A task processing method and device, equipment, and storage medium

Abstract

The application discloses a task processing method and device, equipment and a storage medium, wherein the method comprises the following steps: a first digital twin network (DTN) unit receives a task request sent by a user; the task request comprises parameter information indicating task requirements of a target task; a DTN unit set for processing the target task corresponding to the task request is determined according to the parameter information included in the task request and node information of each second DTN unit in N1 second DTN units; N1 is an integer greater than or equal to 1; wherein the first DTN unit is deployed in a central unit (CU) of a base station, the N1 second DTN units are deployed in N2 distributed units (DUs) of the base station, and the DTN unit set comprises N3 second DTN units; the DTN unit set belongs to the N1 second DTN units; N2 and N3 are integers less than or equal to N1.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of communication, and in particular to a task processing method and device, equipment and storage medium. BACKGROUND

[0002] The task processing of a digital twin network (DTN) includes real-time collection and processing of data, strategy generation and verification, configuration or adjustment of the verified strategy, etc. For some scenarios with high requirements on time delay, the deployment mode and interaction process of the digital twin network deployed in a cloud server or a mobile edge computing (MEC) cannot meet the requirement of low time delay. For example, if the digital twin network is deployed in a cloud server, the data collection and the configuration of the simulation strategy through the digital twin network will increase the transmission time delay, and the centralized task processing of the digital twin network requires a longer time than the distributed task processing. Although the deployment mode of the MEC can perform distributed task processing to reduce the time delay of the task processing, the time delay caused by the data collection and the strategy configuration is unavoidable, and the transmission bandwidth resource is wasted. In summary, the current scheme of deploying the digital twin network in a cloud server or in the MEC in a distributed manner has problems of large time delay of the task processing of the digital twin network, waste of bandwidth resources, and low efficiency of the task processing. SUMMARY

[0003] To solve the above technical problems, embodiments of the present application provide a task processing method and device, equipment and storage medium.

[0004] In a first aspect, embodiments of the present application provide a task processing method applied to a first digital twin network (DTN) unit, wherein the first DTN unit is deployed in a central unit (CU) of a base station, and the method comprises:

[0005] receiving a task request sent by a user, wherein the task request includes parameter information indicating a task requirement of a target task;

[0006] determining a DTN unit set for processing a target task corresponding to the task request according to the parameter information included in the task request and node information of each second DTN unit in N1 second DTN units, wherein N1 is an integer greater than or equal to 1;

[0007] wherein the N1 second DTN units are deployed in N2 distributed units (DUs) of the base station, the DTN unit set includes N3 second DTN units, the DTN unit set belongs to the N1 second DTN units, and N2 and N3 are integers less than or equal to N1.

[0008] In a second aspect, an embodiment of the present application provides a task processing method, applied to a target second DTN unit, the target second DTN unit being deployed in a distributed unit (DU) of a base station, and the method comprising:

[0009] sending node information of the target second DTN unit to a first DTN unit;

[0010] The node information is used by the first DTN unit to determine a DTN unit set for processing a target task corresponding to a task request according to parameter information included in the task request sent by a user and node information of each second DTN unit in N1 second DTN units, the first DTN unit being deployed in a CU of the base station, the DTN unit set belonging to the N1 second DTN units, the DTN unit set including N3 second DTN units, N1 being an integer greater than or equal to 1, and N2 and N3 being integers less than or equal to N1.

[0011] In a third aspect, an embodiment of the present application provides a task processing method, applied to a third DTN unit, the third DTN unit being deployed in a CU of a base station, and the method comprising:

[0012] receiving processing results of subtasks of a target task sent by each second DTN unit in a DTN unit set;

[0013] collecting the processing results of the subtasks of the target task sent by each second DTN unit in the DTN unit set to obtain a task processing result of the target task.

[0014] In a fourth aspect, an embodiment of the present application provides a task processing apparatus, applied to a first DTN unit, the first DTN unit being deployed in a CU of a base station, and the apparatus comprising:

[0015] a first receiving unit, configured to receive a task request sent by a user, the task request including parameter information indicating task requirements of a target task;

[0016] a determining unit, configured to determine a DTN unit set for processing a target task corresponding to the task request according to parameter information included in the task request and node information of each second DTN unit in N1 second DTN units, N1 being an integer greater than or equal to 1;

[0017] The N1 second DTN units are deployed in N2 distributed units (DUs) of the base station, the DTN unit set including N3 second DTN units, the DTN unit set belonging to the N1 second DTN units, and N2 and N3 being integers less than or equal to N1.

[0018] In a fifth aspect, an embodiment of the present application provides a task processing apparatus, applied to a target second DTN unit, the target second DTN unit being deployed in a DU of a base station, and the apparatus comprising:

[0019] a first sending unit, configured to send node information of the target second DTN unit to a first DTN unit;

[0020] The node information is used by the first DTN unit to determine a DTN unit set for processing a target task corresponding to a task request according to parameter information included in the task request sent by a user and node information of each second DTN unit in N1 second DTN units, the first DTN unit is deployed in a CU of a base station, the DTN unit set belongs to the N1 second DTN units, the DTN unit set includes N3 second DTN units, N1 is an integer greater than or equal to 1, and N2 and N3 are integers less than or equal to N1.

[0021] In a sixth aspect, an embodiment of the present application provides a task processing apparatus, applied to a third DTN unit, the third DTN unit being deployed in a CU of a base station, and the apparatus comprising:

[0022] a fourth receiving unit, configured to receive processing results of subtasks of a target task sent by each second DTN unit in a DTN unit set;

[0023] a summarizing unit, configured to summarize the processing results of the subtasks of the target task sent by each second DTN unit in the DTN unit set to obtain a task processing result of the target task.

[0024] In a seventh aspect, an embodiment of the present application provides a task processing device, comprising a processor and a memory for storing a computer program capable of running on the processor, and the processor is configured to run the computer program to implement the task processing method in the first aspect, or implement the task processing method in the second aspect, or implement the task processing method in the third aspect.

[0025] In an eighth aspect, an embodiment of the present application provides a computer storage medium, which stores a computer program, and the computer program is executed by a processor to implement the task processing method in the first aspect, or implement the task processing method in the second aspect, or implement the task processing method in the third aspect.

[0026] In a ninth aspect, an embodiment of the present application provides a computer program product, including a computer program, characterized in that the computer program, when executed by a processor, implements the task processing method in the first aspect of the above embodiment, or implements the task processing method in the second aspect of the above embodiment, or implements the task processing method in the third aspect of the above embodiment.

[0027] The embodiment of the present application can reduce the digital twin task processing delay from three aspects, i.e., the delay of the UE request, the transmission delay caused by data collection, and the configuration delay caused by the saving strategy, by sinking the digital twin network unit to the gNB and deploying in a distributed manner, thereby improving the processing efficiency of the digital twin task. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 It is a schematic diagram of a digital twin network deployment manner;

[0029] Figure 2 It is a schematic diagram of a digital twin network deployment manner provided by the embodiment of the present application;

[0030] Figure 3 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 1 ;

[0031] Figure 4 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 2 ;

[0032] Figure 5 It is a schematic diagram of a computing power allocation weight determination manner;

[0033] Figure 6 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 3 ;

[0034] Figure 7 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 4 ;

[0035] Figure 8 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 5 ;

[0036] Figure 6 It is a flowchart of the task processing method provided by the embodiment of the present application Figure 3 ;

[0037] Figure 10 It is a digital twin task request and processing flow;

[0038] Figure 11 It is a schematic diagram of a task cooperation process;

[0039] Figure 12 A structure of a task processing device provided by an embodiment of the present application Figure 1 ;

[0040] Figure 13 A structure of a task processing device provided by an embodiment of the present application Figure 2 ;

[0041] Figure 14 A structure of a task processing device provided by an embodiment of the present application Figure 3 ;

[0042] Figure 15 A structure of a task processing device provided by an embodiment of the present application DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the accompanying drawings for the embodiments of the present application to make a clear and complete description of the technical solutions of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0044] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be defined and explained in the subsequent drawings.

[0045] The term “and / or” in the present document is only used to describe an association relationship, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the term “at least one” in the present document means any one of multiple or any combination of at least two of multiple, for example, including at least one of A, B and C can mean including any one or more elements selected from the set consisting of A, B and C.

