Information adjustment methods, devices and related equipment

By adjusting information to meet service accuracy requirements, the problem of service accuracy caused by limited storage space is solved, and accuracy can be met when wireless transmission channels or user needs change.

CN123075451APending Publication Date: 2026-11-10VIVO SOFTWARE TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510591257.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-08
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

When a device has limited storage space and cannot store enough inference models or execution algorithms, it may fail to meet the user's service accuracy requirements when the wireless transmission channel or user needs change.

Method used

By acquiring the business requirements of the target business, information is adjusted to meet service accuracy, including model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, and communication reliability, thereby achieving information adjustment.

Benefits of technology

It meets the service accuracy requirements of users when wireless transmission channels or user needs change without requiring a large amount of storage space.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123075451A_ABST
    Figure CN123075451A_ABST
Patent Text Reader

Abstract

This application discloses an information adjustment method, apparatus, and related equipment, belonging to the field of communication technology. The information adjustment method of this application includes: a first network element obtaining a first service requirement of a target service; the first network element adjusting first information according to the first service requirement, wherein the first service requirement includes a first service precision, and the first information includes at least one of the following: model precision, calculation precision, data precision, artificial intelligence (AI) processing precision, AI inference precision, and communication reliability information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to an information adjustment method, apparatus, and related equipment. Background Technology

[0002] When the wireless transmission channel changes or user needs change, the currently executed algorithm or inference model may be unable to meet the user's service accuracy requirements. In related technologies, the service accuracy can be met by changing the execution algorithm or the inference model. However, changing the execution algorithm or the inference model requires storing multiple inference models or multiple execution algorithms. But when the device's storage space is limited and cannot store enough inference models or execution algorithms, it will lead to the problem of not being able to meet the user's service accuracy requirements when the wireless transmission channel changes or user needs change. Summary of the Invention

[0003] This application provides an information adjustment method, apparatus, and related equipment, which can solve the problem that when the device's storage space is limited and cannot store enough inference models or execution algorithms, the service accuracy cannot be met when the wireless transmission channel changes or the user's needs change.

[0004] Firstly, an information adjustment method is provided, the method comprising:

[0005] The first network element acquires the first service requirement of the target service;

[0006] The first network element adjusts the first information according to the first service requirement, wherein the first service requirement includes the first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability information.

[0007] Secondly, an information adjustment device is provided, comprising:

[0008] The first receiving module is used to obtain the first business requirements of the target business.

[0009] The first processing module is used to adjust the first information according to the first business requirement, wherein the first business requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability information.

[0010] Thirdly, a network-side device is provided, comprising a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method described in the first aspect.

[0011] Fourthly, a network-side device is provided, including a processor and a communication interface, wherein the communication interface is used to acquire a first service requirement of a target service; the processor is used to adjust first information according to the first service requirement, wherein the first service requirement includes a first service precision, and the first information includes at least one of the following: model precision, calculation precision, data precision, artificial intelligence (AI) processing precision, AI inference precision, and communication reliability information.

[0012] Fifthly, a readable storage medium is provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0013] In a sixth aspect, a chip is provided, the chip including a processor and a communication interface coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0014] In a seventh aspect, a computer program / program product is provided, the computer program / program product being stored in a storage medium, the computer program / program product being executed by at least one processor to perform the steps of the method as described in the first aspect.

[0015] In this embodiment, the first network element acquires a first service requirement of the target service; the first network element adjusts first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability. In the above scheme, the first information is adjusted so that the adjusted first information can meet the first service accuracy in the first service requirement, thus meeting the user's service accuracy when the wireless transmission channel changes or user needs change, without occupying a large amount of storage space to store the inference model or algorithm. Attached Figure Description

[0016] Figure 1 This diagram illustrates the structure of a communication system to which embodiments of this application can be applied.

[0017] Figure 2 A flowchart illustrating the information adjustment method according to an embodiment of this application;

[0018] Figure 3One of the schematic diagrams illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0019] Figure 4 The second schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0020] Figure 5 The third schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0021] Figure 6 The fourth schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0022] Figure 7 Fifth schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0023] Figure 8 Sixth schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0024] Figure 9 Schematic diagram seven illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0025] Figure 10 Eighth schematic diagram illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0026] Figure 11 Schematic diagram nine illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0027] Figure 12 Schematic diagram ten illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0028] Figure 13 Schematic diagram eleven illustrating the analysis of factors affecting service accuracy in the embodiments of this application;

[0029] Figure 14 One of the interactive schematic diagrams illustrating the information adjustment method of this application embodiment;

[0030] Figure 15 The second interactive schematic diagram illustrating the information adjustment method of this application embodiment;

[0031] Figure 16 A schematic diagram of the modules of the information adjustment device according to an embodiment of this application;

[0032] Figure 17 A structural block diagram illustrating a communication device according to an embodiment of this application;

[0033] Figure 18 This is one of the structural block diagrams of the network-side device according to an embodiment of this application;

[0034] Figure 19 This is a second structural block diagram illustrating an embodiment of the network-side device of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0036] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, the scope of protection for "A or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. In addition, the terms "A and / or B," "at least one of A and B," and "at least one of A or B" also cover at least the above three scenarios. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0037] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc., in the instruction sent. An indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.

[0038] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G communication system.

[0039] Figure 1This diagram illustrates a block diagram of a wireless communication system applicable to embodiments of this application. The wireless communication system includes a terminal 11 and a network-side device 12. The terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home devices (home appliances with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game consoles, personal computers (PCs), ATMs, or self-service machines, etc. Wearable devices include: smartwatches, smart bracelets, smart earphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment. Network-side equipment 12 may include access network equipment or core network equipment, wherein access network equipment may also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (AS), or Wireless Fidelity (WiFi) nodes, etc.The term "base station" can be referred to as Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit / Receive Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The term "base station" is not limited to specific technical terms. It should be noted that the embodiments in this application only use base stations in NR systems as examples for description and do not limit the specific type of base station.

[0040] Core network equipment, also known as core network nodes, core network functions, or core network elements, includes, but is not limited to, at least one of the following: Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), and Binding Support Function. Support Functions (BSF), Application Functions (AF), Location Management Functions (LMF), Gateway Mobile Location Centres (GMLC), and Network Data Analytics Functions (NWDAF), etc. It should be noted that this application embodiment only uses core network equipment in the NR system as an example for description, and does not limit the specific type of core network equipment. If the name of the core network equipment mentioned in this application embodiment changes in subsequent protocol versions (e.g., 6G), it will also be within the scope of protection of this application.

[0041] Optionally, the core network equipment can be implemented by one or more functional modules in a single device, or by multiple devices working together; this application does not specifically limit this. It is understood that the aforementioned functional modules can be network elements in hardware devices, software functional modules running on dedicated hardware, or virtualized functional modules instantiated on a platform (e.g., a cloud platform).

[0042] To enable those skilled in the art to better understand the embodiments of this application, the following description will be provided first.

[0043] I. 5G Communication Quality of Service (QoS) Parameters;

[0044] The 3rd Generation Partnership Project (3GPP) standard specifies the use of QoS mechanisms to guarantee the quality of service for UEs at the communication layer. From the perspective of the transport or physical layer, regardless of the data being transmitted, only the corresponding communication-layer metrics such as bandwidth, bit error rate, and latency need to be guaranteed. 5G QoS parameters are used for core network control plane functions (such as SMF and UPF), radio access network functions (such as gNB), and UEs to map the data stream to be transmitted to appropriate transmission channels (such as PDU session, DRB, etc.), and for resource scheduling and other parameter configurations, such as for radio resource control. In 5G, according to relevant protocols, standardized 5G QoS characteristic parameters are shown in Table 1. The mapping between the 5G QoS Identifier (5QI) and the QoS parameter set defined by the protocol standard is shown in Table 1. In actual deployment, operators can customize QoS levels based on the QoS parameter set.

[0045] Table 1

[0046]

[0047] II. QoS parameters for AI services;

[0048] Future 6G systems can provide AI services. Typical QoS parameters for AI services include latency, service accuracy, or service precision. Among these, service accuracy is affected by inference accuracy and communication transmission reliability.

[0049] III. Task Management (TM) Functionality;

[0050] TM can be used to achieve one or more of the following functions:

[0051] 1. Task Verification: Verify whether a task is a legitimate task requested by a User Equipment (UE) or Application Function (AF). Task verification includes at least one of the following: verifying whether the AF allows requesting a task template, whether the task request comes from a legitimate UE (whether the UE has a task subscription), whether the task request comes from a legitimate AF (depending on whether there is a task agreement between the TM and the AF), and whether it exceeds the maximum number of tasks associated with the UE or AF.

[0052] Optionally, if the task request originates from AF, the task verification function can be implemented by NEF.

[0053] 2. Task Template Generation: A task template is determined through negotiation with the AF (Automatic Feedback Unit). This task template is a customized task description provided by the network to the UE. It can be understood as a questionnaire where the requester "checks" mandatory or optional parameters to express their task requirements. The task template includes at least one of the following:

[0054] Task template identifier, used to identify task templates;

[0055] Task description;

[0056] Description of the Quality of Experience (QoE) of the task;

[0057] Task policy or Quality of Service (QoS) description;

[0058] Task parameters.

[0059] 3. Task Translation or Decomposition: Translate or decompose the task request into task instances, where each task instance is an ongoing task. Each task instance includes at least one of the following:

[0060] (1) Task description;

[0061] (2) Task type, used to indicate whether the task is a single task, a combined task, or multiple tasks, wherein the single task is any one of communication, perception, computing, or artificial intelligence (AI), and the combined task is any combination of the above tasks.

[0062] (3) Task strategy or QoS;

[0063] (4) Security-related parameters, used to encrypt transmitted signaling or data;

[0064] (5) Subtask type, which indicates the type of subtask and includes at least one of the following: perception, computation, AI;

[0065] (6) Subtask policy or QoS, used to indicate the quality of service requirements of the subtask.

[0066] If the subtask type is a perception subtask, then the subtask policy or QoS is the perception subtask configuration value (including specific configuration information);

[0067] If the subtask type is a computational subtask, then the subtask policy or QoS is the configuration value of the computational subtask (including specific configuration information);

[0068] If the subtask type is AI subtask, then the subtask policy or QoS is the AI ​​subtask configuration value (including specific configuration information);

[0069] (7) Subtask timing relationships, including at least one of the following:

[0070] a. Subtask execution order, used to indicate the temporal relationship of subtask execution, including at least one of the following implementation methods:

[0071] TM informs the subsequent subtask execution node of the information of the preceding subtask execution node;

[0072] As the central control node for task execution, the preceding subtask execution node informs the TM of the results of the preceding subtask, and the TM informs the subsequent subtask execution node of the results of the preceding subtask.

[0073] For example, if a task includes a perception subtask and a computation subtask, and the TM determines that the perception subtask is a preceding subtask, then the TM informs the computation execution function of the address of the perception execution function. After the computation execution function receives data from the perception execution function, it can start executing the computation subtask.

[0074] b. Subtask bearer establishment order, used to indicate the timing relationship of subtask bearer establishment, including at least one of the following:

[0075] Randomly generated;

[0076] Concurrent creation;

[0077] Establish according to priority.

[0078] The timing relationship of the subtasks can be part of the subtask policy or QoS.

