short batch windows in the event of a disaster, etc.

[0029] 4 is an explanatory diagram for describing an example of generating the people flow estimation model 330. The distributed platform 200 may acquire people flow data 320 in the coverage area 302 for each of the multiple radio base stations 300, and generate the people flow estimation model 330 using the people flow data 320.

[0030] The people flow data 320 may include normal time people flow data 322 indicating the flow of people by date and time within the coverage area 302 during normal times when no particular events or the like are occurring. The normal time people flow data 322 may include the number of people per time unit at each location in the coverage area 302. The time unit may be seconds, minutes, hours, etc. The normal time people flow data 322 may further include the direction of movement of people, movement speed, etc.

[0031] The people flow data 320 may include event-time people flow data 324 indicating the flow of people when an event occurs within the coverage area 302. The event-time people flow data 324 may include the number of people per time unit at each location in the coverage area 302 when an event such as a concert or fireworks display occurs. The time unit may be seconds, minutes, hours, etc. The event-time people flow data 324 may further include the movement direction and movement speed of people. The people flow data 320 may include event-time people flow data 324 for each type of event.

[0032] The people flow data 320 may include disaster people flow data 326 indicating the flow of people within the coverage area 302 when a disaster occurs. The disaster people flow data 326 may include the number of people per time unit at each location in the coverage area 302 when a disaster occurs. The time unit may be seconds, minutes, hours, etc. The disaster people flow data 326 may further include the direction of movement and movement speed of people. The people flow data 320 may include disaster people flow data 326 for each type of disaster.

[0033] The distribution infrastructure 200 may use people flow data 320 including normal-state people flow data 322 to generate a people flow estimation model 330 that takes date and time as input and outputs people flow. The distribution infrastructure 200 may use people flow data 320 including normal-state people flow data 322 and event-time people flow data 324 to generate a people flow estimation model 330 that takes date and time and an event as input and outputs people flow. The distribution infrastructure 200 may use people flow data 320 including normal-state people flow data 322 and disaster-time people flow data 326 to generate a people flow estimation model 330 that takes date and time and a disaster as input and outputs people flow. The distribution infrastructure 200 may use people flow data 320 including normal-state people flow data 322, event-time people flow data 324, and disaster-time people flow data 326 to generate a people flow estimation model 330 that takes date and time, an event, and a disaster as input and outputs people flow. This makes it possible to estimate with relatively high accuracy the flow of people in situations where no event or disaster has occurred, as well as the flow of people in situations where an event has occurred and in situations where a disaster has occurred.

[0034] The distribution infrastructure 200 may estimate people flow using the people flow estimation model 330 for each of the multiple coverage areas 302 that make up the RAN 310, and may predict the demand of the RAN 310 based on the estimation results. The distribution infrastructure 200 may predict the demand of the RAN 310, assuming that areas with a higher number of people flow have a higher wireless communication load. The distribution infrastructure 200 may estimate the demand of the RAN 310 from the people flow estimated by the people flow estimation model 330, using a people flow demand estimation model that takes people flow within the RAN 310 as input and RAN demand as output, and that is generated by machine learning using actual data on people flow within the RAN 310 and demand for the RAN. The distribution infrastructure 200 may store the people flow demand estimation model in advance.

[0035] As described above, the people flow estimation model 330 needs to be very conscious of the characteristics of the area, and therefore, by optimizing it for the area, it becomes possible to create a more efficient model. However, in this case, if the model is to be optimized for the target area, it is necessary to collect a large amount of actual data generated for each target radio base station 300 and update the model. This type of processing can impose a very high load, so it is desirable to reduce the number of models by, for example, appropriately clustering the radio base stations 300.

[0036] Fig. 5 is an explanatory diagram for describing an example of generating a people flow estimation model 330. In the example shown in Fig. 5, the distribution infrastructure 200 classifies multiple radio base stations 300 that make up the RAN 310 into multiple clusters 340, and generates a people flow estimation model 330 for each of the multiple clusters 340. The distribution infrastructure 200 may generate the people flow estimation model 330 for each of the multiple clusters 340 using people flow data 320 within the coverage area 302 of the multiple radio base stations 300 included in the cluster 340.

[0037] The distributed infrastructure 200 may use a known clustering method such as k-means clustering to classify the multiple radio base stations 300. The clustering method used by the distributed infrastructure 200 is not limited to k-means clustering, and any method may be used.

[0038] The distributed infrastructure 200, for example, collects people flow data 320 for all of the multiple radio base stations 300 and classifies the multiple radio base stations 300 using the collected people flow data 320. The distributed infrastructure 200 classifies the multiple radio base stations 300, for example, by grouping together people flow data 320 with high similarity. This makes it possible to appropriately classify multiple radio base stations 300 with similar people flow trends.