[0046] The digital twin network obtains a digital model of the entire network in a virtual space through data collection, processing and parameterized modeling of each network entity and function in the network system. Through digital twinning, real-time state monitoring, trajectory prediction, fault prediction and other functions of network entities and user services are realized, and the network is self-evolving. The digital twinning of the network can also verify new functions, services and optimization features in the digital field to avoid negative effects and achieve high-level network autonomy and "zero-touch maintenance". It can serve the network itself and complete the twinning tasks involved in the planning, construction, maintenance and optimization process, and also provide digital twinning-related services for user equipment (UE, User Equipment), thereby accurately ensuring the quality of service (QoS, Quality of Service) for users.

[0047] Figure 1 A schematic diagram of a digital twin network deployment method is shown in FIG. 1. As shown in FIG. 1, the digital twin network can be deployed in the cloud or MEC. Figure 1

[0048] In the case of centralized deployment of the digital twin network in the cloud, data collection and simulation strategy configuration through the digital twin network will increase its transmission delay, and centralized task processing of the twin requires a longer time than distributed task processing.

[0049] In the case of distributed deployment of the digital twin network in the MEC, the MEC will have a data request delay when obtaining data from the base station, and the data collection and transmission will also have a delay. Furthermore, after the DTN strategy verification, network parameter optimization or configuration needs to be issued to the base station, which also increases the transmission delay.

[0050] The digital twin network can simulate and verify strategies, and many strategies will be issued to the physical network for parameter modification or configuration, which can greatly reduce the modification / configuration delay. Therefore, the embodiments of the present application propose a deployment method of an embedded digital twin network to solve the problem of low delay and waste of transmission resources.

[0051] Figure 2 A schematic diagram of the digital twin network deployment method provided by the embodiments of the present application is shown in FIG. 2. Figure 2 ​As shown, the embodiment of the present application can solve the three problems of large delay, waste of bandwidth resources, and low task processing efficiency caused by centralized deployment of cloud servers or distributed deployment of MEC in digital twin network by providing a scheme of embedding DTN in next-generation base station (gNB). The embodiment of the present application reduces the delay in three links by sinking the digital twin network unit to gNB and deploying in a distributed manner. First, the user request delay is reduced. Second, the data transmission delay is reduced. The DTN acquires the required data from the gNB in real time at the data source, greatly reducing the data transmission delay. Third, the network parameter optimization or configuration is performed after the DTN strategy verification, which can be directly configured in the gNB, saving the strategy distribution delay.

[0052] Figure 2 In the embodiment, the digital twin network units with different functions are integrated in different modules of the gNB, the digital twin network-centralized unit (DTN-CU) is deployed in the user plane (gNB-CU-CP) of the central unit of the next-generation base station and the control plane (gNB-CU-UP) of the central unit of the next-generation base station, and the digital twin network-distributed unit (DTN-DU) is deployed in the distributed unit (gNB-DU) of the next-generation base station. The DTN-CU and the plurality of DTN-DUs dynamically form a digital twin network system to provide twin services for users. The deployment manner of the digital twin network provided by the embodiment can reduce the digital twin task processing delay from three aspects of UE request delay, data acquisition caused transmission delay, and configuration delay caused by saving strategy distribution.

[0053] Figure 2 In the deployment scheme of the digital twin network, the DTN-CU and each DTN-DU will be updated in real time according to the changes of the physical network, and the acquisition manner can be distributed by the cloud or locally constructed under the condition that the gNB computing resources meet the requirements. The DTN distributed by the cloud will drive the DTN to update in real time according to the local data of the gNB. On the current standard architecture, the following improvements need to be made in the gNB:

[0054] 1. Programmable design can be performed. The control plane supports programming in P4 and other languages, and the logic and program suitable for the function of the digital twin network can be developed on demand;

[0055] 2. Enhance computing power and support in-network computing. For digital twin related business requests, the digital twin network unit in the gNB can be used for processing;

[0056] 3. Support the interactive process of user digital twin task requests;

[0057] 4. The need to increase the gNB internal and DTN-CU, DTN-DU interaction interface, and the relevant interface with external network element registration, processing involved.

[0058] The main functions of the DTN-CU in the gNB-CU-CP include:

[0059] 1. Basic functions of wireless digital twin network, such as accurate data perception, efficient modeling, intelligent decision-making, simulation verification, etc.

[0060] 2. Task decomposition function, when receiving user requests, it needs to implement intelligent analysis of multiple users and multiple tasks, and classify and issue them to DTN-DU according to different processing types. In addition, it needs to intelligently decompose large tasks of the same type for a single user, split the large task into different sub-tasks, and enable DTN-DU to perform collaborative computing, federated verification, etc. According to intelligent distribution to different DTN-DU.

[0061] 3. Dynamic formation and intelligent scheduling function; according to the information reported by DTN-DU, generate a global state diagram, convert UE requests into constraint conditions, quickly generate the optimal DTN-DU combination with the global state diagram, dynamically form the twin environment suitable for different tasks, and intelligently schedule the twin sub-tasks to different DTN-DU.

[0062] 4. Strategy translation function; if there is a need for network policy configuration, parameter modification, etc. in the twin task, the DTN-CU needs to translate the intelligent strategy generated by the DTN unit into network configuration language and issue it through gNB.

[0063] The main functions of the DTN-CU in the gNB-CU-UP include:

[0064] 1. Basic functions of digital twin network, such as accurate data perception, efficient modeling, intelligent decision-making, simulation verification, holographic visualization, etc.

[0065] 2. Result summary and policy feedback function; it can uniformly summarize and analyze the task processing results of each DTN-DU, such as policy conflict detection, result statistical analysis, etc., form a twin task report, and feedback to the user, at the same time, feedback the policy result to the DTN-CU in the gNB-CU-CP.

[0066] The main functions of the DTN-DU in the gNB-DU include:

[0067] 1. Basic functions of digital twin network, such as accurate data perception, efficient modeling, intelligent decision-making, simulation verification, holographic visualization, etc. It can support the needs of twin tasks, such as task prediction, simulation verification, intelligent decision-making, etc.

[0068] 2. The DTN-DU can realize distributed cooperative calculation of tasks, and can also perform federal verification, such as cooperative training and mutual verification.

[0069] Based on the above Figure 2 The deployment scheme of the digital twin network is introduced in combination with Figures 3 to 11 the related embodiments of the processing of the digital twin task.

[0070] Figure 3 The flowchart of the task processing method provided in the embodiments of the present application is shown in Figure 1 The application is applied to a first DTN unit and includes the following steps:

[0071] S301: receiving a task request sent by a user;

[0072] S302: determining a DTN unit set for processing a target task corresponding to the task request according to parameter information included in the task request and node information of each second DTN unit in N1 second DTN units.

[0073] In the embodiments of the present application, the first DTN unit is deployed in the CU of the base station; and the N1 second DTN units are deployed in N2 distributed units (DUs) of the base station.

[0074] In combination with Figure 2 The first DTN unit is a DTN-CU in the gNB-CU-CP; and the second DTN unit is a DTN-DU deployed in each gNB-DU. The first DTN unit and the N1 second DTN units dynamically form a digital twin network system.

[0075] In the embodiments of the present application, the task request includes parameter information indicating the task requirements of the target task. In some embodiments, the parameter information in the task request includes one or more of the following: task type, task waiting time delay, and task accuracy.

[0076] In practical applications, the UE initiates a task request to the DTN-CU of the gNB-CU-CP, and the task request parameters include but are not limited to [Task Type, Maximum latency, Task Accuracy], Task Type 1 is parameter optimization, Task Type 2 is decision generation, Task Type 3 is policy pre-validation, Task Type 4 is performance simulation, Task Type 5 is state, behavior, etc. prediction, Task Type 6 is data perception processing, Task Type 7 is visualization, if the UE initiates one request, it can be represented as Task Type A, if the UE initiates multiple requests, it can be represented as Task Type A&B&C; wherein Maximum latency represents the maximum waiting time delay of the UE, which is a constraint condition for task distribution strategy generation; Task Accuracy is only set for decision tasks, especially intelligent decision based on artificial intelligence (AI, Artificial Intelligence).

[0077] In some embodiments, the node information includes node state information and node connectivity information; wherein,

[0078] The node state information includes one or more of the following: number of task queues, server type, memory occupancy rate, server available computing power, supported model type, supported model accuracy, data volume, data type;

[0079] The node connectivity information includes one or more of the following: relationship information between the first DTN unit and each of the N1 second DTN units, relationship information between each of the N1 second DTN units; the relationship information includes one or more of the following: physical connection relationship, task connectivity relationship.