[0079] Due to differences in scenarios and requirements, the task instances can also be classified as follows: The task instances include general parameters and specific parameters. The general parameters describe the service quality requirements of the complete task and include at least one of the following: task description, task type, task policy or QoS, and security-related parameters. The specific parameters describe the service quality requirements of the specifically decomposed subtasks and include at least one of the following: subtask type, subtask policy or QoS, and subtask execution order.

[0080] It is worth noting that TM can also determine task policies or QoS, and subtask policies or QoS, through any of the following methods:

[0081] Obtain from PCF or Policy Billing;

[0082] The task policy or QoS, and subtask policy or QoS are determined based on information obtained from PCF or policy accounting functions.

[0083] 4. Assign a task identifier or task session identifier; optionally, assign a subtask identifier or subtask session identifier.

[0084] 5. Assign task slices or task network names (DNNs), and optionally, assign subtask slices or subtask DNNs.

[0085] 6. Select a subtask management function or a subtask execution function. The TM selects a subtask management function or a subtask execution function that can meet the subtask QoT. Specifically, the TM can query the NRF, UDM, or UDR to determine a suitable subtask management function or subtask execution function, or the TM can determine a suitable subtask management function or subtask execution function based on stored or maintained information (such as whether the task is supported, computing power, etc.).

[0086] The tasks in this embodiment can be understood as specific business tasks, such as a 6G system providing integrated services including communication, information, sensing, and computing on demand based on external business requests. In one optional implementation, the subtask types within the target task can be classified in multiple ways. In another optional implementation, one or more subtasks obtained by decomposing the target task can be understood as subtasks associated with the target task.

[0087] For example, the subtasks may include, but are not limited to, communication subtasks, perception subtasks, computing subtasks, data subtasks, location subtasks, Internet Protocol Multimedia System (IMS) subtasks, AI subtasks, etc.

[0088] For example, the subtasks may also include, but are not limited to, computational subtasks, information processing subtasks, multimedia subtasks, immersive subtasks, and communication subtasks. Specifically, computational subtasks include AI-related computational tasks and non-AI-related computational tasks; information processing subtasks include location information processing and synesthetic information processing; and immersive subtasks include extended reality (XR) tasks and virtual reality (VR) tasks.

[0089] In this embodiment of the application, the service quality requirements of the task include at least one of the following:

[0090] Task feedback time or task delay: This indicates the maximum time from when a task consumer or requester, such as a terminal or a second network-side device, initiates a task request to when it receives a task response.

[0091] Task accuracy: used to indicate the probability of successfully completing a task or to indicate the degree to which the UE's expected value or the task provider's promised value matches the actual value;

[0092] Overall energy consumption: used to indicate the energy consumption required to complete a task, including the energy consumption of the network or terminal, the network including but not limited to the core network and access network;

[0093] Data density: Used to indicate the number of bits transmitted per unit time.

[0094] The term "AI unit / AI model" as used in this application can also be referred to as AI unit, AI model, ML (machine learning) model, ML unit, AI structure, AI function, AI characteristic, machine learning model, neural network, neural network function, neural network functionality, etc. Alternatively, the term "AI unit / AI model" can refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc., related to AI. It can also refer to a processing method, algorithm, function, module, or unit for a specific dataset. Furthermore, it can refer to a processing method, algorithm, function, module, or unit running on AI / ML related hardware such as GPUs, NPUs, TPUs, and ASICs. This invention does not impose specific limitations in these areas. Optionally, the specific dataset includes the input and / or output of the AI ​​unit / AI model.

[0095] Optionally, the identifier of the AI ​​unit / AI model may be an AI model identifier, an AI structure identifier, an AI algorithm identifier, or an identifier of a specific dataset associated with the AI ​​unit / AI model, or an identifier of a specific scenario, environment, channel characteristics, or device related to the AI / ML, or an identifier of a function, characteristic, capability, or module related to the AI / ML. The present invention does not specifically limit this.

[0096] The network elements involved in the embodiments of this application are described below:

[0097] In this embodiment, the function of the first network element includes task management.

[0098] The functions of the second network element include dynamic QoS management and charging policy control. Dynamic QoS management involves adjusting parameters such as bandwidth and latency in real time based on service type (e.g., URLLC / eMBB / mMTC). In 6G, it can adjust corresponding QoS parameters in real time according to different service types in the sensing and computing technologies. Charging policy control involves interacting with the Charging Function (CHF) to formulate differentiated charging rules (e.g., prioritizing traffic packets). The second network element can be referred to as at least one of the following: Policy Charging Function, PCF, Policy Node, etc.

[0099] The third network element includes data-related functional nodes, for example, the third network element includes the fourth network element; or it includes computing power-related functional nodes, for example, the third network element includes the sixth network element; or it includes AI-related functional nodes, for example, the aforementioned third network element includes the fifth network element. The fourth network element is a sensing or data-related functional node, and its functions include: determining the required sensing resources, sensing units, and data sources according to the sensing service category and the internal needs of the sensing service; or, determining the required data sources and data processing units according to the data service category and the internal needs of the data service. The fourth network element can be referred to as at least one of the following: sensing management function, data management function, sensing function, or data source.

[0100] The fifth network element is used for AI model management, model ID allocation, model classification, model acquisition (determining the appropriate model based on internal needs), model storage, and model version management. This fifth network element can be referred to as at least one of the following: AI model management function, AI model control function, AI algorithm control function, algorithm control function, or algorithm control plane.

[0101] The sixth network element is used for computing business resource management. Specifically, it allocates, updates, and releases computing resources as needed; and updates and maintains the status of computing resource nodes. This sixth network element can be referred to as at least one of the following: computing power control function, computing power control plane, and AI resource management function.

[0102] The seventh network element is used for business processing and implementation of business strategy rules. Specifically, it performs AI business calculation processing (user plane execution process) according to the commands of the sixth network element or the first network element; and updates the AI ​​resource status to the sixth network element or the first network element. The seventh network element can be referred to as at least one of the following: AI resource node, computing power execution node, computing power execution function, computing power execution plane.

[0103] The eighth network element is a functional node related to communication functions, including AMF, SMF, or UPF.

[0104] The functions of the ninth network element include determining the required AI resources, AI units, and data sources based on the AI ​​business category and the internal needs of the AI ​​business; and coordinating various elements (such as AI resources, AI units, and data sources). This ninth network element can be referred to as AI business management function, business management function, or collaborative control function.

[0105] In this embodiment of the application, the first network element can be a newly added network element, such as TM. The first network element can also be co-located with existing network elements, for example, co-located with AMF or PCF. The first network element can also be co-located with the ninth network element or the sixth network element.

[0106] It should be noted that when two network elements are co-located, the interaction process between the two network elements is an internal process of the network element.

[0107] The information adjustment method provided in this application will be described in detail below with reference to the accompanying drawings, through some embodiments and application scenarios.

[0108] like Figure 2 As shown in the figure, this application provides an information adjustment method, including:

[0109] Step 201: The first network element obtains the first service requirement of the target service.

[0110] The first service requirement mentioned above refers to the service requirement of the target service sent by the terminal.

[0111] Step 202: The first network element adjusts the first information according to the first service requirement, wherein the first service requirement includes the first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability.

[0112] The aforementioned first service accuracy includes at least one of the following: model accuracy, computational accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability.

[0113] In some embodiments of this application, the model accuracy is obtained based on a first parameter, which includes at least one of the following: the quantization accuracy of the model parameters, the data type of the model parameters, the number of parameters in the model, the structure of the model, and the degree of pruning of the model structure. In the embodiments of this application, model accuracy may refer to the inference accuracy of the model.

[0114] In some embodiments of this application, model precision = F1(a1,a2,a3). Here, a1 represents the ideal inference precision of the selected AI model, i.e., the AI ​​inference precision when the number of digits in the numerical representation of the input data, the number of digits in the numerical representation of the model parameters, and the number of digits in the stored numerical representation of the inference or computation results obtained from the model structure / parameter quantity are all without precision loss due to insufficient numerical representation digits or further pruning; a2 is the degree of pruning; and a3 is the number of digits in the numerical representation of the AI ​​model parameters. F1() is a mathematical function, the specific function selection of which is determined by the operator's strategy and configuration. For example, F1() = a1*a2*a3, F1() = a1*a3, etc.

[0115] In some embodiments of this application, the calculation accuracy is obtained based on a second parameter, which includes at least one of the following: the number of bits used to represent the numerical values ​​of the input data configured during calculation, the number of bits used to represent the numerical values ​​of the model parameters configured during calculation, the number of bits used to represent the numerical values ​​of the intermediate variables configured during calculation, the number of bits used to represent the numerical values ​​of the calculation results configured during calculation, and the calculation algorithm configured during calculation.

[0116] In some embodiments of this application, for AI-related services, the computational precision is defined as F2(a1, a3, b1, b2). Here, a1 represents the ideal inference precision of the selected AI model, i.e., the AI ​​inference precision when the number of digits in the numerical representation of the input data, the numerical representation of the model parameters, and the stored numerical representation of the inference or computation results obtained from the model structure / parameters are all free from precision loss due to insufficient numerical representation digits or further pruning; a3 is the number of digits in the numerical representation of the AI ​​model parameters; b1 is the number of digits in the numerical representation of the input data; and b2 is the number of digits in the stored numerical representation of the inference or computation results obtained from the model structure / parameters. F2() is a mathematical function, the specific function selection of which is determined by the operator's strategy and configuration. For example, F2() = a3 * b1 * b2, or F1() = b1 * b2, etc.

[0117] In some embodiments of this application, for non-AI services, the computational precision is defined as F3(c1, b1, b2). Here, c1 represents the ideal precision of the selected algorithm, i.e., the algorithm precision when there is no loss of precision due to insufficient numerical representation bits in the input data, the stored numerical representation bits of the computation result, etc.; b1 represents the numerical representation bits of the input data; and b2 represents the stored numerical representation bits of the computation result. F3() is a mathematical function, the specific function selection of which is determined by the operator's strategy and configuration. For example, F3() = b1 * b2, etc.

[0118] In some embodiments of this application, data precision is obtained through a third parameter; the third parameter includes at least one of the following: the number of bits used to represent the numerical value during data acquisition, the number of bits used to represent the numerical value of the data to be transmitted, data integrity information, and data timeliness information. In the embodiments of this application, data precision is used to indicate the accuracy and / or level of detail of the data.

[0119] In some embodiments of this application, the AI ​​inference accuracy is obtained through a fourth parameter; the fourth parameter includes at least one of the following: the model accuracy, the computational accuracy;

[0120] In some embodiments of this application, AI inference precision = F4(a1,a2,b1,b2,b3). Here, a1 represents the ideal inference precision of the selected AI model, i.e., the AI ​​inference precision when the number of digits in the numerical representation of the input data, the number of digits in the numerical representation of the model parameters, and the number of digits in the stored numerical representation of the inference or computation results obtained from the model structure / parameter quantity are all without precision loss due to insufficient numerical representation digits or further pruning; a2 is the pruning degree; a3 is the number of digits in the numerical representation of the AI ​​model parameters; b1 is the number of digits in the numerical representation of the input data; b2 is the number of digits in the stored numerical representation of the computation results. F4() is a mathematical function, the specific function selection being determined by the operator's strategy and configuration. For example, F4() = a1*a2*a3*b1*b2, F4() = a1*a2*a3, etc.

[0121] In some embodiments of this application, the AI ​​processing accuracy is obtained through a fifth parameter; the fifth parameter includes at least one of the following: the model accuracy, the calculation accuracy, and the data accuracy.