[0039] The distributed infrastructure 200 classifies the multiple radio base stations 300, for example, based on the locations of the multiple radio base stations 300. The distributed infrastructure 200 classifies the multiple radio base stations 300, for example, by grouping together those that are close to each other. Because the flow of people in the coverage areas 302 of radio base stations 300 that are close to each other is likely to show similar trends, the multiple radio base stations 300 can be appropriately classified in this way.

[0040] The distributed infrastructure 200 classifies the multiple radio base stations 300, for example, based on the surrounding environments of the multiple radio base stations 300. The distributed infrastructure 200 classifies the multiple radio base stations 300 by grouping together radio base stations 300 with similar environments, such as roads and buildings, around the radio base stations 300. Since the flow of people in the coverage areas 302 of radio base stations 300 with similar surrounding environments is likely to show similar trends, the multiple radio base stations 300 can be appropriately classified.

[0041] The distributed infrastructure 200, for example, collects posting data posted by UEs 30 currently serving the radio base stations 300 for each of the radio base stations 300, embeds the collected posting data to generate base station vector data, and classifies the radio base stations 300 based on the base station vector data of the radio base stations 300. The posting data is, for example, data posted to a social networking service (SNS). The distributed infrastructure 200 classifies the radio base stations 300 by grouping together base station vector data with high similarity. Similar posting data via the radio base stations 300 is likely to indicate similar pedestrian flows or similarity on some scale, and this allows the radio base stations 300 with high relevance to be appropriately classified.

[0042] As described above, the distributed infrastructure 200 executes a RAN control function and an AI processing function, and therefore executes functions with different SLAs and importance. Therefore, the distributed infrastructure 200 may manage functions according to the different SLAs and importance, for example, by dividing functions into those with high priority and very high SLAs (such as RAN control), those with medium priority and high SLAs (those with high real-time characteristics, such as AI execution processing), and those with low priority and low SLAs (such as batch processing that allows task queuing, such as AI learning processing).

[0043] 6 is an explanatory diagram for explaining management of the computational resources 210 of the distributed infrastructure 200. In the example shown in Fig. 6, the distributed infrastructure 200 manages the computational resources 210 by dividing them into a RAN occupation area 212 where only the RAN control function is available, an AI occupation area 214 where only the AI ​​processing function is available, and a spot area 216 where both the RAN control function and the AI ​​processing function are available.

[0044] 7, 8, 9, and 10 are explanatory diagrams for explaining the allocation of computational resources 210 of the distributed infrastructure 200. The distributed infrastructure 200 determines the amount of computational resources to be allocated to the vRAN 203 according to the results of the demand prediction for the RAN 310. If the amount of computational resources required by the vRAN 203 is greater than the RAN occupation area 212, the distributed infrastructure 200 preferentially allocates the spot area 216 to the vRAN 203. The distributed infrastructure 200 does not allocate the AI ​​occupation area 214 to the vRAN 203.

[0045] As illustrated in FIG. 7 , the distribution infrastructure 200 allocates up to the entire RAN occupation area 212 and the spot area 216 to the vRAN 203 .

[0046] 8 , when there is a remainder 217 of the spot area 216 that has not been allocated to the vRAN 203, the distributed infrastructure 200 may allocate the remainder 217 to the AI ​​processing 204. This makes it possible to increase the amount of computational resources allocated to the vRAN 203 in response to an increase in demand for the RAN 310 while maintaining a minimum level of execution of the AI ​​processing 204, and when the demand for the RAN 310 is not very high, it is possible to increase the allocation to the AI ​​processing 204. If it is predicted that the demand for the RAN will increase after this allocation, the distributed infrastructure 200 may allocate the remainder 217 allocated to the AI ​​processing 204 to the vRAN 203.

[0047] The distributed infrastructure 200 may allocate resources preferentially to the AI ​​execution process out of the AI ​​execution process that executes AI and the AI ​​learning process that learns AI. For example, the distributed infrastructure 200 allocates the AI ​​occupied area 214 and the remaining portion 217 to the AI ​​execution process, and if there is a portion of the remaining portion 217 that is not allocated to the AI ​​execution process, allocates the remaining portion to the AI ​​learning process. This makes it possible to prioritize the execution of the AI ​​execution process, which requires greater real-time performance, out of the AI ​​execution process and the AI ​​learning process.