[0080] In practical applications, each DTN-DU reports the node information of each DTN-DU to the DTN-CU in Key-Value format through the F1 interface, and the node information includes but is not limited to [{number of task queues: n1}; {server type: CPU / GPU / DPU / TPU / NPU / BPU…}; {CPU / GPU / TUP / NPU…occupancy rate: n2%}; {memory occupancy rate: n3%}; {server available computing power: n4}; {supported AI model type: fault detection, state prediction, data enhancement, parameter optimization…}; {supported AI model accuracy: n5%}; {data volume: N}; {data type: Tensor, feature name…}] and the like.

[0081] In the embodiment of the present application, the DTN unit set includes N3 second DTN units; the DTN unit set belongs to the N1 second DTN units; N1 is an integer greater than or equal to 1; N2 and N3 are integers less than or equal to N1.

[0082] In actual application, the DTN-CU selects the optimal DTN-DU combination from the N1 DTN-DUs according to the parameter information in the task request and the node information of each DTN-DU to form the dynamic digital twin network. After selecting the optimal DTN-DU combination, the DTN-CU splits the task and distributes it to the DTN-DUs within the formation range for task processing.

[0083] The embodiment of the present application can reduce the digital twin task processing delay from three aspects: UE request delay, transmission delay caused by data collection, and configuration delay caused by saving strategy, by sinking the digital twin network unit to the gNB and deploying in a distributed manner, thereby improving the processing efficiency of the digital twin task.

[0084] Figure 4 Flowchart of the task processing method provided by the embodiment of the present application Figure 2 , applied to a first DTN unit, comprising the following steps:

[0085] S401: receiving a task request sent by a user;

[0086] S402: constructing a global state graph according to the node information of each second DTN unit in the N1 second DTN units;

[0087] S403: converting the parameter information in the task request into a constraint condition of the global state graph;

[0088] S404: searching the global state graph using the constraint condition to obtain N4 groups of second DTN units satisfying the constraint condition;

[0089] S405: selecting the DTN unit set from the N4 groups of second DTN units, with the connectivity of the network topology of each second DTN unit in the N4 groups of second DTN units being minimized as the target.

[0090] In the above step S402, the DTN-CU in the gNB-CU-CP constructs a global state graph according to the DU reporting information The node V information dynamically stores and updates the information reported by each DTN-DU in the form of Key-Value, and the edge E is used to store the physical topology relationship between the DTN-CU and each DTN-DU and the connection relationship between tasks.

[0091] In some embodiments, the step S403 described above comprises the following steps:

[0092] S4031: taking the task type of the searched second DTN unit set in the global state graph as a first constraint sub-condition, which meets the task type in the parameter information;

[0093] S4032: taking the task accuracy of the searched second DTN unit set in the global state graph as a second constraint sub-condition, which is greater than or equal to the task accuracy in the parameter information;

[0094] S4033: taking the task latency parameter in the parameter information as the computing power requirement of the target task, and taking the total computing power that the searched second DTN unit set in the global state graph can provide as a third constraint sub-condition, which is greater than the computing power requirement of the target task.

[0095] In the step S403 described above, the constraint condition comprises the first constraint sub-condition, the second constraint sub-condition and the third constraint sub-condition; the first constraint sub-condition is used to constrain the task type of the searched second DTN unit set in the global state graph to meet the task type of the task request; the second constraint sub-condition is used to constrain the task accuracy of the searched second DTN unit set in the global state graph to be greater than or equal to the task accuracy in the task request; and the third constraint sub-condition is used to constrain the task processing time consumption of the searched second DTN unit set in the global state graph to be less than or equal to the task latency in the task request.

[0096] Next, the implementation mode of converting the parameter information in the task request into the constraint condition of the global state graph is introduced.

[0097] In actual application, the global state graph constructed according to the node information of each DTN-DU cannot directly identify the request information of the UE, so it is necessary to convert the UE request information into the constraint condition searchable by the global state graph for input, so as to obtain all independent subsets of the optimal node combination. The UE request information is converted into the constraint condition Q, that is, under the conditions of meeting the service type and ensuring the service accuracy to be greater than the required accuracy of the UE, the DTN-DU combination with the minimum task processing time consumption is selected, which can be represented by the following formula:

[0098] (1)

[0099] Since the TaskTime in the formula Q cannot be searched in the global state graph, it is necessary to be further mapped into the condition stored in the global state graph for retrieval, so here the demand of the UE for the Maximum Latency is mapped into the demand for the computing power provided by the server.

[0100] In practical applications, considering that the processing time of a task depends on the ratio of the total computing power required to complete the task to the total computing power provided by the DTN-DU nodes, the smaller the ratio, the shorter the processing time, that is:

[0101] (2)

[0102] The constraint Q is equivalent to:

[0103] (3)

[0104] In the above formula (3) This is the third constraint sub-condition in the embodiments of this application.

[0105] In this embodiment of the application, each group of second DTN units in the N4 group of second DTN units includes one or more second DTN units; N4 is an integer greater than or equal to 1.

[0106] In practical applications, DTN-CU uses Find the global state graph to obtain That is, all independent subsets in the global state graph G that meet the requirements, namely, all minimum DTN-DU unit combination strategies that satisfy the task type, accuracy, and total computing power requirements. The set. In gNB, there are multiple DTN-DU combination strategies that meet the computing power and task requirements. In this embodiment, the combination with the lowest task processing time is selected. .

[0107] Regarding step S405 above, in practical applications, for all independent subsets in the selected global state graph G that meet the requirements, the optimization objective is to minimize the connectivity of each DTN-DU within the combination. The optimal combination is obtained from all independent subsets, and computational power allocation weights are learned based on Graph Convolution Neural Networks (GCN) for task deconstruction. For specific implementation details, please refer to [link / reference]. Figure 5 .

[0108] In some implementations, step S405 above includes the following steps:

[0109] S4051: Transform the third constraint sub-condition into a fourth constraint sub-condition;

[0110] S4052: input the global state graph, the computing power vector corresponding to each second DTN unit in the global state graph, and the computing power requirement of the target task into a graph convolutional neural network defined based on the global dynamic graph, combine the fourth constraint sub-condition to obtain a group of second DTN units with the minimum connectivity of the network topology in the N4 groups of second DTN units and the computing power allocation weight of each second DTN unit in the group of second DTN units with the minimum connectivity of the network topology;

[0111] S4053: determine the group of second DTN units with the minimum connectivity of the network topology as the DTN unit set for processing the target task corresponding to the task request.

[0112] In the embodiments of the application, the fourth constraint sub-condition is to select a group of second DTN units with the minimum reciprocal value of the computing power value in the N4 groups of second DTN units as the DTN unit set; wherein for each group of second DTN units in the N4 groups of second DTN units, the computing power value of the group of second DTN units is the cumulative value of the product of the computing power allocation value of each second DTN unit included in the group of second DTN units and the computing power allocation weight of each second DTN unit.

[0113] Considering that in actual distributed processing, the parallel task processing time will be affected by the network topology G, the actual DTN-DU node For example, if has a higher connectivity (i.e. more connections) with other DTN-DU nodes, considering that allocating a higher load to this node may hinder the execution efficiency of other parallel tasks, therefore, it is considered to preferentially allocate a lighter operation load to this node. If is a node with a lower connectivity (i.e. fewer connections) with other DTN-DU nodes, allocating a heavier load to this node tends not to affect the execution of other parallel tasks, therefore, it is considered to preferentially allocate a heavier operation (computing power) load to this node. Based on this constraint condition can be further optimized as:

[0114] (4)

[0115] wherein, is the fourth constraint sub-condition, indicating that the task processing time of the DTN-DU node combination satisfying the constraint condition is the minimum.

[0116] Each parameter in formula (4) represents as follows:

[0117] (5)

[0118] (6)

[0119] represents the computing power allocated by the DTN-DU node v. A weight is assigned to the computing power corresponding to the node v, which is affected by the node connectivity and can be obtained by defining a graph neural network.

[0120] (7)

[0121] U is a vector composed of the computing power corresponding to the node v. is a graph convolution network (GCN) defined based on the global dynamic graph G, represents Task-oriented learning parameters.

[0122] The UE demand constraint condition Q (task type, accuracy, total computing power demand) is satisfied, and the task processing time is optimal DTN-DU combination strategy The strategy combination can be obtained by optimizing formula 4, and W represents the computing power allocation weight corresponding to the strategy combination.