[0122] In some embodiments of this application, service accuracy = F5(y1,y2,y3), where y1 is communication reliability (e.g., bit error rate), y2 is AI processing accuracy (e.g., AI inference accuracy), and y3 is the completeness / accuracy / timeliness of data collected from the data plane; F5() is a mathematical function, the specific function selection of which is determined by the operator's strategy and configuration. For example, F5() = y1*y2*y3, F5() = y1*y2, etc.

[0123] AI processing accuracy = AI inference precision = F4(a1,a2,b1,b2,b3). Where a1 is the ideal inference precision of the selected AI model, i.e., the AI ​​inference precision when the number of digits in the numerical representation of the input data, the number of digits in the numerical representation of the model parameters, and the number of digits in the stored numerical representation of the inference or computation results obtained from the model structure / parameter quantity are all without precision loss due to insufficient numerical representation digits or further pruning; a2 is the degree of pruning; a3 is the number of digits in the numerical representation of the AI ​​model parameters; b1 is the number of digits in the numerical representation of the input data; b2 is the number of digits in the stored numerical representation of the computation results; F4() is a mathematical function, the specific function selection is determined by the operator's strategy and configuration. For example, F4() = a1*a2*a3*b1*b2, F4() = a1*a2*a3, etc.

[0124] In some embodiments of this application, service accuracy = F5(y1,y2), where y1 is communication reliability (e.g., bit error rate), y2 is AI processing accuracy (e.g., including inference accuracy and the completeness / accuracy / timeliness of data collected from the data plane); F5() is a mathematical function, the specific function selection being determined by the operator's strategy and configuration. For example, F5() = y1*y2, etc.

[0125] AI processing accuracy = F6(a1,a2,b1,b2,b3,y3). Where a1 is the ideal inference accuracy of the selected AI model, i.e., the AI ​​inference accuracy when the numerical representation of the input data, the numerical representation of the model parameters, and the stored numerical representation of the inference or computation results obtained from the model structure / parameter quantity are all free from accuracy loss due to insufficient numerical representation or further pruning; a2 is the pruning degree; a3 is the numerical representation of the AI ​​model parameters; b1 is the numerical representation of the input data; b2 is the stored numerical representation of the computation results; y3 is the completeness / accuracy / timeliness of the data collected from the data plane; F6() is a mathematical function, the specific function selection is determined by the operator's strategy and configuration. For example, F6() = a1*a2*a3*b1*b2*y3, F6() = a1*a2*a3*y3, F6() = a1*a2*a3*b1*y3, etc.

[0126] In some embodiments of this application, communication reliability is obtained through a sixth parameter, which includes the Packet Error Rate (PER), determined based on the number of erroneously received packets and the total number of packets sent.

[0127] The numerical values ​​in the embodiments of this application can also be described as bit width.

[0128] In this embodiment, service accuracy can also be described as service precision rate, service accuracy, or service precision. The accuracy of an AI service is the degree to which the output of an AI model matches the actual or expected result.

[0129] For example: Can an image recognition model correctly identify objects? Can a speech recognition model accurately transcribe speech content?

[0130] In classification problems, "precision" usually refers to accuracy, which is a metric that measures how many samples predicted as positive by the classifier are actually positive. Precision is particularly concerned with reducing false positives (FP), that is, how many samples labeled as positive by the classifier are actually positive.

[0131] In image recognition tasks, the definition of "accuracy" can vary depending on the specific task and context. Here are some common definitions and their explanations:

[0132] 1. Pixel-level precision;

[0133] Suitable for image segmentation tasks (such as semantic segmentation);

[0134] Pixel Accuracy={Number of Correctly Classified Pixels} / {Total Number of Pixels}

[0135] "Correctly Classified Pixels" refers to the number of pixels whose classification results match the true labels, while "Total Number of Pixels" represents the total number of pixels.

[0136] For example, suppose an image has 1000 pixels, and 800 pixels are classified correctly according to the true label. Then, Pixel Accuracy = 800 / 1000 = 80%.

[0137] 2. Top-k Accuracy;

[0138] Suitable for image classification tasks;

[0139] Top-k accuracy is defined as follows: if the true label appears in the top k most probable predicted categories, it is considered correct; otherwise, it is incorrect.

[0140] For example, (k=1): Suppose there are 100 test images, of which 90 have true labels that match the predicted category with the highest probability, then: Top-k Accuracy = 90%.

[0141] In this embodiment, the number of bits used to represent a numerical value refers to the number of binary bits used in a computer to store or represent a numerical value. It directly affects the range, precision, and storage efficiency of the numerical value.

[0142] Numerical representations are generally divided into floating-point numbers and fixed-point numbers.

[0143] For fixed-point numbers, the parameters that affect the precision of fixed-point numbers include: fractional bits, integer bits, and total bit width.

[0144] For floating-point numbers, the parameters affecting their precision include the total number of bits and the number of bits in the mantissa. Other parameters affecting floating-point precision may include the number of decimal significant digits, or dig.

[0145] The mantissa (significant and bits) is the number of binary bits used to store significant digits in the IEEE 754 standard, which determines the absolute precision.

[0146] In the IEEE 754 floating-point standard, dig (Decimal digits of precision) is a key parameter used to describe the minimum number of decimal significant digits that a certain floating-point format can reliably represent.

[0147] In this embodiment, the first network element acquires a first service requirement of the target service; the first network element adjusts first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability. In the above scheme, the first information is adjusted so that the adjusted first information can meet the first service accuracy in the first service requirement, thus meeting the user's service accuracy when the wireless transmission channel changes or user needs change, without occupying a large amount of storage space to store the inference model or algorithm.

[0148] In some embodiments of this application, the model accuracy is adjusted by a first parameter; the first parameter includes at least one of the following: the quantization accuracy of the model parameters, the data type of the model parameters, the number of parameters of the model, the structure of the model, and the degree of pruning of the model structure.

[0149] And / or, the calculation precision is adjusted by a second parameter; the second parameter includes at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the model parameters configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation.

[0150] And / or, the data precision is adjusted by a third parameter; the third parameter includes at least one of the following: the number of bits used to represent the numerical value during data acquisition, the number of bits used to represent the numerical value of the data to be transmitted, data integrity information, and data timeliness information;

[0151] And / or, the AI ​​inference accuracy is adjusted by a fourth parameter; the fourth parameter includes at least one of the following: the data accuracy, the model accuracy, and the computational accuracy;

[0152] And / or, the AI ​​processing accuracy is adjusted by a fifth parameter; the fifth parameter includes at least one of the following: the model accuracy, the calculation accuracy, and the data accuracy.

[0153] And / or, communication reliability is obtained through a sixth parameter, which includes PER.

[0154] In some embodiments, PER = (number of erroneously received packets / total number of packets sent) * 100%;

[0155] The criteria for determining erroneously received packets include at least one of the following:

[0156] Cyclic Redundancy Check (CRC) failed.

[0157] The Media Access Control (MAC) layer has exceeded the maximum number of Hybrid Automatic Repeat reQuest (HARQ) retransmissions.

[0158] Packet loss that cannot be recovered by the Radio Link Control (RLC) layer's Automatic Repeat reQuest (ARQ).

[0159] In some embodiments of this application, the first service requirement further includes a first service latency, and the method further includes:

[0160] The first network element adjusts at least one of the communication delay, computing delay, and data acquisition delay based on the first service delay.

[0161] In this embodiment of the application, the first service delay includes at least one of communication delay, computation delay, and data acquisition delay.

[0162] In this application embodiment, communication delay can also be described as transmission delay, or as end-to-end packet delay, etc.

[0163] In some implementations, the aforementioned communication delay includes at least one of the following:

[0164] The delay in data packet transmission from UE to UPF;

[0165] The delay in data transmission from UPF to UE;

[0166] The time delay for data packets to travel from the UE to the UPF and then back to the UE;

[0167] The delay of a data packet traveling from point X to point Y;

[0168] The delay of a data packet traveling from point X to point Y and then back to point X;

[0169] Among them, point X includes at least one of the following: UE, eighth network element (e.g., UPF), fourth network element, fifth network element, sixth network element, and seventh network element;

[0170] Point Y is different from point X. Point Y includes at least one of the following: UE, eighth network element (e.g., UPF), fourth network element, fifth network element, sixth network element, and seventh network element.

[0171] Here, at least one of the communication delay, computation delay, and data acquisition delay is adjusted to meet the aforementioned first service delay.

[0172] In some embodiments of this application, the first network element adjusts the first information according to the first service requirement, including:

[0173] The first network element adjusts the first information based on the service QoS information or task QoS information corresponding to the first service precision.

[0174] In some embodiments of this application, the aforementioned service QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS;

[0175] The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information;

[0176] The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy;

[0177] The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

[0178] In some embodiments of this application, the task QoS information includes model QoS information, data QoS information, computing power QoS information, and communication QoS information;

[0179] The model QoS information includes at least one of the following: model accuracy, model inference latency, model training latency, and model monitoring latency; the data QoS information includes data accuracy, data acquisition latency, and data processing latency; the computing power QoS information includes computing latency or computing accuracy; and the communication QoS information includes at least one of the following: communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

[0180] The communication QoS information in this application embodiment can also be understood as a Quality of Service Profile or QoS Rules.

[0181] In this embodiment, the first network element converts the first service precision sent by the terminal into service QoS information or task QoS information, and then adjusts the first information based on the service QoS information or task QoS information.

[0182] In some embodiments of this application, the first network element adjusts the first information according to the first service requirement, including:

[0183] The first network element receives a first message, which includes a notification of changes in first Quality of Service (QoS) information; the first network element adjusts the first information according to the first message and the first service requirement.

[0184] The first QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS;

[0185] The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information;

[0186] The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy;

[0187] The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

[0188] In this embodiment, after the terminal sends the first service requirement of the target service, when the network detects a change in the wireless transmission channel, that is, a change in transmission reliability, the network can send the aforementioned first message to the first network element so that the first network element can adjust the first information according to the change notification of the first QoS information in the first message to meet the first service accuracy in the first service requirement.

[0189] In some embodiments of this application, the first network element acquires the first service requirement of the target service, including:

[0190] The first network element receives a second message, which includes: the first service requirement, or task modification related information, wherein the task modification related information is used to determine the first service requirement;

[0191] The first network element determines the first service requirement based on the second message;

[0192] The task modification information includes at least one of the following:

[0193] (1) The identifier of the first strategy, wherein the first strategy includes the strategy corresponding to the second business requirement, and the second business requirement is the business requirement before modification;

[0194] (2) Indication of the reason for updating the strategy;

[0195] In some embodiments of this application, the reason for updating the policy includes updating the QoS policy.

[0196] In some embodiments of this application, the reason indication for the update strategy includes at least one of the following: information on changes in the terminal's computing power; information on changes in the terminal's display accuracy; and information on changes in the terminal's computational requirements.

[0197] For example, the reasons for updating the policy include at least one of the following:

[0198] Reduced UE computing power leads to reduced display accuracy.

[0199] UE computing power is reduced, but display accuracy remains unchanged;

[0200] Improved UE computing power leads to improved display accuracy;

[0201] UE computing power is improved, but display accuracy remains unchanged;

[0202] With the UE computing power and display accuracy remaining unchanged, the computational requirements per unit time increase.

[0203] With the UE computing power and display accuracy remaining unchanged, the computational requirements per unit time decrease.

[0204] In this application embodiment, display precision is understood to mean at least one of the following:

[0205] The number of pixels that can be resolved within a unit of field of view can also be described as pixels per degree (PPD), where PPD = number of pixels per eye / field of view.