[0048] The distributed infrastructure 200 may allocate resources preferentially to the RAN control AI execution process between the RAN control AI execution process related to RAN control and the non-RAN control AI execution process not related to RAN control. For example, the distributed infrastructure 200 allocates the AI ​​occupied area 214 and the remaining part 217 to the RAN control AI execution process, and if there is a remaining part of the remaining part 217 that is not allocated to the RAN control AI execution process, allocates the remaining part to the non-RAN control AI execution process. This makes it possible to preferentially execute the RAN control AI execution process that is considered to have a higher SLA between the RAN control AI execution process and the non-RAN control AI execution process.

[0049] 9 , when not all of the spot area 216 is assigned to the vRAN 203, the distributed infrastructure 200 may assign all of the spot area 216 to the AI ​​processing 204. The distributed infrastructure 200 may assign the AI ​​occupation area 214 and the spot area 216 to the AI ​​execution processing, and when there is a remaining portion of the spot area 216 that is not assigned to the AI ​​execution processing, the distributed infrastructure 200 may assign the remaining portion to the AI ​​learning processing. The distributed infrastructure 200 may assign the AI ​​occupation area 214 and the spot area 216 to the RAN-controlled AI execution processing, and when there is a remaining portion of the spot area 216 that is not assigned to the RAN-controlled AI execution processing, the distributed infrastructure 200 may assign the remaining portion to the non-RAN-controlled AI execution processing.

[0050] 10 , even if the amount of computational resources required by vRAN 203 is smaller than that of RAN occupation area 212, distributed infrastructure 200 does not allocate RAN occupation area 212 to AI processing 204. This allows the remaining portion of RAN occupation area 212 to be immediately allocated to vRAN 203 when demand for RAN 310 increases. Furthermore, if RAN occupation area 212 were allocable to AI processing 204, an increase in demand for RAN 310 while AI processing 204 is using part of RAN occupation area 212 would force the AI ​​processing 204 to be immediately stopped. However, by managing the RAN occupation area 212 so as not to be allocated to AI processing 204, such a situation can be prevented.

[0051] While the Sustainable Development Goals (SDGs) are being called for, there is a problem of power consumption. For example, by unnecessarily activating and running a high-performance GPU, there is a possibility that a large amount of output will be wasted. Therefore, it is important to create a mechanism for using efficient high-performance GPUs and shutting down unnecessary GPUs. In particular, during the nighttime, the demand for the RAN 310 decreases, so it is important to reduce the waste of power there. However, it is considered necessary to respond when a sudden demand for the RAN 310 occurs.

[0052] 11 and 12 are explanatory diagrams for explaining the power saving function of the distributed infrastructure 200. The distributed infrastructure 200 may transition hardware corresponding to unused portions of the spot area 216 to a power saving state. For example, in a case where the spot area 216 is realized by multiple GPUs, the distributed infrastructure 200 transitions one or more GPUs corresponding to unused portions of the spot area 216 to a power saving state.

[0053] The power saving state may be a state in which the power is turned off, a sleep state, a standby state, or a state in which an application that is normally running is not running.

[0054] 11, when a portion of the spot area 216 is not being used, the distribution infrastructure 200 may transition the hardware corresponding to the unused portion of the spot area 216 to a power-saving state. As shown in Fig. 12, when the entire spot area 216 is not being used, the distribution infrastructure 200 may transition the hardware corresponding to the entire spot area 216 to a power-saving state. This allows for appropriate reduction in power consumption.

[0055] The distributed infrastructure 200 may transition hardware corresponding to unused portions of the spot area 216 to different power-saving states depending on the allocation rate of the RAN occupation area 212 to the vRAN 203. For example, when the allocation rate of the RAN occupation area 212 is lower than a first threshold, the distributed infrastructure 200 controls the power supply of the hardware corresponding to unused portions of the spot area 216 to be turned off. When the allocation rate of the RAN occupation area 212 is higher than the first threshold and lower than a second threshold higher than the first threshold, the distributed infrastructure 200 controls the power supply of the hardware corresponding to unused portions of the spot area 216 to be transitioned to a sleep state. In this way, when the load of the vRAN 203 is low and sufficiently contained within the RAN occupation area 212, the power supply can be turned off to enhance the power-saving effect. Furthermore, when the load of the vRAN 203 is relatively high and there is a high possibility that it will exceed the RAN occupation area 212, the power supply can be transitioned to a sleep state, thereby enabling the spot area 216 to be quickly allocated when the load of the vRAN 203 increases.

[0056] 13 shows an example of the functional configuration of the distributed infrastructure 200. The distributed infrastructure 200 includes a storage unit 222, a data acquisition unit 224, a classification unit 226, a model generation unit 228, a demand forecasting unit 230, a forecast result acquisition unit 232, an AI management unit 234, a resource allocation unit 236, an execution unit 238, a resource management unit 240, and a state control unit 242. Note that it is not essential for the distributed infrastructure 200 to include all of these units.