[0123] The technical scheme of the embodiments of the present application introduces a global state graph, specifically constructs a global state graph according to the node information of a plurality of DTN-DUs, and takes the parameters in the user's task request as a constraint condition of the global state graph. The global state graph containing a plurality of DTN-DUs is searched by using the constraint condition converted from the task request, and one or more DTN-DU combinations satisfying the task processing computing power and time consumption are obtained. In addition, the present application also selects the DTN-DU combination with the minimum connectivity from one or more DTN-DU combinations to ensure that the time consumption of the selected target task for processing the task request is the minimum.

[0124] Figure 6 The flowchart of the task processing method provided by the embodiments of the present application Figure 3 , applied to a first DTN unit, comprising the following steps:

[0125] S601: The target task corresponding to the user's task request is divided into N5 subtasks;

[0126] S602: The N5 subtasks are distributed to each second DTN unit in the DTN unit set for processing;

[0127] S603: Receive the task processing result of the target task sent by the third DTN unit.

[0128] In the above step S601, N5 is an integer greater than or equal to 1.

[0129] In some embodiments, the above step S601 comprises:

[0130] split the target task into N5sub-tasks according to the computing power values of each second DTN unit in the DTN unit set; wherein the computing power value of each second DTN unit in the DTN unit set is obtained according to the computing power allocation value and the computing power allocation weight of the second DTN unit.

[0131] In some embodiments, the above step S602 comprises: distributing the N5sub-tasks to each second DTN unit in the DTN unit set according to the computing power value of each DTN unit in the DTN unit set for processing.

[0132] In practical applications, when the DTN-CU splits the task requested by the user, the computing power allocation value and the computing power allocation weight of each DTN-DU node in the optimal DTN-DU combination obtained in the above step S405 are combined to deconstruct and distribute the target task to each DTN-DU node in the optimal DTN-DU combination.

[0133] In the embodiments of the present application, for the DTN-DU combination finally selected for processing the task, each DTN-DU in the combination completes its own sub-task as needed.

[0134] In practical applications, different tasks can be processed in the following cases:

[0135] Case 1: The sub-tasks processed by each DTN-DU are independent, and each DTN-DU independently processes the sub-task allocated to it;

[0136] Case 2: The sub-tasks processed by each DTN-DU are interrelated and can be cooperatively calculated, federated verified, etc. In this case, each DTN-CU informs each DTN-DU of the global state graph information, and each DTN-DU can quickly find a DTN-DU suitable for cooperating in task cooperative calculation and federated verification according to the global state graph information, and then sends the cooperative calculation task to the corresponding DTN-DU for processing.

[0137] In the embodiments of the present application, for each second DTN unit in the DTN unit set, the second DTN unit sends the processing result of the sub-task to a third DTN unit after processing the sub-task distributed by the first DTN unit.

[0138] In the embodiments of the present application, the third DTN unit is deployed in the CU of the base station. In combination with Figure 2 , the third DTN unit is a DTN-CU deployed in the gNB-CU-UP.

[0139] ​In the embodiment of the present application, the task processing result is obtained by the third DTN unit aggregating the processing results of the sub-tasks sent by each second DTN unit in the set of DTN units.

[0140] As shown in Figure 2 , after completing the respective sub-tasks, the DTN-DU in the gNB-DU sends the processing results of the completed sub-tasks to the DTN-CU in the gNB-CU-UP, and the DTN-CU in the gNB-CU-UP aggregates and analyzes the processing results of the sub-tasks sent by each DTN-DU to generate a twin task report.

[0141] In actual application, after generating the twin task report, the DTN-CU in the gNB-CU-UP can feed back the task processing result to the DTN-CU in the gNB-CU-CP and feed back the twin task report to the UE. The DTN-CU in the gNB-CU-CP judges whether policy adjustment or parameter update configuration is needed according to the feedback task processing result, and if needed, performs policy translation and parameter configuration through the gNB-CU-CP as needed.

[0142] The embodiment of the present application can reduce the digital twin task processing delay from three aspects of UE request delay, transmission delay caused by data collection, and configuration delay caused by saving policy distribution, by sinking the digital twin network unit to the gNB and deploying in a distributed manner, thereby improving the processing efficiency of the digital twin task.

[0143] Figure 7 The flow of the task processing method provided by the embodiment of the present application is shown in Figure 4 , which is applied to a target second DTN unit and includes the following steps:

[0144] S701: Send the node information of the target second DTN unit to a first DTN unit.

[0145] In the embodiment of the present application, the target second DTN unit is deployed in the DU of the base station.

[0146] In combination with Figure 2 , the target second DTN unit is a DTN-DU deployed in the gNB-DU; more specifically, the target second DTN unit is a DTN-DU unit for processing a sub-task in a target task.

[0147] In the embodiments of the present application, the node information is used for the first DTN unit to determine a DTN unit set for processing a target task corresponding to a task request according to parameter information included in the task request sent by a user and node information of each second DTN unit in the N1 second DTN units; the first DTN unit is deployed in a CU of a base station, and the DTN unit set belongs to the N1 second DTN units; the DTN unit set includes N3 second DTN units; N1 is an integer greater than or equal to 1; N2 and N3 are integers less than or equal to N1.

[0148] In the embodiments of the present application, the first DTN unit is deployed in a CU of a base station; and the N1 second DTN units are deployed in N2 distributed units (DUs) of the base station.

[0149] In combination Figure 2 , the first DTN unit is a DTN-CU in a gNB-CU-CP; and the second DTN units are DTN-DUs deployed in respective gNB-DUs. The first DTN unit and the N1 second DTN units dynamically form a digital twin network system.

[0150] In some embodiments, the node information includes node state information and node connectivity information; wherein,

[0151] The node state information includes one or more of the following: number of queued tasks, server type, memory occupancy rate, available computing power of the server, supported model type, supported model accuracy, data volume, data type;

[0152] The node connectivity information includes one or more of the following: relationship information between the first DTN unit and each second DTN unit in the N1 second DTN units, and relationship information between each second DTN unit in the N1 second DTN units; the relationship information includes one or more of the following: physical connection relationship, task connectivity relationship.

[0153] In actual application, each DTN-DU reports node information of each DTN-DU to a DTN-CU in a Key-Value format through an F1 interface, and the node information includes but is not limited to [{number of queued tasks: n1}; {server type: CPU / GPU / DPU / TPU / NPU / BPU…}; {CPU / GPU / TUP / NPU… occupancy rate: n2%}; {memory occupancy rate: n3%}; {available computing power of the server: n4}; {supported AI model type: fault detection type, state prediction type, data enhancement type, parameter optimization type…}; {supported AI model accuracy: n5%}; {data volume: N}; {data type: Tensor, feature name…}] and the like.

[0154] In the embodiments of the present application, the task request includes parameter information indicating the task demand of the target task. In some embodiments, the parameter information in the task request includes one or more of the following: task type, task latency, and task accuracy.

[0155] In actual application, the UE initiates a task request to the DTN-CU of the gNB-CU-CP, and the task request parameters include but are not limited to [Task Type, Maximum latency, Task Accuracy], Task Type 1 is parameter optimization, Task Type 2 is decision generation, Task Type 3 is policy pre-verification, Task Type 4 is performance simulation, Task Type 5 is state, behavior, etc. prediction, Task Type 6 is data perception processing, Task Type 7 is visualization, if the UE initiates one request, it can be represented as Task Type A, and if the UE initiates multiple requests, it can be represented as Task Type A&B&C; wherein Maximum latency represents the maximum latency of the UE, which is a constraint condition for generating a task distribution strategy; Task Accuracy is only set for decision tasks, especially intelligent decision-making based on artificial intelligence (AI).

[0156] In actual application, the DTN-CU selects the optimal DTN-DU combination from N1 DTN-DUs for the construction of a dynamic digital twin network according to the parameter information in the task request and the node information of each DTN-DU. After selecting the optimal DTN-DU combination, the DTN-CU splits the task and distributes it to the DTN-DUs within the corresponding construction range for task processing.

[0157] The embodiments of the present application can reduce the digital twin task processing delay from three aspects: UE request delay, data acquisition caused transmission delay, and configuration delay caused by saving strategy, by sinking the digital twin network unit to the gNB and deploying it in a distributed manner, thereby improving the processing efficiency of the digital twin task.