[0206] Field of View (FOV); the larger the FOV, the more total pixels are needed to obtain the same PPD.

[0207] Pixel density refers to the number of effective pixels per unit area (Pixels Per Inch, PPI).

[0208] (3) Instructions for modifying the second business requirement;

[0209] In some embodiments of this application, the modification indication of the second service requirement includes at least one of the following: information on changes in the computing power of the terminal; information on changes in the display accuracy of the terminal; and information on changes in the computational requirements of the terminal.

[0210] For example, if the terminal's computing power change information is used to indicate a reduction in the terminal's computing power, the network side will unload the terminal's computing power according to the indication. Conversely, if the terminal's computing power change information is used to indicate an increase in the terminal's computing power, the network side will increase the terminal's computing power according to the indication.

[0211] Based on the modification instructions for the aforementioned second service requirement, the first network element shall determine at least one of the following:

[0212] More computing power needs to be offloaded from the network;

[0213] Less computing power needs to be offloaded by the network;

[0214] The network needs to provide lower service accuracy;

[0215] The network needs to provide higher service accuracy.

[0216] (4) Instructions for the second business requirement;

[0217] (5) First business requirement, which is the modified business requirement.

[0218] For example, the terminal first sends a second service request, and then the user's request changes. The user sends the aforementioned second message to the first network element so that the first network element can determine the changed first service request based on the second message, and then adjust the first information according to the first service precision in the first service request to meet the first service precision in the first service request.

[0219] In some embodiments of this application, the first information is adjusted, including:

[0220] The first network element updates the first information according to the first service requirement to obtain the second information;

[0221] The second information includes at least one of the following:

[0222] (1) The second QoS information that needs to be updated includes at least one of the following: computational latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

[0223] (2) Updated third QoS information, the third QoS information including at least one of the following: computation delay, model accuracy, data accuracy, computation accuracy, model inference delay, model training delay, model monitoring delay, data acquisition delay, data processing delay, communication reliability information, priority level, resource type, communication transmission rate, and communication delay; wherein, the priority level is used to determine the priority of resource scheduling for QoS streams;

[0224] (3) First change amount, the first change amount is used to indicate the change amount between the third QoS information and the QoS information corresponding to the first information;

[0225] (4) First configuration information, wherein the first configuration information is used to indicate parameter configuration information corresponding to the third QoS information, and the parameter configuration information includes configuration information of at least one of the first parameter, the second parameter and the third parameter;

[0226] (5) A first action instruction, wherein the first action instruction is used to increase or decrease the parameter corresponding to the second QoS information, and the parameter corresponding to the second QoS information includes at least one of a first parameter, a second parameter, a third parameter, a fourth parameter and a fifth parameter;

[0227] For example, the first action indicates the degree of pruning corresponding to increasing or decreasing model accuracy. Alternatively, it indicates the number of bits used to represent the numerical value of the computation result in the computation time configuration corresponding to increasing or decreasing computation accuracy.

[0228] For example, if the computation latency is 30ms, the first action instruction can ensure the 30ms computation latency by adjusting the number of bits used to represent the numerical value of the computation result configured during computation from 32 bits to 16 bits.

[0229] (6) Indication of the reason for updating the strategy;

[0230] Optionally, the reason for updating the policy may include updating the QoS policy.

[0231] Optionally, upon receiving the first message, the reason indication for the update policy includes a change in the first QoS information, for example, including at least one of the following: change information in communication QoS, change information in computation quality QoS, and change information in data QoS.

[0232] Optionally, upon receiving the second message, the reason for the update strategy may include at least one of the following: information on changes in the terminal's computing power; information on changes in the terminal's display precision; or information on changes in the terminal's computational requirements.

[0233] (7) The identifier of the first strategy, wherein the first strategy is the strategy corresponding to the first service requirement requested by the terminal before receiving the first message, or the first strategy is the strategy corresponding to the second service requirement requested by the terminal before receiving the second message.

[0234] (8) The identifier of the second policy, which is the policy corresponding to the third QoS information.

[0235] In some embodiments of this application, the first network element updates the first information according to the first service requirement to obtain the second information, including:

[0236] The first network element sends a third message to the second network element based on the first service accuracy, the third message being used to request an update to the first information;

[0237] The first network element obtains the fourth message sent by the second network element, the fourth message being a response message to the third message, and the fourth message including the second information;

[0238] The first network element obtains the second information based on the fourth message;

[0239] The third message includes at least one of the following:

[0240] (1) Identification of the first strategy;

[0241] (2) Indication of the reason for updating the strategy;

[0242] Optionally, upon receiving the first message, the reason indication for the update policy includes a change in the first QoS information, for example, including at least one of the following: change information in communication QoS, change information in computation quality QoS, and change information in data QoS.

[0243] Optionally, upon receiving the second message, the reason for the update strategy may include at least one of the following: information on changes in the terminal's computing power; information on changes in the terminal's display precision; or information on changes in the terminal's computational requirements.

[0244] For example, upon receiving the first message, the reason for the update strategy includes at least one of the following:

[0245] The communication transmission rate (e.g., bit rate) decreases, but the communication reliability remains unchanged;

[0246] Communication transmission rate (e.g., bit rate) decreases, and communication reliability decreases;

[0247] Communication transmission rate (e.g., bit rate) decreases, while communication reliability increases;

[0248] The communication transmission rate (e.g., bit rate) remains unchanged, but the communication reliability decreases;

[0249] The communication transmission rate (e.g., bit rate) remains unchanged, while communication reliability is improved;

[0250] The communication transmission rate (e.g., bit rate) is increased, while the communication reliability remains unchanged;

[0251] Communication transmission rates (e.g., bit rate) increase, but communication reliability decreases;

[0252] Communication transmission rates (e.g., bit rate) are increased, and communication reliability is improved.

[0253] (3) The first QoS information;

[0254] (4) The second QoS information.

[0255] The second network element in this embodiment is a policy node.

[0256] In some embodiments of this application, the above information adjustment method further includes:

[0257] The first network element sends a fifth message to the third network element. The fifth message is used to request modification of the first sub-QoS information. The parameters in the first sub-QoS information are a subset of the parameters in the second QoS information. The first sub-QoS information includes model QoS information, data QoS information, computing power QoS information, and communication QoS information.

[0258] The first network element obtains the sixth message sent by the third network element, wherein the sixth message is a response message to the fifth message;

[0259] The model QoS information includes at least one of the following: model accuracy, model inference latency, model training latency, and model monitoring latency; the data QoS information includes data accuracy, data acquisition latency, and data processing latency; the computing power QoS information includes computing latency or computing accuracy; and the communication QoS information includes at least one of the following: communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

[0260] The priority level in the embodiments of this application is used to determine the priority of resource scheduling for QoS flows.

[0261] In some embodiments of this application, the fifth message includes at least one of the following:

[0262] (1) First sub-QoS information;

[0263] (2) Updated second sub-QoS information, which is determined based on the third QoS information;

[0264] In one implementation, the parameters included in the second sub-QoS information are a subset of the third QoS information.

[0265] (3) A second change amount, which is used to indicate the change amount between the second sub-QoS information and the sub-QoS information corresponding to the first information; the sub-QoS information corresponding to the first information is determined based on the QoS information corresponding to the first information.

[0266] (4) Second configuration information, wherein the second configuration information is used to indicate parameter configuration information corresponding to the second sub-QoS information;

[0267] (5) A second action instruction, the second action instruction being used to increase or decrease the parameter corresponding to the first sub-QoS information, the parameter corresponding to the first sub-QoS information including at least one of a first parameter, a second parameter, a third parameter, a fourth parameter and a fifth parameter;

[0268] (6) Indication of the reason for updating the strategy;

[0269] Optionally, upon receiving the first message, the reason indication for the update policy includes a change in the first QoS information, for example, including at least one of the following: change information in communication QoS, change information in computation quality QoS, and change information in data QoS.

[0270] Optionally, upon receiving the second message, the reason for the update strategy may include at least one of the following: information on changes in the terminal's computing power; information on changes in the terminal's display precision; or information on changes in the terminal's computational requirements.

[0271] (7) The identifier of the first sub-policy, wherein the first sub-policy is the policy corresponding to the first sub-QoS information;

[0272] (8) The identifier of the second sub-policy, which is the policy corresponding to the second sub-QoS information.

[0273] In some embodiments of this application, when the first sub-QoS information includes the model QoS information, the third network element includes an AI-related functional node; for example, the third network element includes the fifth network element.

[0274] Alternatively, if the first sub-QoS information includes the data QoS information, the third network element includes a data-related functional node; for example, the third network element includes the aforementioned fourth network element.

[0275] Alternatively, if the first sub-QoS information includes the computing power QoS information, the third network element includes a functional node related to computing power. For example, the third network element includes the aforementioned sixth network element.

[0276] In some embodiments of this application, the first information is adjusted according to the first business requirement, including at least one of the following:

[0277] When the communication QoS change notification indicates a decrease in transmission rate, and the communication reliability information indicates that the communication reliability remains unchanged, the adjustment of the first information includes: reducing the calculation latency;

[0278] When the communication QoS change notification indicates a decrease in transmission rate and a decrease in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy;

[0279] When the communication QoS change notification indicates a decrease in transmission rate and an increase in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy;

[0280] When the communication QoS change notification indicates that the transmission rate remains unchanged but the communication reliability is reduced, the adjustment of the first information includes at least one of the following: improving model accuracy; improving data accuracy; improving calculation accuracy;

[0281] When the communication QoS change notification indicates that the transmission rate remains unchanged and the communication reliability is improved, the adjustment to the first information includes at least one of the following: reducing model accuracy; reducing data accuracy; reducing calculation accuracy;

[0282] When the communication QoS change notification indicates an increase in transmission rate and the communication reliability remains unchanged, the adjustment to the first information includes: increasing the calculation delay;

[0283] When the communication QoS change notification indicates an increase in transmission rate and an increase in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy;

[0284] When the communication QoS change notification indicates an increase in transmission rate and a decrease in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy.

[0285] In some embodiments of this application, the first information is adjusted according to the first business requirement, including at least one of the following:

[0286] When the second message indicates that the terminal's computing power has decreased and the display accuracy has decreased, the adjustment to the first message includes at least one of the following: increasing the computing speed, decreasing the computing accuracy, decreasing the algorithm accuracy, decreasing the data accuracy, and increasing the communication speed;

[0287] When the second message indicates that the terminal's computing power is reduced but the display accuracy remains unchanged, the adjustment to the first information includes at least one of the following: increasing the computing speed and increasing the communication speed;

[0288] When the second message indicates that the terminal's computing power and display accuracy have increased, the adjustment to the first information includes at least one of the following: reducing the computing speed, increasing the computing accuracy, increasing the algorithm accuracy, increasing the data accuracy, and reducing the communication speed;

[0289] When the computing power of the second message terminal is increased while the display accuracy remains unchanged, the adjustment of the first information includes at least one of the following: reducing the computing rate and reducing the communication rate;

[0290] When the second message indicates that the terminal's computational demand per unit time increases, the adjustment to the first message includes at least one of the following: increasing the computational rate and increasing the communication rate;

[0291] When the second message indicates that the terminal's computational requirements per unit time are reduced, the adjustment to the first message includes at least one of the following: reducing the computational rate and reducing the communication rate.

[0292] In some embodiments of this application,

[0293] When the AI ​​unit is provided by a network-side device, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the network-side device, the model structure of the AI ​​unit from the network-side device, and the degree of pruning of the model structure of the AI ​​unit from the network-side device.