[0057] The data acquisition unit 224 acquires various data and stores the acquired data in the storage unit 222.

[0058] The data acquisition unit 224 acquires, for example, data related to the subordinate radio base station 300. The data acquisition unit 224 may acquire location data of the radio base station 300. The location data of the radio base station 300 indicates, for example, the latitude and longitude of the radio base station 300. The data acquisition unit 224 may acquire surrounding environment data indicating the surrounding environment of the radio base station 300. The surrounding environment data may be, for example, three-dimensional map data of the area around the radio base station 300.

[0059] The data acquisition unit 224 may acquire performance data of the wireless communication service provided by the wireless base station 300. The performance data may include data of the UEs 30 that are within the range of the wireless base station 300. The performance data may include the amount of communication by the UEs 30 that are within the range of the wireless base station 300. The data acquisition unit 224 may acquire posted data posted by the UEs 30 that are within the range of the wireless base station 300.

[0060] The data acquisition unit 224 may acquire people flow data 320 within the coverage area 302 of the wireless base station 300. The data acquisition unit 224 may receive people flow data 320 from an external source. The data acquisition unit 224 may acquire people flow data 320 generated in the distributed platform 200. The people flow data 320 may include normal people flow data 322. The people flow data 320 may include event people flow data 324. The people flow data 320 may include disaster people flow data 326.

[0061] The classification unit 226 classifies the multiple radio base stations 300 into multiple clusters. The classification unit 226 may classify the multiple radio base stations 300 using a known clustering method such as k-means clustering. The clustering method used by the distribution infrastructure 200 is not limited to k-means clustering, and any method may be used.

[0062] The classification unit 226 classifies the multiple radio base stations 300, for example, by using the people flow data 320 of the multiple radio base stations 300. The classification unit 226 classifies the multiple radio base stations 300, for example, by grouping together people flow data 320 with high similarity.

[0063] The classification unit 226 classifies the multiple radio base stations 300, for example, based on the location data of the multiple radio base stations 300. The distribution infrastructure 200 classifies the multiple radio base stations 300, for example, by grouping together those that are close to each other.

[0064] The classification unit 226 classifies the multiple radio base stations 300, for example, based on the surrounding environment data of the multiple radio base stations 300. The classification unit 226 classifies the multiple radio base stations 300, for example, by grouping together radio base stations 300 that have similar environments, such as roads and buildings, around the radio base stations 300.

[0065] The classification unit 226 generates base station vector data for each of the multiple radio base stations 300, for example, by embedding posted data posted by UEs 30 currently in the coverage area of ​​the radio base station 300, and classifies the multiple radio base stations 300 based on the base station vector data of the multiple radio base stations 300. The classification unit 226 classifies the multiple radio base stations 300, for example, by grouping together base station vector data with high similarity.

[0066] The model generation unit 228 generates the people flow estimation model 330. The model generation unit 228 stores the generated people flow estimation model 330 in the storage unit 222. The model generation unit 228 may generate the people flow estimation model 330 by performing machine learning using the people flow data 320 for each of the multiple radio base stations 300.

[0067] The model generation unit 228 may use the people flow data 320 including the normal-state people flow data 322 to generate a people flow estimation model 330 that takes date and time as input and outputs people flow. The model generation unit 228 may use the people flow data 320 including the normal-state people flow data 322 and the event-state people flow data 324 to generate a people flow estimation model 330 that takes date and time and an event as input and outputs people flow. The model generation unit 228 may use the people flow data 320 including the normal-state people flow data 322 and the disaster-state people flow data 326 to generate a people flow estimation model 330 that takes date and time and a disaster as input and outputs people flow. The model generation unit 228 may use the people flow data 320 including the normal-state people flow data 322, the event-state people flow data 324, and the disaster-state people flow data 326 to generate a people flow estimation model 330 that takes date and time, an event, and a disaster as input and outputs people flow.

[0068] The model generation unit 228 may generate a people flow estimation model 330 for each of the multiple clusters classified by the classification unit 226 by performing machine learning using people flow data 320 within the coverage areas 302 of the multiple radio base stations 300 included in the cluster.

[0069] The demand prediction unit 230 may estimate people flow for each of the multiple coverage areas 302 that make up the RAN 310 using the people flow estimation model 330, and may predict the demand of the RAN 310 based on the estimation result. The demand prediction unit 230 may predict the demand of the RAN 310, assuming that areas with a higher people flow have a higher wireless communication load. The demand prediction unit 230 may estimate the demand of the RAN 310 from the people flow estimated by the people flow estimation model 330, using a people flow demand estimation model that takes people flow within the RAN 310 as input and outputs the demand of the RAN, the people flow demand estimation model being generated by machine learning using actual data on people flow within the RAN 310 and the demand of the RAN. The people flow demand estimation model may be stored in advance in the storage unit 222.