[0158] Figure 8 Flowchart of the task processing method provided by the embodiments of the present application Figure 5 , applied to a target second DTN unit, comprising the following steps:

[0159] S801: sending node information of the target second DTN unit to a first DTN unit;

[0160] S802: receiving a target sub-task of the target task distributed by the first DTN unit;

[0161] S803: processing the target subtask of the target task distributed by the first DTN unit to obtain a processing result of the target subtask;

[0162] S804: sending the processing result of the target subtask of the target task distributed by the first DTN unit to a third DTN unit.

[0163] In the step S802, the target subtask is a subtask in N5 subtasks distributed by the first DTN unit to the second DTN unit after the first DTN unit splits the target task into the N5 subtasks.

[0164] In actual application, the DTN-CU combines the computing power distribution values and the computing power distribution weights of each DTN-DU node in the optimal DTN-DU combination obtained in the step S405 to perform deconstruction and distribution of the target task on each DTN-DU node in the optimal DTN-DU combination when splitting the task requested by the user.

[0165] In the embodiment of the application, each DTN-DU in the DTN-DU combination finally selected for processing the task completes the respective subtask as needed.

[0166] In an embodiment, the step S803 includes:

[0167] performing task cooperative operation with the subtasks processed by the other second DTN units in the DTN unit set except the target second DTN unit to obtain the processing result of the target subtask.

[0168] In actual application, the task can be processed according to the following cases:

[0169] Case 1: the subtasks processed by each DTN-DU are independent, and each DTN-DU independently processes the assigned subtask;

[0170] Case 2: the subtasks processed by each DTN-DU are associated with each other and can be cooperatively calculated, federated verified, etc. In this case, each DTN-CU performs global state graph information notification on each DTN-DU, and each DTN-DU can quickly find a DTN-DU suitable for cooperating in task cooperative calculation and federated verification according to the global state graph information, and then sends the cooperative calculation task to the corresponding DTN-DU for processing.

[0171] ​In step S804, the third DTN unit is deployed in the CU of the base station, and is configured to aggregate the processing results of the sub-tasks sent by each second DTN unit in the set of DTN units to obtain the task processing result of the target task.

[0172] In the embodiments of the present application, for each second DTN unit in the set of DTN units, the second DTN unit sends the processing result of the sub-task to the third DTN unit after completing the processing of the sub-task distributed by the first DTN unit.

[0173] In the embodiments of the present application, the third DTN unit is deployed in the CU of the base station. In combination with Figure 2 , the third DTN unit is a DTN-CU deployed in the gNB-CU-UP.

[0174] In the embodiments of the present application, the task processing result is obtained by the third DTN unit aggregating the processing results of the sub-tasks sent by each second DTN unit in the set of DTN units.

[0175] As shown in Figure 2 , the DTN-DU in the gNB-DU sends the processing result of the completed sub-task to the DTN-CU in the gNB-CU-UP after completing the respective sub-task, and the DTN-CU in the gNB-CU-UP aggregates and analyzes the processing results of the sub-tasks sent by each DTN-DU to generate a twin task report.

[0176] In actual applications, after generating the twin task report, the DTN-CU in the gNB-CU-UP can feed back the task processing result to the DTN-CU in the gNB-CU-CP and feed back the twin task report to the UE. The DTN-CU in the gNB-CU-CP determines whether policy adjustment or parameter update configuration is needed according to the feedback task processing result, and if needed, performs policy translation and parameter configuration through the gNB-CU-CP.

[0177] The embodiments of the present application can reduce the digital twin task processing delay from three aspects of UE request delay, transmission delay caused by data collection, and configuration delay caused by saving policy distribution, by sinking the digital twin network unit to the gNB and deploying in a distributed manner, thereby improving the processing efficiency of the digital twin task.

[0178] Figure 9 The flowchart of the task processing method provided by the embodiments of the present application is shown in Figure 6 , which is applied to a third DTN unit, and the method comprises:

[0179] S901: receiving the processing result of the subtask of the target task sent by each second DTN unit in the DTN unit set;

[0180] S902: aggregating the processing result of the subtask of the target task sent by each second DTN unit in the DTN unit set to obtain the task processing result of the target task.

[0181] In the embodiment of the application, the third DTN unit is deployed in the CU of the base station. In combination with Figure 2 , the third DTN unit is a DTN-CU deployed in the gNB-CU-UP.

[0182] The subtask is a subtask in the N5 subtasks distributed to the second DTN unit after the first DTN unit splits the target task into N5 subtasks.

[0183] In actual application, the DTN-CU, in combination with the computing power distribution value and the computing power distribution weight of each DTN-DU node in the selected optimal DTN-DU combination, performs deconstruction and distribution of the target task on each DTN-DU node in the optimal DTN-DU combination when splitting the task requested by the user.

[0184] In the embodiment of the application, for the DTN-DU combination finally selected for processing the task, each DTN-DU in the combination completes its own subtask as needed.

[0185] In actual application, the task can be processed according to the following cases:

[0186] Case 1: The subtasks processed by each DTN-DU are independent, and each DTN-DU independently processes the assigned subtask;

[0187] Case 2: The subtasks processed by each DTN-DU are interrelated and can be cooperatively calculated, federated verified, etc. In this case, each DTN-CU performs global state graph information announcement to each DTN-DU, and each DTN-DU can quickly find a DTN-DU suitable for cooperating in task cooperative calculation and federated verification according to the global state graph information, and then sends the cooperative calculation task to the corresponding DTN-DU for processing.

[0188] In the embodiment of the application, the task processing result is obtained by the third DTN unit aggregating the processing result of the subtask sent by each second DTN unit in the DTN unit set.

[0189] As Figure 2As shown, the DTN-DU in the gNB-DU sends the processing result of the completed subtask to the DTN-CU in the gNB-CU-UP after completing the respective subtask, and the DTN-CU in the gNB-CU-UP analyzes the processing result of the subtask sent by each DTN-DU, and generates a twin task report.

[0190] In some embodiments, the third DTN unit sends the task processing result of the target task to the first DTN unit and / or the terminal after obtaining the task processing result.

[0191] In actual application, after generating the twin task report, the DTN-CU in the gNB-CU-UP can feed back the task processing result to the DTN-CU in the gNB-CU-CP and feed back the twin task report to the UE. The DTN-CU in the gNB-CU-CP judges whether policy adjustment or parameter update configuration is needed according to the feedback task processing result, and if needed, performs policy translation and configures parameters through the gNB-CU-CP.

[0192] The embodiment of the present application can reduce the digital twin task processing delay from three aspects of UE request delay, transmission delay caused by data collection, and configuration delay caused by saving policy, by sinking the digital twin network unit to the gNB and deploying in a distributed manner, thereby improving the processing efficiency of the digital twin task.

[0193] Figure 10 A digital twin task request and processing flow.

[0194] Based on Figure 2 As shown in the digital twin network deployment mode, to complete the deployment of the digital twin network and the processing of the digital twin task, it is necessary to first implement a twin task chain building process, and perform relevant information registration and capability subscription.

[0195] The twin task chain building, relevant information registration and capability subscription need to be implemented through network element information registration. The network element information registration includes: introducing a network repository function (NRF) network element on the radio access network (RAN) side, providing network function registration and discovery of the radio side network element, which can enable network functions to discover each other and communicate through API.

[0196] The introduction of NRF on the RAN side can be realized in two ways: one is to sink the existing standardized NRF in the architecture, keep the function unchanged, add new interfaces between the NRFs on the RAN side, support the registration, discovery and other functions of the newly added network elements in the RAN, and increase the corresponding standardized interfaces; the other is to newly introduce a network element with NRF function on the RAN side to perform the registration and discovery functions of the network element.

[0197] The network element information registration includes steps 1.1 to 1.2 in the above. Figure 10

[0198] 1.1, the gNB registers with the NRF, and the registration information includes DTN-CU and DTN-DU information.

[0199] The registration information can include [IP Address, DTN-type, Support Service Type, ServiceCapability] and the like; the DTN-type includes three types: DTN-CU in gNB-CU-CP, DTN-CU in gNB-CU-UP, and DTN-DU; the Support Service Type at least includes data sensing / processing, policy generation, parameter optimization, configuration / policy simulation verification, state behavior prediction, network visualization and the like; the Service Capability at least includes AI capability, operation parallelism capability, storage capability and the like.

[0200] 1.2, each DTN-DU registers the capability with the DTN-CU.