[0294] When the AI ​​unit is provided by the terminal, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the terminal, the model structure of the AI ​​unit from the terminal, and the degree of pruning of the model structure of the AI ​​unit from the terminal.

[0295] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the terminal is represented by the number of bits, and the numerical value of the first data is represented by the number of bits;

[0296] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the network-side device, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the network-side device is represented by the number of bits, and the numerical value of the first data is represented by the number of bits;

[0297] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal and the network side device, the data accuracy is adjusted by at least one of the following: the value of the data collected by the network side device is represented by the number of bits, the value of the data collected by the terminal is represented by the number of bits, and the value of the first data is represented by the number of bits.

[0298] When the service precision is set to AI-specific service precision, the calculation precision is adjusted by at least one of the following: the number of bits used to represent the numerical values ​​of the input data configured during calculation, the number of bits used to represent the numerical values ​​of the model parameters configured during calculation, the number of bits used to represent the numerical values ​​of the intermediate variables configured during calculation, and the number of bits used to represent the numerical values ​​of the calculation results configured during calculation.

[0299] When the service precision is for non-AI services, the calculation precision is adjusted by at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation.

[0300] The first data includes data to be transmitted that the network sends to the terminal.

[0301] In this embodiment, the first network element acquires a first service requirement of the target service; the first network element adjusts first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability. In the above scheme, the first information is adjusted so that the adjusted first information can meet the first service accuracy in the first service requirement, thus meeting the user's service accuracy when the wireless transmission channel changes or user needs change, without occupying a large amount of storage space to store the inference model or algorithm.

[0302] The information adjustment method of this application will be described below with reference to embodiments.

[0303] Example 1: Example 1 is an overall design for computing accuracy, which can adjust the corresponding computing resources when communication reliability changes or the user's demand for service accuracy changes, so as to meet the user's demand for service accuracy as much as possible.

[0304] 1. Definition of calculation precision;

[0305] like Figure 3 As shown, computational accuracy can be defined as the degree of similarity between the first result and the second result. The first result can also be described as the result of executing the first algorithm, and the second result can also be described as the result of executing the second algorithm.

[0306] The first result is the inference or calculation result obtained by representing the input data at point B (after wireless transmission, before computation) in terms of the number of bits, the model parameters in terms of the number of bits, and the model structure / parameter quantity. The second result is the inference or calculation result obtained by representing the input data at point C (during computation) in terms of the number of bits, the model parameters in terms of the number of bits, and the model structure and parameter quantity.

[0307] For example, for AI-related services, let's assume:

[0308] The first precision is the AI ​​inference precision obtained by inferring the numerical values ​​of the input data at point B (after wireless transmission and before calculation), the numerical values ​​of intermediate variables / inference results (assuming the numerical values ​​of intermediate variables / inference results are B4, B4 is greater than or equal to B1*B1 to ensure no computational loss), the numerical values ​​of model parameters, and the model structure / parameter quantity through AI unit inference; where B1 is the bit width of the input data at point B.

[0309] The second accuracy is the AI inference accuracy obtained by performing inference by the AI unit based on the number of bits for numerical representation of the input data at point C (during calculation), the number of bits for numerical representation of the intermediate variables / inference results, the number of bits for numerical representation of the model parameters, and the model structure / parameter quantity;

[0310] Then the calculation accuracy = second accuracy / first accuracy.

[0311] For non-AI services, assume that:

[0312] The first accuracy is the algorithm accuracy obtained by performing algorithm calculation based on the number of bits for numerical representation of the input data at point B (after wireless transmission and before calculation), and the number of bits for stored numerical representation of the calculation result (assuming no accuracy loss due to calculation), and the selected algorithm;

[0313] The second accuracy is the algorithm accuracy obtained by performing algorithm calculation based on the number of bits for numerical representation of the input data at point C (during calculation), the number of bits for stored numerical representation of the calculation result, and the selected algorithm.

[0314] Then the calculation accuracy = second accuracy / first accuracy.

[0315] In Figure 2 , point A is the data or model provided by the UE, point B is the data or model after wireless transmission and before calculation, point C is the data and model when the AI unit performs inference at the AI resource node or the algorithm is executed at the calculation node, point D is the result data after calculation and before result transmission, and point F is the result data received by the UE.

[0316] The accuracy loss of the input data bit width from A1 to B1 is the accuracy loss caused by transmission, and this accuracy loss is uncontrollable;

[0317] The accuracy loss from B to C is a controllable accuracy loss, which can keep the accuracy unchanged. When the accuracy loss at point B is the same as the accuracy loss at point C, the calculation accuracy is determined by model accuracy + data accuracy.

[0318] When the service accuracy at point C < the service accuracy at point B,

[0319] AI service accuracy = model accuracy * data accuracy * F (pruning / quantization) * transmission reliability;

[0320] Non-AI service accuracy = algorithm accuracy * data accuracy * F (quantization) * transmission reliability;

[0321] B1→C1: To speed up the calculation, reduce the number of bits for numerical representation of the data, or keep it unchanged;

[0322] B2→C2: To speed up the calculation, reduce the number of bits for numerical representation of the model parameters, or keep it unchanged;

[0323] B3→C3: To speed up the calculation, the model can be pruned or left unchanged;

[0324] (1) For point A:

[0325] Point A may include data provided by the UE, and optionally, it may also include a model provided by the UE. Therefore, factors related to service accuracy at point A include:

[0326] A1: The numerical representation of the input data provided by the UE is in bits;

[0327] A2: The UE provides the number of bits used to represent the numerical values ​​of the model's parameters;

[0328] A3: UE provides the neural network structure of the model;

[0329] A4: UE provides the number of parameters or nodes in the model's neural network;

[0330] At point A, the service accuracy can be adjusted by at least one of the above methods.

[0331] (2) For the distance from point A to point B;

[0332] In this process, the loss of service accuracy from point A to point B is due to accuracy loss caused by wireless transmission. This includes at least one of the following:

[0333] The data provided by the UE is quantized before being sent to the network;

[0334] When the wireless channel deteriorates, the UE reduces the number of bits used to represent the numerical values ​​of the input data, thereby reducing the amount of data sent to the network.

[0335] When the wireless channel deteriorates, the UE reduces the number of bits in the numerical representation of the model parameters, thereby reducing the amount of data sent to the network;

[0336] When the wireless channel deteriorates, the UE prunes the model to reduce the amount of data sent to the network;

[0337] When the wireless channel deteriorates, the UE reduces the number of model parameters (e.g., by dropping out certain parameters) to reduce the amount of data sent to the network.

[0338] When the wireless channel deteriorates, point B does not receive complete data, causing some model parameters to be dropped out;

[0339] When the wireless channel deteriorates, point B does not receive all the correct data, causing some model parameters to be modified.

[0340] (3) For the distance from point B to point C;

[0341] The loss of service accuracy from point B to point C is controllable. That is, it is possible to choose a point B and point C where no accuracy loss occurs, or to allow a loss of service accuracy between points B and C under at least one of the following conditions.

[0342] 1. Limited memory resources;

[0343] 2. Computation latency is limited;

[0344] 3. Communication reliability is higher than the first threshold;

[0345] 4. Communication latency exceeds the second threshold.

[0346] (4) For the distance from point B to point C, the factors affecting the calculation accuracy include at least one of the following:

[0347] The input data is represented by a number of digits;

[0348] The numerical representation of model parameters has digits.

[0349] The degree of pruning for AI units / models;

[0350] The number of parameters in the model;

[0351] The numerical value of the intermediate variable is represented by a number of digits;

[0352] The numerical representation of the inference / calculation results in digits.

[0353] (5) The loss of service accuracy caused by the transition from point B to point C includes at least one of the following:

[0354] Reduce the number of bits in the input data numerical representation, C1 <B1;

[0355] Reduce the number of bits in the numerical representation of model parameters, C2 <B2;

[0356] Prune the AI ​​units / models;

[0357] Dropout is applied to the model's parameters;

[0358] Reduce the number of digits in the numerical representation of intermediate variables, C4 <C2*C1;

[0359] Reduce the number of bits used to represent the numerical values ​​of inference or calculation results, C5 <C2*C1;

[0360] (6) For the distance from point C to point D;

[0361] The process from point C to point D is that the network maps the stored settlement result to feedback data. At this time, if the wireless transmission channel is good, the number of bits for numerical representation of the calculation result may be equal to the number of bits for numerical representation of the feedback data, that is, C5=D1; if the wireless transmission channel deteriorates, the number of bits for numerical representation of the calculation result may be greater than the number of bits for numerical representation of the feedback data, that is, C5>D1, so as to reduce the amount of transmitted data, and ensure service latency by sacrificing calculation accuracy.

[0362] (7) For the part from point D to point E;

[0363] In this process, the service accuracy loss caused from point D to point E belongs to the accuracy loss caused by wireless transmission. It includes at least one of the following:

[0364] The data provided by the network is quantized before being sent to the UE;

[0365] When the wireless channel deteriorates, the network reduces the number of bits for numerical representation of the feedback data, so as to reduce the data sent to the UE;

[0366] When the wireless channel deteriorates, the UE does not receive complete data, resulting in the loss of some feedback data;

[0367] When the wireless channel deteriorates, the UE does not receive all data correctly, resulting in inaccuracy of some feedback data.

[0368] Embodiment 2: Service accuracy of AI-type services;

[0369] Embodiment 2-1: As shown in Figure 4 , the service accuracy influencing factors when the inference AI unit is provided by the network and the model input data comes from the UE include data accuracy, model accuracy, calculation accuracy, and transmission reliability;

[0370] (1) The data accuracy includes at least one of the following:

[0371] The number of bits for numerical representation / data type of the data provided by the UE (as model input);

[0372] The number of bits for numerical representation / data type of the feedback data sent to the UE in the downlink;

[0373] The data accuracy may be controlled by at least one of a data management function, a data execution function, and a cooperative management function.

[0374] (2) The model accuracy includes at least one of the following:

[0375] The number of bits for numerical representation of the model parameters from the network;

[0376] The structure / number of nodes of the model from the network;

[0377] The pruning configuration used for calculation can be understood as the pruning degree described above.

[0378] Model accuracy can be controlled by at least one of the following: AI model management function, AI control function, AI model library and collaborative management function.

[0379] (3) The calculation accuracy includes at least one of the following:

[0380] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0381] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0382] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0383] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0384] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0385] (4) Transmission reliability includes at least one of the following:

[0386] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0387] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the result data.

[0388] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0389] It should be noted that in this embodiment 2-1, point A can be the UE's transmitting end, point B can be the aforementioned second network element and fifth network element, the second network element is used to determine communication reliability, and the fifth network element is used to obtain AI units from the network; point C can be the sixth network element or the fifth network element, point D can be the second network element or the fourth network element, and point E can refer to the UE's receiving end.

[0390] Example 2-2: As Figure 5 As shown, the inference AI unit is provided by the network, and the factors affecting the service accuracy when the model input data comes from the network include data accuracy, model accuracy, computational accuracy, and transmission reliability.

[0391] (1) Data precision includes at least one of the following:

[0392] The numerical representation of data collected from the network (as input to the model, or as input to the computational algorithm) in terms of bit depth / data type;

[0393] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0394] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions;

[0395] (2) Model accuracy includes at least one of the following:

[0396] The number of bits used to represent the numerical values ​​of the model parameters from the network;

[0397] The structure / number of nodes of the model from the network;

[0398] Pruning configuration used for calculation;

[0399] Model accuracy can be controlled by at least one of the following: AI model management function, AI control function, AI model library and collaborative management function.