[0070] The prediction result acquisition unit 232 acquires the prediction result of predicting the demand of the RAN 310. The prediction result acquisition unit 232 may acquire the prediction result of the demand of the RAN 310 predicted by the demand prediction unit 230 from the demand prediction unit 230. The prediction result acquisition unit 232 may receive the prediction result of the demand of the RAN 310 from an external source.

[0071] The AI ​​management unit 234 manages AI processing. For example, the AI ​​management unit 234 acquires and manages an AI processing execution request that requests the execution of AI processing.

[0072] The resource allocation unit 236 determines the amount of computational resources of the distributed infrastructure 200 to be allocated to the RAN control function and the amount of computational resources of the distributed infrastructure 200 to be allocated to the AI ​​processing function, and allocates them to the RAN control function and the AI ​​processing function. The resource allocation unit 236 may determine the amount of computational resources of the distributed infrastructure 200 to be allocated to the RAN control function and the amount of computational resources of the distributed infrastructure 200 to be allocated to the AI ​​processing function, based on the prediction result of the demand for RAN 310 acquired by the prediction result acquisition unit 232, and allocate them to the RAN control function and the AI ​​processing function.

[0073] The resource allocation unit 236 may allocate computational resources preferentially to the RAN control function out of the RAN control function and the AI ​​processing function. For example, the resource allocation unit 236 determines the amount of computational resources to be allocated to the RAN control function based on the predicted results of demand for the system 10, and then, if there is surplus computational resources, allocates the surplus to the AI ​​processing function.

[0074] The execution unit 238 executes RAN control functions and AI processing functions on the distributed infrastructure 200 according to the amount of computational resources allocated by the resource allocation unit 236.

[0075] The resource management unit 240 manages the computing resources on the distributed infrastructure 200 by dividing them into a RAN-occupied area where only the RAN control function is available, an AI-occupied area where only the AI ​​processing function is available, and a spot area where both the RAN control function and the AI ​​processing function are available.

[0076] The resource allocation unit 236 may allocate the RAN occupation area to the RAN control function. The resource allocation unit 236 may allocate the AI ​​occupation area to the AI ​​processing function. The resource allocation unit 236 may preferentially allocate the spot area to the RAN control function out of the RAN control function and the AI ​​processing function. For example, if the amount of computational resources required by the RAN control function is greater than the RAN occupation area, the resource allocation unit 236 preferentially allocates the spot area to the RAN control function.

[0077] If there is a remainder (sometimes referred to as a first remainder) of the spot area that is not assigned to the RAN control function, the resource allocation unit 236 may allocate the first remainder to the AI ​​processing function. If the entire spot area is not assigned to the RAN control function, the entire spot area may be assigned to the AI ​​processing function as the first remainder. After allocating part or all of the spot area to the AI ​​processing function, if an increase in demand for the RAN 310 is predicted, the resource allocation unit 236 may reallocate part or all of the spot area allocated to the AI ​​processing function to the RAN control function.

[0078] As described above, the AI ​​processing function may include an AI execution process and an AI learning process. The resource allocation unit 236 may allocate the AI ​​occupation area and the first remainder to the AI ​​execution process, and if there is a remainder (sometimes referred to as a second remainder) of the first remainder that is not allocated to the AI ​​execution process, may allocate the second remainder to the AI ​​learning process.

[0079] As described above, the AI ​​execution process may include a RAN-controlled AI execution process and a non-RAN-controlled AI execution process. Of the RAN-controlled AI execution process and the non-RAN-controlled AI execution process, the resource allocation unit 236 may allocate the AI ​​occupation area and the first remainder portion preferentially to the RAN-controlled AI execution process.

[0080] When at least a portion of the spot area is not assigned to the RAN control function, the resource allocation unit 236 may decide, depending on the demand for AI processing, whether to allocate the portion of the spot area not assigned to the RAN control function to the AI ​​processing function, or to transition the hardware corresponding to the portion of the spot area not assigned to the RAN control function to a power-saving state.

[0081] The resource allocation unit 236 determines the demand for AI processing, for example, based on the status of AI processing execution requests managed by the AI ​​management unit 234. For example, when at least a portion of the spot area is not allocated to the RAN control function, the resource allocation unit 236 determines the amount of computational resources to be allocated to the AI ​​processing function within the spot area based on the demand for AI processing, and if there is a surplus, decides to transition the hardware corresponding to the surplus to a power saving state.