[0201] The registration information includes [IP Address, Computing Capability, Process Type] and the like, and each DTN-DU can actively or passively report the information to the DTN-CU, the active way can be active reporting in the same frequency or variable frequency, and the passive way can be reporting when the DTN-CU sends a reporting request to the DTN-DU. The Computing Capability includes server types (CPU / GPU / DPU / TUP / NPU / BPU and the like), and the Process Type at least includes data sensing / processing, policy generation, parameter / configuration document / policy simulation verification and the like (the fields can be a subset A&B&C or the whole set ALL);

[0202] 2, the UE sends a query twin service subscription object message to the NRF.

[0203] 3, the NRF returns the appropriate DTN-CU IP to the UE according to the registration information.

[0204] ​4. UE sends twin task request to DTN-CU in gNB-CU-CP, such as parameter optimization, simulation verification, etc.

[0205] 5. DTN-DU in each gNB-DU reports data in Key-Value format; the reported data includes the self-node information of each node.

[0206] 6. DTN-CU constructs and maintains a global state diagram data table, converts UE request into constraint conditions, and searches for optimal DTN-DU by Min(Linkdistance) / { (= Service Type) & (0 < deal time < Maximum latency) & (>= Task Accuracy)} to automatically organize DTN environment and intelligently distribute twin tasks to DTN-DU.

[0207] 7. DTN-CU notifies each DTN-DU of the global state diagram data table. Each DTN-DU can quickly find suitable DTN-DU for task coordination and federation verification according to the global state diagram information, and then sends the coordination task to the corresponding DTN-DU for processing.

[0208] 8. DTN-DU completes the distributed task on demand, and each DTN-DU can perform coordinated calculation and federation verification. At the same time, the latest data is synchronized in real time from the network file system (NFS, Network File System) / operation management and maintenance (OAM, Operation Administration and Maintenance) network element to ensure the accuracy of the task.

[0209] 9. Data synchronization on demand.

[0210] Here, to ensure the real-time and accuracy of DTN-DU, the latest data can be selectively obtained from NFS / OAM in real time, and the twin environment can be updated in real time and dynamically to ensure the accuracy of the task to the greatest extent. A data interaction interface between the new gNB and NFS / OAM is added.

[0211] 10. Task result feedback.

[0212] DTN-DU feeds back the processed task result to DTN-CU in gNB-CU-UP.

[0213] 11. User plane DTN-CU performs result summary analysis and generates a twin task report.

[0214] DTN-CU in gNB-CU-UP performs result summary analysis and generates a twin task report.

[0215] 12、Result feedback.

[0216] The DTN-CU in the gNB-CU-UP feeds back the result to the DTN-CU in the gNB-CU-CP, and feeds back the twin task report to the UE.

[0217] 13、The DTN-CU performs policy translation on demand, and configures parameters through the gNB-CU-CP.

[0218] The DTN-CU judges whether policy adjustment or parameter update configuration is needed according to the feedback result, and performs policy translation on demand and configures parameters through the gNB-CU-CP if needed.

[0219] 14、Send task report.

[0220] The DTN-CU in the gNB-CU-UP feeds back the result to the DTN-CU in the gNB-CU-CP, and feeds back the twin task report to the UE.

[0221] Figure 11 It is a schematic diagram of a task cooperation process; taking Co-Training model training as an example: two models with the same structure but different features are trained in two DTN-DUs at the same time, and the labels output by the two are input into each other's model, and are corrected by each other, and finally converge into one model, solving the problem of improving model generalization based on a small amount of labeled data, based on the mutual cooperation and computing power between different DTN-DUs, greatly improving the model convergence speed, thereby improving the decision-making efficiency.

[0222] Figure 11 The model training process in the above includes: training Model1 in DTN-DU A and training Model2 in DTN-DU B, and the total feature is , and the Feature Set of the labeled data is , , , as the input of Model1, as the input of Model2, and each completes model training. and represent the features corresponding to and respectively, but without a labeled data set, and are added to Model1 and Model2 respectively to obtain the predicted result = Model1( ), = Model2( ), and DTN-DU A feeds back Send to DTN-DU B, add as a pseudo label , DTN-DU B will Send to DTN-DU A, add as a pseudo label , continue to train the model respectively, take the distance between the two output results of the same sample as the model training termination condition, if the threshold is reached, end the model training, and get the final unified model for intelligent decision-making.

[0223] Figure 11 The sub-tasks processed by each DTN-DU are related to each other, can be cooperatively calculated, and each DTN-DU can quickly find a DTN-DU suitable for task cooperative calculation according to the global state diagram information, and then send the cooperative calculation task to the corresponding DTN-DU for processing.

[0224] Figure 12 Structure of a task processing device provided by an embodiment of the application Figure 1 , applied to a first DTN unit deployed in a CU of a base station, the device comprises:

[0225] A first receiving unit 1201 configured to receive a task request sent by a user; the task request comprises parameter information indicating a task requirement of a target task;

[0226] A determining unit 1202 configured to determine, according to the parameter information included in the task request and node information of each second DTN unit in N1 second DTN units, a DTN unit set for processing a target task corresponding to the task request; N1 is an integer greater than or equal to 1.

[0227] The N1 second DTN units are deployed in N2 distribution units DUs of a base station, and the DTN unit set comprises N3 second DTN units; the DTN unit set belongs to the N1 second DTN units; N2 and N3 are integers less than or equal to N1.

[0228] In some embodiments, the parameter information comprises one or more of the following: task type, task waiting time delay, and task accuracy.

[0229] In some embodiments, the node information comprises node state information and node connectivity information; wherein,

[0230] The node state information comprises one or more of the following: number of queued tasks, server type, memory occupancy rate, available computing power of the server, supported model type, supported model accuracy, data volume, and data type.

[0231] The node connection information includes one or more of the following: relationship information between the first DTN unit and each of the N1 second DTN units, relationship information between each of the N1 second DTN units; the relationship information includes one or more of the following: a physical connection relationship, a task connection relationship.

[0232] In some embodiments, the determination unit is configured to construct a global state graph according to node information of each of the N1 second DTN units; convert the parameter information into a constraint condition of the global state graph; search the global state graph using the constraint condition to obtain N4 groups of second DTN units that satisfy the constraint condition.

[0233] The constraint condition includes a first constraint sub-condition, a second constraint sub-condition, and a third constraint sub-condition; the first constraint sub-condition is used to constrain the task type of the second DTN unit set in the searched global state graph to satisfy the task type of the task request; the second constraint sub-condition is used to constrain the task accuracy of the second DTN unit set in the searched global state graph to be greater than or equal to the task accuracy in the task request; the third constraint sub-condition is used to constrain the task processing time of the second DTN unit set in the searched global state graph to be less than or equal to the task waiting time delay in the task request; each group of second DTN units in the N4 groups of second DTN units includes one or more second DTN units; N4 is an integer greater than or equal to 1.

[0234] In some embodiments, the determination unit is configured to use the task type of the second DTN unit set in the searched global state graph to satisfy the task type in the parameter information as a first constraint sub-condition; use the task accuracy of the second DTN unit set in the searched global state graph to be greater than or equal to the task accuracy in the parameter information as a second constraint sub-condition; map the task waiting time delay parameter in the parameter information to the computing power requirement of the target task, and use the total computing power that the second DTN unit set in the searched global state graph can provide to be greater than the computing power requirement of the target task as a third constraint sub-condition.

[0235] In some embodiments, the determination unit is configured to select the DTN unit set from the N4 groups of second DTN units with the minimum connectivity of the network topology of each of the N4 groups of second DTN units as the target.

[0236] In some embodiments, the determining unit is configured to convert the third constraint sub-condition into a fourth constraint sub-condition; the fourth constraint sub-condition is to select a group of second DTN units with the minimum reciprocal value of the computing power value from the N4 groups of second DTN units as the DTN unit set; wherein, for each group of second DTN units from the N4 groups of second DTN units, the computing power value of the group of second DTN units is the cumulative value of the product of the computing power allocation value of each second DTN unit included in the group of second DTN units and the computing power allocation weight of each second DTN unit; the global state graph, the computing power vector corresponding to each second DTN unit in the global state graph, and the computing power demand of the target task are input into a graph convolutional neural network defined based on the global dynamic graph, and the fourth constraint sub-condition is combined to obtain a group of second DTN units with the minimum network topology connectivity from the N4 groups of second DTN units and the computing power allocation weight of each second DTN unit in the group of second DTN units with the minimum network topology connectivity; the group of second DTN units with the minimum network topology connectivity is determined as the DTN unit set for processing the target task corresponding to the task request.