[0400] (3) The calculation accuracy includes at least one of the following:

[0401] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0402] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0403] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0404] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0405] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0406] (4) Transmission reliability includes at least one of the following:

[0407] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0408] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0409] It should be noted that in this embodiment 2-2, point A can be the UE's transmitting end, point B can be the aforementioned fourth and fifth network elements, with the fifth network element used to obtain AI units from the network; point C can be the sixth or fifth network element, point D can be the eighth or fourth network element, and point E can refer to the UE's receiving end.

[0410] Examples 2-3: Figure 6 As shown, when the inference AI unit is provided by the network and the model input data comes from the UE and the network, the factors affecting the service accuracy include data accuracy, model accuracy, calculation accuracy, and transmission reliability.

[0411] (1) Data precision includes at least one of the following:

[0412] The numerical representation of data collected from the network (as input to the model, or as input to the computational algorithm) in terms of bit depth / data type;

[0413] The number of bits / data type in the numerical representation of the data collected by the UE (as input to the model, or as input to the calculation algorithm);

[0414] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0415] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0416] (2) Model accuracy includes at least one of the following:

[0417] The number of bits used to represent the numerical values ​​of the model parameters from the network;

[0418] The structure / number of nodes of the model from the network;

[0419] Pruning configuration used for calculation;

[0420] Model accuracy can be controlled by at least one of the following: AI model management function, AI control function, AI model library and collaborative management function.

[0421] (3) The calculation accuracy includes at least one of the following:

[0422] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0423] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0424] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0425] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0426] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0427] (4) Transmission reliability includes at least one of the following:

[0428] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0429] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0430] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0431] It should be noted that in this embodiment 2-3, point A can be the UE's transmitting end, point B can be the second, fourth and fifth network elements mentioned above, the second network element is used to determine communication reliability, the fourth network element is used to acquire input data, and the fifth network element is used to acquire AI units from the network; point C can be the sixth or fifth network element, point D can be the second or fourth network element, and point E can refer to the UE's receiving end.

[0432] Examples 2-4: Figure 7 As shown, when the inference AI unit is sent to the network by the UE and the model input data comes from the UE, the factors affecting the service accuracy include data accuracy, model accuracy, calculation accuracy, and transmission reliability.

[0433] (1) Data precision includes at least one of the following:

[0434] The number of bits / data type in the numerical representation of the data collected by the UE (as input to the model, or as input to the calculation algorithm);

[0435] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0436] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0437] (2) Model accuracy includes at least one of the following:

[0438] The number of bits used to represent the numerical values ​​of the model parameters from the UE;

[0439] The structure / number of nodes of the model from the UE;

[0440] Pruning configuration used for calculation;

[0441] Model accuracy can be controlled by at least one of the following: UE, AI model management function, AI control function, AI model library, and collaborative management function.

[0442] (3) The calculation accuracy includes at least one of the following:

[0443] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0444] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0445] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0446] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0447] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0448] (4) Transmission reliability includes at least one of the following:

[0449] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0450] The transmission reliability of the model parameters of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the model parameters.

[0451] The transmission reliability of the model structure of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the bit error rate and packet loss rate of the model structure.

[0452] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the result data.

[0453] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0454] It should be noted that in this embodiment 2-4, point A can be the UE's transmitting end, point B can be the aforementioned second network element, which is used to determine communication reliability; point C can be the sixth or fifth network element, point D can be the second or fourth network element, and point E can refer to the UE's receiving end.

[0455] Examples 2-5: Figure 8 As shown, when the inference AI unit is sent to the network by the UE, the factors affecting the service accuracy when the model input data comes from the network include data accuracy, model accuracy, calculation accuracy, and transmission reliability.

[0456] (1) Data precision includes at least one of the following:

[0457] The numerical representation of data collected from the network (as input to the model, or as input to the computational algorithm) in terms of bit depth / data type;

[0458] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0459] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0460] (2) Model accuracy includes at least one of the following:

[0461] The number of bits used to represent the numerical values ​​of the model parameters from the UE;

[0462] The structure / number of nodes of the model from the UE;

[0463] Pruning configuration used for calculation;

[0464] Model accuracy can be controlled by at least one of the following: UE, AI model management function, AI control function, AI model library, and collaborative management function.

[0465] (3) The calculation accuracy includes at least one of the following:

[0466] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0467] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0468] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0469] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0470] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0471] (4) Transmission reliability includes at least one of the following:

[0472] The transmission reliability of the model parameters of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the model parameters.

[0473] The transmission reliability of the model structure of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the bit error rate and packet loss rate of the model structure.

[0474] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the result data.

[0475] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0476] It should be noted that in this embodiment 2-5, point A can be the UE's transmitting end, point B can be the second network element and the fourth network element mentioned above, the second network element is used to determine the communication reliability, and the fourth network element is used to obtain input data; point C can be the sixth network element or the fifth network element, point D can be the second network element or the fourth network element, and point E can refer to the UE's receiving end.

[0477] Examples 2-6: Figure 9 As shown, when the inference AI unit is sent to the network by the UE, and the model input data comes from the UE and the network, the factors affecting the service accuracy include data accuracy, model accuracy, calculation accuracy, and transmission reliability.

[0478] (1) Data precision includes at least one of the following:

[0479] The number of bits / data type in the numerical representation of the data collected by the UE (as input to the model, or as input to the calculation algorithm);

[0480] The numerical representation of data collected from the network (as input to the model, or as input to the computational algorithm) in terms of bit depth / data type;

[0481] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0482] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0483] (2) Model accuracy includes at least one of the following:

[0484] The number of bits used to represent the numerical values ​​of the model parameters from the UE;

[0485] The structure / number of nodes of the model from the UE;

[0486] Pruning configuration used for calculation;

[0487] Model accuracy can be controlled by at least one of the following: UE, AI model management function, AI control function, AI model library, and collaborative management function.

[0488] (3) The calculation accuracy includes at least one of the following:

[0489] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0490] The numerical representation of the model parameters configured during calculation is in bits / data type;

[0491] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0492] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0493] The accuracy of the calculation can be controlled by at least one of the following: AI resource management function, calculation control function, calculation execution node and collaborative management function.

[0494] (4) Transmission reliability includes at least one of the following:

[0495] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0496] The transmission reliability of the model parameters of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the following: the number of bits, the data type, the bit error rate, and the packet loss rate of the model parameters.

[0497] The transmission reliability of the model structure of the AI ​​unit transmitted by the UE uplink can be represented by at least one of the bit error rate and packet loss rate of the model structure.

[0498] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0499] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0500] It should be noted that in this embodiment 2-6, point A can be the UE's transmitting end, point B can be the second network element and the fourth network element mentioned above, the second network element is used to determine the communication reliability, and the fourth network element is used to obtain input data; point C can be the sixth network element or the fifth network element, point D can be the second network element or the fourth network element, and point E can refer to the UE's receiving end.

[0501] Example 3: Service accuracy of non-AI services;

[0502] Example 3-1: As Figure 10 As shown, when the calculation execution algorithm is provided by the network and the model input data comes from the UE, the factors affecting the service accuracy include data accuracy, algorithm accuracy, calculation accuracy, and transmission reliability.

[0503] (1) Data precision includes at least one of the following:

[0504] The numerical representation of the data provided by the UE (as algorithm input) is in bits / data type;

[0505] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0506] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0507] (2) The algorithm accuracy includes at least one of the following:

[0508] Algorithms from the network;

[0509] The algorithm used for computation is selective;

[0510] Algorithm accuracy can be controlled by at least one of the following: algorithm control function, algorithm library, and collaborative management function.

[0511] (3) The calculation accuracy includes at least one of the following:

[0512] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0513] The algorithm configured during computation;

[0514] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0515] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0516] The accuracy of the calculation can be controlled by at least one of the following: calculation control function, calculation execution node, and collaborative management function.

[0517] (4) Transmission reliability includes at least one of the following:

[0518] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0519] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0520] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0521] It should be noted that in this embodiment 2-1, point A can be the UE's transmitting end, point B can be the second network element and the fifth network element mentioned above, the second network element is used to determine the communication reliability, and the fifth network element is used to obtain the algorithm from the network; point C can be the sixth network element or the fifth network element, point D can be the second network element or the fourth network element, and point E can refer to the UE's receiving end.

[0522] Example 3-2: As Figure 11 As shown, when the computational execution algorithm is provided by the network and the model input data comes from the UE and the network, the factors affecting the service accuracy include data accuracy, algorithm accuracy, computational accuracy, and transmission reliability.

[0523] (1) Data precision includes at least one of the following:

[0524] The numerical representation of the data provided by the UE (as algorithm input) is in bits / data type;

[0525] The numerical representation of the data provided by the network (as input to the algorithm) is in bits / data type;

[0526] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0527] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0528] (2) The algorithm accuracy includes at least one of the following:

[0529] Algorithms from the network;

[0530] The algorithm used for computation is selective;

[0531] Algorithm accuracy can be controlled by at least one of the following: algorithm control function, algorithm library, and collaborative management function.

[0532] (3) The calculation accuracy includes at least one of the following:

[0533] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0534] The algorithm configured during computation;

[0535] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0536] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0537] The accuracy of the calculation can be controlled by at least one of the following: calculation control function, calculation execution node, and collaborative management function.

[0538] (4) Transmission reliability includes at least one of the following:

[0539] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0540] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0541] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0542] It should be noted that in this embodiment 3-2, point A can be the UE's transmitting end, point B can be the eighth network element, the fourth network element, and the fifth network element mentioned above, wherein the eighth network element is used to determine communication reliability, the fifth network element is used to obtain the execution algorithm from the network, and the fourth network element is used to obtain input data; point C can be the sixth network element or the fifth network element, point D can be the eighth network element or the fourth network element, and point E can refer to the UE's receiving end.

[0543] Example 3-3: As Figure 12 As shown, the calculation execution algorithm is sent to the network by the UE. When the model input data comes from the UE, the factors affecting the service accuracy include data accuracy, algorithm accuracy, calculation accuracy, and transmission reliability.

[0544] (1) Data precision includes at least one of the following:

[0545] The numerical representation of the data provided by the UE (as algorithm input) is in bits / data type;

[0546] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0547] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0548] (2) The algorithm accuracy includes at least one of the following:

[0549] At least one algorithm from the UE;

[0550] The algorithm used for computation is selective;

[0551] Algorithm accuracy can be controlled by at least one of the following: algorithm control function, algorithm library, and collaborative management function.

[0552] (3) The calculation accuracy includes at least one of the following:

[0553] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0554] The algorithm configured during computation;

[0555] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0556] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0557] The accuracy of the calculation can be controlled by at least one of the following: calculation control function, calculation execution node, and collaborative management function.

[0558] (4) Transmission reliability includes at least one of the following:

[0559] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0560] The UE provides the transmission reliability of the algorithm, which can be represented by at least one of bit error rate and packet loss rate; the UE receives the calculation result data downlink, which can be represented by at least one of the numerical representation bit number, data type, bit error rate, and packet loss rate of the result data.

[0561] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0562] It should be noted that in this embodiment 3-3, point A can be the UE's transmitting end, point B can be the aforementioned eighth network element, wherein the eighth network element is used to determine communication reliability; point C can be the sixth or fifth network element, point D can be the eighth or fourth network element, and point E can refer to the UE's receiving end.