[0082] When the resource allocation unit 236 determines to transition the hardware corresponding to the unused portion of the spot area to a power saving state, the state control unit 242 may transition the hardware corresponding to the unused portion of the spot area to a different power saving state depending on the allocation rate of the RAN occupation area to the RAN control function. As described above, the power saving state may be a state in which the power is turned off. The power saving state may be a sleep state. The power saving state may be a standby state. The power saving state may also be a state in which an application that is normally running is not running.

[0083] The state control unit 242 may control the power supply to be turned off for hardware corresponding to unused parts of the spot area when the allocation rate of the RAN occupied area is lower than a predetermined first threshold, and may control the power supply to be turned off for hardware corresponding to unused parts of the spot area when the allocation rate of the RAN occupied area is higher than the first threshold and lower than a second threshold higher than the first threshold, to transition the hardware corresponding to unused parts of the spot area to a sleep state.

[0084] 14 is an explanatory diagram for providing an overview of the architecture 400 in the system 10. The distributed infrastructure 200 clusters the multiple radio base stations 300 under its control using historical usage data for the RAN 310, people flow data, event data, and disaster data in the area corresponding to the RAN 310, and these data for other areas, to select a model creation area and generate an AI model (such as a people flow estimation model) for the selected model creation area. The distributed infrastructure 200 uses the AI ​​model to generate forecast data for the demand of the RAN 310. The distributed infrastructure 200 also collects AI demand and assigns priorities to AI processing execution requests. The distributed infrastructure 200 registers low-priority AI processing execution requests in a queue.

[0085] Based on these, the distributed infrastructure 200 allocates resources for the RAN control function from the RAN occupied area and the spot area, and allocates resources to high-priority AI processing from the AI ​​occupied area and, if there is a remainder in the spot area, from the remainder, according to the priority of the AI ​​processing execution request. If there is a remainder in the spot area after this allocation, the remainder is allocated to the AI ​​processing registered in the queue.

[0086] The distribution infrastructure 200 may further execute an emergency allocation process in the event of an abnormal situation, such as a sudden increase in demand for the RAN 310. The occurrence of such an abnormal situation may be detected by using an abnormal situation determination model that has been generated and stored in advance. For example, by using historical usage data during an abnormal situation, it is possible to generate an abnormal situation determination model that can predict the occurrence of an abnormal situation from the historical usage data. For example, by using people flow data during an abnormal situation, it is possible to generate an abnormal situation determination model that can predict the occurrence of an abnormal situation from the people flow data.

[0087] When an abnormal situation occurs, the distributed infrastructure 200 may, for example, detect available resources in a nationwide RAN, and if available, control the use of the available resources as resources to respond to the abnormal situation. If no available resources are available, the distributed infrastructure 200 may control the use of spot areas of its own or other distributed infrastructures 200 as resources to respond to the abnormal situation. While performing these operations, the distributed infrastructure 200 may determine whether its own computing resources include unnecessary resources, and if it determines that unnecessary resources exist, may perform control to transition hardware corresponding to the unnecessary resources to a power-saving state.

[0088] 15 schematically illustrates an example of the hardware configuration of a computer 1200 functioning as the management infrastructure 100 or the distribution infrastructure 200. A program installed on the computer 1200 can cause the computer 1200 to function as one or more "parts" of an apparatus according to the present embodiment, or can cause the computer 1200 to execute operations associated with the apparatus according to the present embodiment or one or more "parts," and / or can cause the computer 1200 to execute a process according to the present embodiment or steps of the process. Such a program can be executed by the CPU 1212 to cause the computer 1200 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

[0089] The computer 1200 according to this embodiment includes a CPU 1212, a GPU 1213, a RAM 1214, and a graphics controller 1216, which are interconnected by a host controller 1210. The computer 1200 also includes input / output units such as a communications interface 1222, a storage device 1224, a DVD drive 1226, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The DVD drive 1226 may be a DVD-ROM drive, a DVD-RAM drive, or the like. The storage device 1224 may be a hard disk drive, a solid-state drive, or the like. The computer 1200 also includes a ROM 1230 and legacy input / output units such as a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0090] The CPU 1212 operates according to programs stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into the graphics controller 1216 itself, and causes the image data to be displayed on the display device 1218.

[0091] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive 1226 reads programs or data from a DVD-ROM 1227 or the like and provides them to the storage device 1224. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0092] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or programs that depend on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a USB port, a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0093] The programs are provided by a computer-readable storage medium such as a DVD-ROM 1227 or an IC card. The programs are read from the computer-readable storage medium, installed in the storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable storage media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and brings about cooperation between the programs and the various types of hardware resources described above. An apparatus or a method may be configured by implementing operations or processing of information in accordance with the use of the computer 1200.