[0237] In some embodiments, the apparatus further comprises:

[0238] a splitting unit configured to split the target task into N5 sub-tasks; N5 is an integer greater than or equal to 1;

[0239] a distribution unit configured to distribute the N5 sub-tasks to each second DTN unit in the DTN unit set for processing.

[0240] In some embodiments, the splitting unit is configured to split the target task into N5 sub-tasks according to the computing power value of each second DTN unit in the DTN unit set; wherein, the computing power value of each second DTN unit in the DTN unit set is obtained according to the computing power allocation value and the computing power allocation weight of the second DTN unit;

[0241] the distribution unit is configured to distribute the N5 sub-tasks to each second DTN unit in the DTN unit set according to the computing power value of each DTN unit in the DTN unit set for processing.

[0242] In some embodiments, for each second DTN unit in the DTN unit set, after processing the sub-task distributed by the first DTN unit, the second DTN unit sends the processing result of the sub-task to a third DTN unit, and the third DTN unit is deployed in a central unit CU of a base station; the apparatus further comprises:

[0243] a second receiving unit, configured to receive a task processing result of the target task sent by the third DTN unit;

[0244] The task processing result is obtained by the third DTN unit by aggregating processing results of subtasks sent by each second DTN unit in the set of DTN units.

[0245] Those skilled in the art should understand that, Figure 12 The implementation functions of each unit in the task processing apparatus shown can be understood with reference to the foregoing related descriptions of the task processing method on the first DTN unit side. Figure 12 The functions of each unit in the task processing apparatus shown can be implemented by a program running on a processor, or by a specific logic circuit.

[0246] Figure 13 Structure and composition of a task processing apparatus provided by an embodiment of the present application Figure 2 applied to a target second DTN unit deployed in a DU of a base station, the apparatus comprising:

[0247] a first sending unit 1301 configured to send node information of the target second DTN unit to a first DTN unit;

[0248] The node information is used by the first DTN unit to determine, according to parameter information included in a task request sent by a user and node information of each second DTN unit in N1 second DTN units, a set of DTN units for processing a target task corresponding to the task request; the first DTN unit is deployed in a CU of a base station, the set of DTN units belongs to the N1 second DTN units; the set of DTN units includes N3 second DTN units; N1 is an integer greater than or equal to 1; N2 and N3 are integers less than or equal to N1.

[0249] In some embodiments, in the case where the target second DTN unit belongs to the set of DTN units, the apparatus further comprises:

[0250] a third receiving unit configured to receive a target subtask of the target task distributed by the first DTN unit; the target subtask is a subtask in N5 subtasks of the target task distributed by the first DTN unit to the second DTN units after the first DTN unit splits the target task into the N5 subtasks;

[0251] a processing unit configured to process the target subtask of the target task distributed by the first DTN unit to obtain a processing result of the target subtask.

[0252] In some embodiments, the processing unit is configured to perform task coordination operation with the sub-tasks processed by the other second DTN units in the set of DTN units except the target second DTN unit, to obtain a processing result of the target sub-task.

[0253] In some embodiments, the apparatus further comprises:

[0254] a second sending unit configured to send the processing result of the target sub-task of the target task distributed to the first DTN unit to a third DTN unit;

[0255] The third DTN unit is deployed in a CU of a base station, and is configured to aggregate the processing results of the sub-tasks sent by the second DTN units in the set of DTN units to obtain a task processing result of the target task.

[0256] Those skilled in the art should understand that, Figure 13 The implementation functions of the units in the task processing apparatus can be understood with reference to the foregoing descriptions of the task processing method on the target second DTN unit side. Figure 13 The functions of the units in the task processing apparatus can be implemented by programs running on a processor, or by specific logic circuits.

[0257] Figure 14 A structural composition of a task processing apparatus provided by an embodiment of the present application is shown in Figure 3 The apparatus is applied to a third DTN unit, which is deployed in a CU of a base station, and comprises:

[0258] a fourth receiving unit 1401 configured to receive the processing results of the sub-tasks of the target task sent by the second DTN units in the set of DTN units;

[0259] an aggregation unit 1402 configured to aggregate the processing results of the sub-tasks of the target task sent by the second DTN units in the set of DTN units to obtain a task processing result of the target task.

[0260] In some embodiments, the apparatus further comprises:

[0261] a third sending unit configured to send the task processing result to the first DTN unit and / or a terminal; the first DTN unit is deployed in a CU of a base station.

[0262] Those skilled in the art should understand that, Figure 14 The implementation functions of the units in the task processing apparatus can be understood with reference to the foregoing descriptions of the task processing method. Figure 14The functions of the units in the task processing apparatus shown can be implemented by programs running on the processor, or by specific logic circuits.

[0263] The embodiment of the present application further provides a task processing device. Figure 15 A hardware structure of a task processing device provided by the embodiment of the present application is shown in a schematic diagram as follows. Figure 15 The task processing device includes a communication component 1503 for data transmission, at least one processor 1501, and a memory 1502 for storing computer programs capable of running on the processor 1501. The various components in the terminal are coupled together through a bus system 1504. It can be understood that the bus system 1504 is used to realize the connection communication between the components. The bus system 1504 includes not only a data bus, but also a power bus, a control bus, and a status signal bus. However, for the purpose of clear illustration, all the buses are marked as the bus system 1504 in the Figure 15 .

[0264] The processor 1501 executes the computer programs to at least perform Figures 3 to 6 , or Figures 7 to 8 , or Figure 9 the steps of the method shown.

[0265] It is to be understood that the memory 1502 can be a volatile memory or a nonvolatile memory, and can also include both volatile and nonvolatile memory. Among them, the nonvolatile memory can be a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a ferromagnetic random access memory (FRAM), a Flash Memory, a magnetic surface memory, an optical disc, or a Compact Disc Read-Only Memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a Random Access Memory (RAM) used as an external cache. By way of example and not limitation, many forms of RAM can be used, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDR SDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Sync Link Dynamic Random Access Memory (SLDRAM), Direct Rambus Random Access Memory (DRRAM).The memory 1502 described in the embodiments of the present application is intended to include, but not limited to, these and any other suitable types of memories.

[0266] The method disclosed in the embodiments of the present application can be applied in the processor 1501 or implemented by the processor 1501. The processor 1501 can be an integrated circuit chip having a processing capability of signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of hardware in the processor 1501 or the instruction in the form of software. The processor 1501 described above can be a general processor, a DSP, or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The processor 1501 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general processor can be a microprocessor or any conventional processor, etc. In combination with the steps of the method disclosed in the embodiments of the present application, the execution can be directly completed by the hardware decoding processor or by the combination of hardware and software modules in the decoding processor. The software module can be located in the storage medium, which is located in the memory 1502. The processor 1501 reads the information in the memory 1502 and combines the hardware to complete the steps of the above method.

[0267] In the exemplary embodiments, the task processing device can be implemented by one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), FPGAs, general-purpose processors, controllers, MCUs, microprocessors (Microprocessors), or other electronic elements, for executing the above-mentioned task processing method.

[0268] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the program is characterized in that, when executed by a processor, it is used for executing at least the steps of the method shown in Figures 3 to 6 , or Figures 7 to 8 , or Figure 9 The computer readable storage medium can be a memory, and the memory can be the memory 1502 as shown in Figure 15 .

[0269] The technical solutions described in the embodiments of the present application can be combined arbitrarily without conflict.

[0270] In several embodiments provided in the present application, it should be understood that the disclosed method and intelligent device can be implemented in other manners. The embodiments described above are merely exemplary, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling, or direct coupling or communication connection between any two components can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0271] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place or distributed on a plurality of network units; part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0272] In addition, each functional unit in each embodiment of the present application can be integrated into a second processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional unit.