[0563] Examples 3-4: Figure 13 As shown, the calculation execution algorithm is sent to the network by the UE. When the model input data comes from the UE and the network, the factors affecting the service accuracy include data accuracy, algorithm accuracy, calculation accuracy, and transmission reliability.

[0564] (1) Data precision includes at least one of the following:

[0565] The numerical representation of the data provided by the UE (as algorithm input) is in bits / data type;

[0566] The numerical representation of the data provided by the network (as input to the algorithm) is in bits / data type;

[0567] The numerical representation of the feedback data sent to the UE in the downlink is represented by a number of bits / data type;

[0568] Data accuracy can be controlled by at least one of the following: data management functions, data execution functions, and collaborative management functions.

[0569] (2) The algorithm accuracy includes at least one of the following:

[0570] At least one algorithm from the UE;

[0571] The algorithm used for computation is selective;

[0572] Algorithm accuracy can be controlled by at least one of the following: algorithm control function, algorithm library, and collaborative management function.

[0573] (3) The calculation accuracy includes at least one of the following:

[0574] The numerical representation of the input data configured during calculation is in terms of the number of bits / data type;

[0575] The algorithm configured during computation;

[0576] The numerical representation of intermediate variables configured during calculation is indicated by the number of bits / data type;

[0577] The number of bits / data type of the numerical representation of the result data configured during calculation;

[0578] The accuracy of the calculation can be controlled by at least one of the following: calculation control function, calculation execution node, and collaborative management function.

[0579] (4) Transmission reliability includes at least one of the following:

[0580] The transmission reliability of the UE uplink transmission of input data can be represented by at least one of the following: the number of bits of numerical representation of the input data, the data type, the bit error rate, and the packet loss rate.

[0581] The UE provides the transmission reliability of the algorithm, which can be represented by at least one of bit error rate and packet loss rate;

[0582] The transmission reliability of the downlink received calculation result data of the UE can be represented by at least one of the following: the number of bits of numerical representation of the result data, the data type, the bit error rate, and the packet loss rate.

[0583] Transmission reliability can be controlled by at least one of the following: session management function, communication management function, AMF, SMF, UPF, and cooperative management function.

[0584] It should be noted that in this embodiment 3-4, point A can be the UE's transmitting end, point B can be the aforementioned eighth network element and fourth network element, wherein the eighth network element is used to determine communication reliability, and the fourth network element is used to acquire input data; point C can be the sixth network element or the fifth network element, point D can be the eighth network element or the fourth network element, and point E can refer to the UE's receiving end.

[0585] Example 4: A network-initiated dynamic adjustment scheme for computational precision;

[0586] When the network detects changes in wireless transmission, i.e., changes in transmission reliability, the overall service accuracy can be kept constant by adjusting the calculation precision.

[0587] like Figure 14 As shown, the scheme includes:

[0588] Step 1: The first network element receives the first message. The first message includes a notification of changes in the first Quality of Service (QoS) information.

[0589] The first message can be sent by the eighth network element.

[0590] The eighth network element mentioned above is a communication function node. For example, the eighth network element is a communication user plane function such as RAN or UPF.

[0591] Among them, the aforementioned first QoS information is QoS information related to the reason for the update strategy, and is QoS information that the first network element has detected as having changed.

[0592] The first QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS;

[0593] The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information;

[0594] The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy;

[0595] The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

[0596] The aforementioned communication transmission rate can also be referred to as communication transmission bandwidth or bit rate.

[0597] The communication transmission rate includes at least one of the following: uplink communication transmission rate, downlink communication transmission rate.

[0598] The communication latency and reliability have been described above and will not be repeated here.

[0599] Step 2: The first network element is determined and the strategy is modified.

[0600] It should be noted that when the first network element and the second network element are co-located, steps 2-1, 2-2, and 2-3 are the internal processes of the first network element or the second network element.

[0601] Step 2-1: The first network element decides to adjust or modify the policy and sends a third message to the second network element. The third message is used to request a policy update.

[0602] The third message includes at least one of the following: the ID of the first policy, an indication of the reason for updating the policy, the first QoS information, and the second QoS information that needs to be updated;

[0603] The reason for updating the policy is indicated as: changes in communication QoS;

[0604] In this embodiment, the reason for updating the strategy includes at least one of the following:

[0605] In this embodiment, the reason for updating the strategy includes at least one of the following:

[0606] The communication transmission rate (e.g., bit rate) decreases, but the communication reliability remains unchanged;

[0607] Communication transmission rate (e.g., bit rate) decreases, and communication reliability decreases;

[0608] Communication transmission rate (e.g., bit rate) decreases, while communication reliability increases;

[0609] The communication transmission rate (e.g., bit rate) remains unchanged, but the communication reliability decreases;

[0610] The communication transmission rate (e.g., bit rate) remains unchanged, while communication reliability is improved;

[0611] The communication transmission rate (e.g., bit rate) is increased, while the communication reliability remains unchanged;

[0612] Communication transmission rates (e.g., bit rate) increase, but communication reliability decreases;

[0613] Communication transmission rates (e.g., bit rate) are increased, and communication reliability is improved.

[0614] Among them, the aforementioned second QoS information is the QoS information to be updated, which is the QoS information that the second network element can adjust.

[0615] In this embodiment, the second QoS information includes at least one of the following:

[0616] Computational latency, model accuracy, data accuracy, computational accuracy;

[0617] Step 2-2: Adjust or modify the strategy of the second network element.

[0618] Specifically, determine the type of wireless link change and identify an adjustment strategy for at least one of the following: model QoS information, data QoS information, computing power QoS information, and communication QoS information.

[0619] Optional:

[0620] Case 1a: The wireless transmission bandwidth is reduced (i.e., the transmission rate is reduced), but the transmission reliability remains unchanged.

[0621] Service latency: Due to the reduction in wireless transmission rate, the transmission latency increases. In order to keep the total service latency constant, it is necessary to reduce the computation latency.

[0622] Service accuracy: Since the transmission reliability remains unchanged, the service accuracy is not affected. Therefore, there is no need to adjust the calculation accuracy, model accuracy, or data accuracy.

[0623] Modify the strategy: Reduce computation latency, for example, by providing more computing resources to ensure that the total service latency remains unchanged.

[0624] Case 1a-2: The wireless transmission bandwidth is reduced (i.e., the transmission rate is reduced), and the transmission reliability is reduced.

[0625] Service latency: Due to the reduction in wireless transmission rate, the transmission latency increases. In order to keep the total service latency constant, it is necessary to reduce the computation latency.

[0626] Service accuracy: Reduced transmission reliability impacts service accuracy; therefore, it's necessary to improve computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0627] Method 1: Improve model accuracy without affecting the overall latency. For example, use an AI model with higher inference accuracy (e.g., reduce pruning or use a more complex model) to ensure that the overall service accuracy meets the requirements;

[0628] Method 2: Improve data accuracy without affecting the overall latency. For example, collect higher-precision input data on the network side as input to the AI ​​model to ensure that the overall service accuracy meets the requirements; premise: the data collected on the network side can also be used as model input.

[0629] Method 3: Improve computational accuracy without affecting the overall latency. For example, increase the number of bits used to represent the input data (point C, closer to point B), the number of bits used to represent the model parameters (point C, closer to point B), the number of bits used to represent the intermediate data (point C, closer to the square of point B), and the number of bits used to represent the result data (point C, closer to the square of point B) to ensure that the overall service accuracy meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0630] Modification Strategy 1: Reduce computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged) and improve model accuracy.

[0631] Modification Strategy 2: Reduce computation latency (e.g., by providing more computing resources to ensure that the total service latency remains unchanged) and improve data accuracy.

[0632] Modification Strategy 3: Reduce computation latency (e.g., by providing more computing resources to ensure that the total service latency remains unchanged) and improve computation accuracy.

[0633] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0634] Case 1a-3: Wireless transmission bandwidth is reduced (i.e., transmission rate is reduced), but transmission reliability is improved.

[0635] Service latency: Due to the reduction in wireless transmission rate, the transmission latency increases. In order to keep the total service latency constant, it is necessary to reduce the computation latency.

[0636] Service accuracy: Improved transmission reliability has impacted service accuracy. Therefore, it is necessary to reduce computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0637] Method 1: Reduce model accuracy without affecting the overall latency. For example, use an AI model with lower inference accuracy (e.g., increase pruning or use a simpler model) to ensure that the overall service accuracy meets the requirements;

[0638] Method 2: Reduce data precision without affecting the overall latency. For example, collect lower precision input data on the network side as input to the AI ​​model to ensure that the overall service precision meets the requirements; premise: the data collected on the network side can also be used as model input.

[0639] Method 3: Reduce computational precision without affecting the overall latency. For example, reduce the number of bits used to represent the input data (point C, less than point B), the number of bits used to represent the model parameters (point C, less than point B), the number of bits used to represent the intermediate data (point C, less than the square of point B), and the number of bits used to represent the result data (point C, less than the square of point B) to ensure that the overall service precision meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0640] Modification Strategy 1: Reduce computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged), and reduce model accuracy.

[0641] Modification Strategy 2: Reduce computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged), and reduce data precision.

[0642] Modification Strategy 3: Reduce computation latency (e.g., by providing more computing resources to ensure that the total service latency remains unchanged), and reduce computation precision.

[0643] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0644] Case 1b: The wireless transmission bandwidth remains unchanged (i.e., the transmission rate remains unchanged), but the transmission reliability decreases.

[0645] Service latency: Since the wireless transmission rate remains constant, there is no need to adjust the computation latency.

[0646] Service accuracy: Due to reduced transmission reliability, service accuracy is affected. Therefore, it is necessary to improve computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0647] Method 1: Improve model accuracy without affecting the overall latency. For example, use an AI model with higher inference accuracy (e.g., reduce pruning or use a more complex model) to ensure that the overall service accuracy meets the requirements;

[0648] Method 2: Improve data accuracy without affecting the overall latency. For example, collect higher-precision input data on the network side as input to the AI ​​model to ensure that the overall service accuracy meets the requirements; premise: the data collected on the network side can also be used as model input.

[0649] Method 3: Improve computational accuracy without affecting the overall latency. For example, increase the number of bits used to represent the input data (point C, closer to point B), the number of bits used to represent the model parameters (point C, closer to point B), the number of bits used to represent the intermediate data (point C, closer to the square of point B), and the number of bits used to represent the result data (point C, closer to the square of point B) to ensure that the overall service accuracy meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0650] Modification Strategy 1: Improve model accuracy.

[0651] Modify Strategy 2: Improve data accuracy.

[0652] Modification Strategy 3: Improve calculation accuracy.

[0653] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0654] Case 1d: The wireless transmission bandwidth remains unchanged (i.e., the transmission rate remains unchanged), but the transmission reliability is improved.

[0655] Service latency: Since the wireless transmission rate remains constant, there is no need to adjust the computation latency.

[0656] Service accuracy: Improved transmission reliability has impacted service accuracy. Therefore, it is necessary to reduce computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0657] Method 1: Reduce model accuracy without affecting the overall latency. For example, use an AI model with lower inference accuracy (e.g., increase pruning or use a simpler model) to ensure that the overall service accuracy meets the requirements;

[0658] Method 2: Reduce data precision without affecting the overall latency. For example, collect lower precision input data on the network side as input to the AI ​​model to ensure that the overall service precision meets the requirements; premise: the data collected on the network side can also be used as model input.