[0094] For example, when communication is performed between computer 1200 and an external device, CPU 1212 may execute a communication program loaded into RAM 1214 and instruct communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of CPU 1212, communication interface 1222 reads transmission data stored in a transmission buffer area provided in RAM 1214, storage device 1224, DVD-ROM 1227, or a recording medium such as an IC card, and transmits the read transmission data to a network, or writes received data received from the network to a reception buffer area or the like provided on the recording medium.

[0095] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, the DVD drive 1226 (DVD-ROM 1227), an IC card, etc. to be read into the RAM 1214, and may perform various types of processing on the data on the RAM 1214. The CPU 1212 may then write back the processed data to the external recording medium.

[0096] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. The CPU 1212 may perform various types of processing on data read from the RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to the RAM 1214. The CPU 1212 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries, each having an attribute value of a first attribute associated with an attribute value of a second attribute, are stored on the recording medium, the CPU 1212 may search for an entry whose attribute value of the first attribute matches a specified condition from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0097] The above-described programs or software modules may be stored in a computer-readable storage medium on or near the computer 1200. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable storage medium, thereby providing the programs to the computer 1200 via the network.

[0098] The blocks in the flowcharts and block diagrams in the present embodiments may represent stages of a process in which an operation is performed or "parts" of a device responsible for performing the operation. Particular stages and "parts" may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable storage medium, and / or a processor provided with computer-readable instructions stored on a computer-readable storage medium. The dedicated circuitry may include digital and / or analog hardware circuits, and may include integrated circuits (ICs) and / or discrete circuits. The programmable circuitry may include reconfigurable hardware circuits, such as field programmable gate arrays (FPGAs) and programmable logic arrays (PLAs), including AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and memory elements.

[0099] A computer-readable storage medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that a computer-readable storage medium having instructions stored thereon comprises an article of manufacture, including instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable storage media may include electronic, magnetic, optical, electromagnetic, and semiconductor storage media. More specific examples of computer-readable storage media may include floppy disks, diskettes, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), electrically erasable programmable read-only memories (EEPROMs), static random access memories (SRAMs), compact disc read-only memories (CD-ROMs), digital versatile discs (DVDs), Blu-ray discs, memory sticks, integrated circuit cards, and the like.

[0100] The computer readable instructions may include either assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0101] The computer-readable instructions may be provided to a general-purpose computer, a special-purpose computer, or another programmable data processing device processor or programmable circuit, either locally or via a local area network (LAN), a wide area network (WAN) such as the Internet, so that the processor or programmable circuit of the programmable data processing device, such as a computer, executes the computer-readable instructions to generate means for performing the operations specified in the flowcharts or block diagrams. Here, the computer may be a personal computer (PC), a tablet computer, a smartphone, a workstation, a server computer, a general-purpose computer, a special-purpose computer, or the like, or may be a computer system in which multiple computers are connected. Such a computer system in which multiple computers are connected is also called a distributed computing system, and is a broad definition of computer. In a distributed computing system, multiple computers collectively execute a program by each executing a portion of the program and passing data between the computers as needed during program execution.

[0102] Examples of processors include computer processors, central processing units (CPUs), processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc. A computer may have one processor or multiple processors. In a multiprocessor system with multiple processors, each processor executes a portion of a program and passes data between processors as needed during program execution, allowing the multiple processors to collectively execute the program. For example, in multitasking, each of the multiple processors may execute a portion of each task in small chunks by switching tasks at each time slice. In this case, which portion of a program each processor executes changes dynamically. Which portion of a program each of the multiple processors executes may also be statically determined by multiprocessor-aware programming.

[0103] Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0104] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a subsequent process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order.

[0105] 10 System, 30 UE, 100 Management infrastructure, 200 Distribution infrastructure, 201 HW, 202 Virtualization infrastructure, 203 vRAN, 204 AI processing, 205 RAN_AI, 206 Non-RAN_AI, 210 Computational resources, 212 RAN occupied area, 214 AI occupied area, 216 Spot area, 217 Remaining part, 222 Memory unit, 224 Data acquisition unit, 226 Classification unit, 228 Model generation unit, 230 Demand forecasting unit, 232 Prediction result acquisition unit, 234 AI management unit, 236 Resource allocation unit, 238 Execution unit, 240 Resource management unit, 242 State control unit, 300 Radio base station, 302 Coverage area, 310 RAN, 320 People flow data, 322 Normal time people flow data, 324 Event time people flow data, 326 Disaster time people flow data, 330 People flow estimation model, 340 Cluster, 400 Architecture, 1200 Computer, 1210 Host controller, 1212 CPU, 1213 GPU, 1214 RAM, 1216 Graphics controller, 1218 Display device, 1220 Input / output controller, 1222 Communication interface, 1224 Storage device, 1226 DVD drive, 1227 DVD-ROM, 1230 ROM, 1240 Input / output chip