[0273] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A task processing method, characterized in that, The method, applied to a first digital twin network (DTN) unit deployed in a base station's central unit (CU), includes: Receive a task request sent by a user; the task request includes parameter information indicating the task requirements of the target task; The set of DTN units for processing the target task corresponding to the task request is determined based on the parameter information included in the task request and the node information of each of the N1 second DTN units; N1 is an integer greater than or equal to 1. Wherein, the N1 second DTN units are deployed in the N2 distribution units (DU) of the base station, and the DTN unit set includes N3 second DTN units; the DTN unit set belongs to the N1 second DTN units; N2 and N3 are both integers less than or equal to N1; The step of determining the set of DTN units for processing the target task corresponding to the task request based on the parameter information included in the task request and the node information of each of the N1 second DTN units includes: A global state diagram is constructed based on the node information of each of the N1 second DTN units; The parameter information is then converted into constraints for the global state diagram. The global state diagram is searched using the constraints to obtain N4 groups of second DTN cells that satisfy the constraints. With the goal of minimizing the connectivity of the network topology of each second DTN unit in the N4 group of second DTN units, the set of DTN units is selected from the N4 group of second DTN units.

2. The method according to claim 1, characterized in that, The parameter information includes one or more of the following: task type, task waiting delay, and task accuracy.

3. The method according to claim 1, characterized in that, The node information includes node status information and node connectivity information; wherein... The node status information includes one or more of the following: number of tasks in the queue, server type, memory usage, available computing power of the server, supported model types, supported model accuracy, data volume, and data type. The node connectivity information includes one or more of the following: relationship information between the first DTN unit and each of the N1 second DTN units, and relationship information between each of the N1 second DTN units; the relationship information includes one or more of the following: physical connection relationship, task connectivity relationship.

4. The method according to claim 2, characterized in that, The constraints include a first constraint sub-condition, a second constraint sub-condition, and a third constraint sub-condition; the first constraint sub-condition is used to constrain the task type of the second DTN unit set in the global state graph being searched to satisfy the task type of the task request; the second constraint sub-condition is used to constrain the task accuracy of the second DTN unit set in the global state graph being searched to be greater than or equal to the task accuracy in the task request; the third constraint sub-condition is used to constrain the task processing time of the second DTN unit set in the global state graph being searched to be less than or equal to the task waiting latency in the task request; each of the N4 groups of second DTN units includes one or more second DTN units; N4 is an integer greater than or equal to 1.

5. The method according to claim 4, characterized in that, The constraint conditions for converting the parameter information into the global state diagram include: The task type of the second DTN unit set in the searched global state graph that satisfies the task type in the parameter information is taken as the first constraint sub-condition. The task accuracy of the second DTN cell set in the searched global state graph is greater than or equal to the task accuracy in the parameter information, which is taken as the second constraint sub-condition. The task waiting delay parameter in the parameter information is mapped to the computing power requirement of the target task, and the total computing power that the second DTN unit set in the searched global state graph can provide is greater than the computing power requirement of the target task is taken as the third constraint sub-condition.

6. The method according to claim 1, characterized in that, The step of selecting the DTN unit set from the N4 groups of second DTN units with the goal of minimizing the connectivity of the network topology of each second DTN unit in the N4 groups includes: The third constraint sub-condition is transformed into a fourth constraint sub-condition; the fourth constraint sub-condition is to select the group of second DTN units with the smallest reciprocal value of computing power value among the N4 groups of second DTN units as the DTN unit set; wherein, for each group of second DTN units in the N4 groups of second DTN units, the computing power value of the group of second DTN units is the sum of the products of the computing power allocation value of each second DTN unit included in the group of second DTN units and the computing power allocation weight of each second DTN unit; The global state graph, the computing power vector corresponding to each second DTN unit in the global state graph, and the computing power requirement of the target task are input into the graph convolutional neural network defined based on the global dynamic graph. Combined with the fourth constraint sub-condition, the group of second DTN units with the smallest network topology connectivity in the N4 groups of second DTN units and the computing power allocation weight of each second DTN unit in the group of second DTN units with the smallest network topology connectivity are obtained. The set of second DTN units with the lowest network topology connectivity is determined as the set of DTN units used to process the target task corresponding to the task request.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: The target task is divided into N5 subtasks; N5 is an integer greater than or equal to 1. The N5 subtasks are distributed to each of the second DTN units in the DTN unit set for processing.

8. The method according to claim 7, characterized in that, The step of dividing the target task into N5 sub-tasks includes: The target task is divided into N5 subtasks based on the computing power value of each second DTN unit in the DTN unit set; wherein, the computing power value of each second DTN unit in the DTN unit set is obtained based on the computing power allocation value and computing power allocation weight of the second DTN unit; Accordingly, the step of distributing the N5 subtasks to each of the second DTN units in the DTN unit set for processing includes: The N5 subtasks are allocated to each second DTN unit in the DTN unit set according to the computing power value of each DTN unit in the DTN unit set for processing.

9. The method according to claim 7, characterized in that, For each second DTN unit in the DTN unit set, after processing the subtasks distributed by the first DTN unit, the second DTN unit sends the processing result of the subtasks to a third DTN unit, the third DTN unit being deployed in the central unit (CU) of the base station. The method further includes: Receive the task processing result of the target task sent by the third DTN unit; The task processing result is obtained by the third DTN unit summarizing the processing results of the sub-tasks sent by each of the second DTN units in the DTN unit set.

10. A task processing method, characterized in that, The method, applied to a target second DTN unit deployed in a base station's distribution unit (DU), includes: Send the node information of the target second DTN unit to the first DTN unit; The node information is used by the first DTN unit to determine a set of DTN units for processing the target task corresponding to the task request, based on the parameter information included in the task request sent by the user and the node information of each of the N1 second DTN units. The first DTN unit is deployed in the CU of the base station, and the set of DTN units belongs to the N1 second DTN units. The set of DTN units includes N3 second DTN units. N1 is an integer greater than or equal to 1, and N2 and N3 are both integers less than or equal to N1. The DTN unit set is obtained by the first DTN unit constructing a global state graph based on the node information of each of the N1 second DTN units; converting the parameter information into constraints of the global state graph; searching the global state graph using the constraints to obtain N4 groups of second DTN units that satisfy the constraints; and selecting from the N4 groups of second DTN units with the goal of minimizing the connectivity of the network topology of each of the N4 groups of second DTN units.

11. The method according to claim 10, characterized in that, If the target second DTN unit belongs to the DTN unit set, the method further includes: Receive the target subtask of the target task distributed by the first DTN unit; wherein, the target subtask is a subtask among the N5 subtasks distributed by the first DTN unit to the second DTN unit after the target task is split into N5 subtasks; The target subtasks of the target task distributed by the first DTN unit are processed to obtain the processing results of the target subtasks.

12. The method according to claim 11, characterized in that, The process of processing the target subtasks of the target task distributed by the first DTN unit to obtain the processing results of the target subtasks includes: The task is collaboratively processed with the subtasks processed by other second DTN units in the DTN unit set, excluding the target second DTN unit, to obtain the processing result of the target subtask.

13. The method according to claim 11 or 12, characterized in that, The method further includes: The processing results of the target sub-tasks of the target task distributed by the first DTN unit are sent to the third DTN unit; The third DTN unit is deployed in the CU of the base station. The third DTN unit is used to summarize the processing results of the sub-tasks sent by each of the second DTN units in the DTN unit set to obtain the task processing result of the target task.

14. A task processing method, characterized in that, The method, applied to a third DTN unit deployed in the CU of a base station, includes: Receive the processing results of the subtasks of the target task sent by each second DTN unit in the DTN unit set; The task processing result of the target task is obtained by summarizing the processing results of the sub-tasks of the target task sent by each second DTN unit in the DTN unit set; wherein, the DTN unit set is constructed by the first DTN unit based on the node information of each of the N1 second DTN units included in the task request sent by the user; the parameter information included in the task request sent by the user is converted into the constraints of the global state graph; the global state graph is searched using the constraints to obtain N4 groups of second DTN units that satisfy the constraints; and the N4 groups of second DTN units are selected with the goal of minimizing the connectivity of the network topology of each of the N4 groups of second DTN units.

15. The method according to claim 14, characterized in that, The method further includes: The task processing result is sent to the first DTN unit and / or the terminal; wherein the first DTN unit is deployed in the CU of the base station.

16. A task processing device, characterized in that, The device includes: a processor and a memory for storing a computer program capable of running on the processor, wherein the processor, when running the computer program, implements the method of any one of claims 1 to 9, or implements the method of any one of claims 10 to 13, or implements the method of any one of claims 14 to 15.

17. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 9, or the method of any one of claims 10 to 13, or the method of any one of claims 14 to 15.

18. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 9, or the method of any one of claims 10 to 13, or the method of any one of claims 14 to 15.

Images