Claims

1. A system comprising: a resource allocation unit that determines, based on the results of forecasting demand for a RAN (Radio Access Network), the amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN and the amount of computational resources of the execution platform to be allocated to an AI processing function that performs AI (Artificial Intelligence) processing, and allocates the resources to the RAN control function and the AI ​​processing function; and an execution unit that executes the RAN control function and the AI ​​processing function on the execution platform in accordance with the amount of computational resources allocated by the resource allocation unit.

2. The system of claim 1, further comprising a demand forecasting unit that estimates people flow for each of a plurality of coverage areas constituting the RAN using a people flow estimation model that estimates people flow within the coverage area, and predicts demand for the RAN based on the estimation results, wherein the resource allocation unit determines the amount of computational resources of the system to be allocated to the RAN control function and the amount of computational resources of the system to be allocated to the AI ​​processing function based on the prediction results of demand for the RAN by the demand forecasting unit.

3. The system according to claim 2, further comprising: a model generation unit that generates the people flow estimation model by performing machine learning using people flow data indicating people flow within a coverage area of ​​a wireless base station.

4. The system described in claim 3, wherein the model generation unit uses normal time people flow data indicating people flow for each date and time, event time people flow data indicating people flow when an event occurs within the coverage area of ​​the wireless base station, and disaster time people flow data indicating people flow when a disaster occurs within the coverage area of ​​the wireless base station to generate the people flow estimation model, which takes date and time, events, and disasters as inputs and outputs people flow.

5. The system according to claim 3 or 4, further comprising a classification unit that classifies a plurality of wireless base stations into a plurality of clusters, wherein the model generation unit generates the people flow estimation model for each of the plurality of clusters by performing machine learning using the people flow data within the coverage areas of the plurality of wireless base stations included in the cluster.

6. The system described in claim 5, wherein the classification unit collects posting data posted by user terminals within the range of each of the plurality of radio base stations, generates base station vector data by embedding the collected posting data, and classifies the plurality of radio base stations into the plurality of clusters based on the base station vector data of the plurality of radio base stations.

7. A system as described in any one of claims 1 to 6, further comprising a resource management unit that divides and manages the computational resources on the one execution platform into a RAN occupied area that can only be used by the RAN control function, an AI occupied area that can only be used by the AI ​​processing function, and a spot area that can be used by both the RAN control function and the AI ​​processing function, wherein the resource allocation unit preferentially allocates the spot area to the RAN control function when the amount of computational resources required by the RAN control function is greater than the RAN occupied area.

8. The system described in claim 7, wherein, when at least a portion of the spot area is not assigned to the RAN control function, the resource allocation unit determines, depending on the demand for the AI ​​processing, whether to allocate the portion of the spot area not assigned to the RAN control function to the AI ​​processing function, or to transition hardware corresponding to the portion of the spot area not assigned to the RAN control function to a power-saving state.

9. The system described in claim 8, further comprising a state control unit that, when it is determined that hardware corresponding to a portion of the spot area not assigned to the RAN control function is transitioned to a power saving state, transitions the hardware corresponding to a portion of the spot area not assigned to the RAN control function to a different power saving state depending on the allocation rate of the RAN occupied area.

10. The system described in claim 9, wherein the state control unit controls to turn off the power of hardware corresponding to a portion of the spot area not allocated to the RAN control function when the allocation rate of the RAN occupied area is lower than a predetermined first threshold, and controls to transition the hardware corresponding to a portion of the spot area not allocated to the RAN control function to a sleep state when the allocation rate of the RAN occupied area is higher than the first threshold and lower than a second threshold higher than the first threshold.

11. A program for causing a computer to execute: a determination step for determining, based on the results of a forecast of RAN demand, the amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN, and the amount of computational resources of said execution platform to be allocated to an AI processing function that performs AI processing; and an execution step for executing said RAN control function and said AI processing function on said execution platform in accordance with the amount of computational resources determined in said determination step.

12. A control method executed by a computer, comprising: a determination step of determining, based on the results of a forecast of RAN demand, the amount of computational resources of an execution platform to be allocated to a RAN control function that controls the RAN, and the amount of computational resources of said execution platform to be allocated to an AI processing function that performs AI processing; and an execution step of executing said RAN control function and said AI processing function on said execution platform in accordance with the amount of computational resources determined in said determination step.

Images