[0659] Method 3: Reduce computational precision without affecting the overall latency. For example, reduce the number of bits used to represent the input data (point C, less than point B), the number of bits used to represent the model parameters (point C, less than point B), the number of bits used to represent the intermediate data (point C, less than the square of point B), and the number of bits used to represent the result data (point C, less than the square of point B) to ensure that the overall service precision meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0660] Modification Strategy 1: Reduce model accuracy.

[0661] Modification Strategy 2: Reduce data precision.

[0662] Modification Strategy 3: Reduce calculation precision.

[0663] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0664] Case 1c: Wireless transmission bandwidth is increased (i.e., transmission rate is increased), while transmission reliability remains unchanged.

[0665] Service latency: As the wireless transmission rate increases, the transmission latency decreases. In order to keep the total service latency constant, the computation latency can be increased.

[0666] Service accuracy: Since the transmission reliability remains unchanged, the service accuracy is not affected. Therefore, there is no need to adjust the calculation accuracy, model accuracy, or data accuracy.

[0667] Modify the strategy: Increase computation latency (e.g., by reducing computational resources to ensure the total service latency remains unchanged).

[0668] Case 1c-2: Increased wireless transmission bandwidth (i.e., increased transmission rate) and improved transmission reliability.

[0669] Service latency: As the wireless transmission rate increases, the transmission latency decreases. In order to keep the total service latency constant, the computation latency can be increased.

[0670] Service accuracy: Improved transmission reliability has impacted service accuracy. Therefore, it is necessary to reduce computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0671] Method 1: Reduce model accuracy without affecting the overall latency. For example, use an AI model with lower inference accuracy (e.g., increase pruning or use a simpler model) to ensure that the overall service accuracy meets the requirements;

[0672] Method 2: Reduce data precision without affecting the overall latency. For example, collect lower precision input data on the network side as input to the AI ​​model to ensure that the overall service precision meets the requirements; premise: the data collected on the network side can also be used as model input.

[0673] Method 3: Reduce computational precision without affecting the overall latency. For example, reduce the number of bits used to represent the input data (point C, less than point B), the number of bits used to represent the model parameters (point C, less than point B), the number of bits used to represent the intermediate data (point C, less than the square of point B), and the number of bits used to represent the result data (point C, less than the square of point B) to ensure that the overall service precision meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0674] Modification Strategy 1: Increase computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged) and reduce model accuracy.

[0675] Modification Strategy 2: Increase computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged) and reduce data precision.

[0676] Modification Strategy 3: Increase computation latency (e.g., by providing more computational resources to ensure that the total service latency remains unchanged) and reduce computational precision.

[0677] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0678] Case 1c-3: Increased wireless transmission bandwidth (i.e., increased transmission rate) but decreased transmission reliability.

[0679] Service latency: As the wireless transmission rate increases, the transmission latency decreases. In order to keep the total service latency constant, the computation latency can be increased.

[0680] Service accuracy: Reduced transmission reliability impacts service accuracy; therefore, it's necessary to improve computational accuracy, model accuracy, and data accuracy to maintain overall service accuracy. Methods include at least one of the following:

[0681] Method 1: Improve model accuracy without affecting the overall latency. For example, use an AI model with higher inference accuracy (e.g., reduce pruning or use a more complex model) to ensure that the overall service accuracy meets the requirements;

[0682] Method 2: Improve data accuracy without affecting the overall latency. For example, collect higher-precision input data on the network side as input to the AI ​​model to ensure that the overall service accuracy meets the requirements; premise: the data collected on the network side can also be used as model input.

[0683] Method 3: Improve computational accuracy without affecting the overall latency. For example, increase the number of bits used to represent the input data (point C, closer to point B), the number of bits used to represent the model parameters (point C, closer to point B), the number of bits used to represent the intermediate data (point C, closer to the square of point B), and the number of bits used to represent the result data (point C, closer to the square of point B) to ensure that the overall service accuracy meets the requirements. Assume that different levels of bit representation can be predefined at point C, and different bit representation levels can be selected when adjustments are needed.

[0684] Modification Strategy 1: Increase computation latency to improve model accuracy.

[0685] Modified Strategy 2: Increase computation latency to improve data accuracy.

[0686] Modification Strategy 3: Increase computation latency to improve computation accuracy.

[0687] Or it could be a combination of strategies 1, 2, and 3 mentioned above.

[0688] Steps 2-3: The first network element receives the fourth message sent by the second network element. The fourth message is used to respond to a policy modification request. The fourth message includes at least one of the following:

[0689] The ID of the first policy (the policy ID before modification), the ID of the second policy, the second QoS information, the reason for updating the policy, the third QoS information after the update, the first change, the first configuration information, and the first action instruction.

[0690] The first change amount is used to indicate the change amount between the third QoS information and the QoS information corresponding to the first information.

[0691] The first configuration information is used to indicate the parameter configuration information corresponding to the third QoS information, and the parameter configuration information includes configuration information of at least one of the first parameter, the second parameter and the third parameter.

[0692] The first action instruction is used to increase or decrease the parameter corresponding to the second QoS information, wherein the parameter corresponding to the second QoS information includes at least one of the first parameter, the second parameter, the third parameter, the fourth parameter, and the fifth parameter.

[0693] Specifically, the second QoS information includes at least one of the following:

[0694] Computational latency, model accuracy, data accuracy, computational accuracy

[0695] The first change includes at least one of the following:

[0696] Reduced computation latency or reduced computation time;

[0697] Increased or decreased model accuracy;

[0698] Improved or reduced algorithm accuracy;

[0699] Increased or decreased data precision;

[0700] Increased or decreased computational accuracy;

[0701] The first action indication is an action indication used to achieve the third QoS information, including at least one of the following:

[0702] Increasing the number of bits used to represent the numerical value of the input data during calculation (point C, which is closer to point B) results in increased calculation precision.

[0703] The numerical representation of model parameters during calculation is increased (point C, which is closer to point B), resulting in increased calculation accuracy.

[0704] Increase the number of bits used to represent intermediate data during calculations (point C, which is closer to the square of point B), resulting in increased calculation precision.

[0705] Increase the number of bits used to represent the numerical result data during calculation (point C, which is closer to the square of point B), corresponding to increased calculation precision;

[0706] Reducing the number of bits used to represent the numerical value of the input data during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0707] Reducing the number of bits used to represent the numerical values ​​of model parameters during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0708] Reducing the number of bits used to represent intermediate data during calculations (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision.

[0709] Reduce the number of bits used to represent the numerical result data during calculation (point C, which is smaller than the square of point B), (corresponding to a reduction in calculation precision);

[0710] Using an AI model with higher inference accuracy (e.g., reducing pruning or using a more complex model) corresponds to increased model accuracy;

[0711] Use AI models with lower inference accuracy (e.g., increase pruning or use simpler models, which correspond to reduced model accuracy);

[0712] Using an algorithm with higher precision corresponds to increased algorithm precision;

[0713] Using an algorithm with lower precision corresponds to a reduction in algorithm precision;

[0714] Higher precision input data is collected on the network side, corresponding to increased data precision;

[0715] The network side collects lower-precision input data, which corresponds to a reduction in data precision;

[0716] Provide more computing resources, resulting in reduced computing latency;

[0717] This provides less computing resources, resulting in increased computing latency. The first configuration information is used to achieve the third QoS information and includes at least one of the following:

[0718] The numerical representation of the input data during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0719] The numerical representation of the model parameters during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0720] The numerical representation of intermediate data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0721] The numerical representation of the result data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0722] Inference AI model indicators (such as indicators of pruning level, or model identifiers, corresponding to increased or decreased model accuracy);

[0723] Algorithm indication, corresponding to increasing or decreasing algorithm precision;

[0724] The network side configures the input data collection settings to increase or decrease the data precision accordingly.

[0725] The allocation of computing resources corresponds to an increase or decrease in computing latency;

[0726] Step 3a-1 (optional): When the second QoS information includes model accuracy and / or algorithm accuracy, the first network element sends a fifth message to the fifth network element. This fifth message is used to modify the AI ​​subtask or algorithm subtask.

[0727] When the second QoS information includes model accuracy and / or algorithm accuracy, the third network element includes the fifth network element.

[0728] The fifth message includes at least one of the following:

[0729] ID of the first sub-policy (policy ID corresponding to the first sub-QoS information), ID of the second sub-policy (policy ID corresponding to the second sub-QoS information), first sub-QoS information, indication of the reason for updating the policy, second change amount, second configuration information, and second action indication;

[0730] The parameters of the first QoS information are a subset of the parameters of the second QoS information, including at least one of the following:

[0731] Model accuracy;

[0732] Algorithm accuracy.

[0733] The parameters of the second sub-QoS information are a subset of the parameters of the third QoS information, including at least one of the following:

[0734] Improved model accuracy;

[0735] Improved algorithm accuracy;

[0736] Reduced model accuracy;

[0737] Reduced algorithm accuracy.

[0738] The second configuration information is configuration information used to achieve the second sub-QoS information, including at least one of the following:

[0739] AI model indicators for inference (such as indicators of pruning level or model identifiers) correspond to increased or decreased model accuracy;

[0740] Algorithm indication, corresponding to increasing or decreasing algorithm precision.

[0741] The second action indication is an action indication used to achieve the second sub-QoS, including at least one of the following:

[0742] Using an AI model with higher inference accuracy (e.g., reducing pruning or using a more complex model) corresponds to increased model accuracy.

[0743] Using an AI model with lower inference accuracy (e.g., increasing pruning or using a simpler model) corresponds to a reduction in model accuracy.

[0744] Using an algorithm with higher precision corresponds to increased algorithm precision.

[0745] Using an algorithm with lower precision corresponds to a reduction in algorithm precision.

[0746] Step 3a-2 (optional): The fifth network element sends a sixth message to the first network element. The sixth message is a response to the fifth message and is used to indicate model information, including at least one of the following: pruning degree indication, model, and model download address.

[0747] Step 3b-1 (optional):

[0748] When the second QoS includes data precision, the first network element sends a fifth message to the fourth network element. This fifth message is used to modify the data subtask.

[0749] When the second QoS information includes data precision, the third network element includes the fourth network element.

[0750] The fifth message includes at least one of the following:

[0751] ID of the first sub-policy (policy ID corresponding to the first sub-QoS information), ID of the second sub-policy (policy ID corresponding to the second sub-QoS information), first sub-QoS information, indication of the reason for updating the policy, second change amount, second configuration information, and second action indication;

[0752] The parameters of the first QoS information are a subset of the parameters of the second QoS information, including data precision.

[0753] The parameters of the second sub-QoS information are a subset of the parameters of the third QoS information, including at least one of the following:

[0754] Increased or decreased data precision.

[0755] The second configuration is the configuration information used to achieve the second sub-QoS, including the network side's configuration for collecting input data, which corresponds to increasing or decreasing the data precision.

[0756] The second action indication is an action indication used to achieve the second sub-QoS, including at least one of the following:

[0757] Higher precision input data is collected on the network side, corresponding to increased data precision;

[0758] The network side collects input data with lower precision, which corresponds to a reduction in data accuracy.

[0759] Step 3b-2 (optional):

[0760] The fourth network element sends a sixth message to the first network element. This sixth message is a response to the fifth message and is used to indicate data source information, including at least one of the following: quantization or numerical representation bits of the data acquisition, and the data source address.

[0761] Step 3c-1 (optional):

[0762] When the second QoS information includes the calculation QoS, the first network element sends a fifth message to the sixth network element to modify the calculation subtask.

[0763] When the second QoS information includes QoS calculation, the third network element includes the sixth network element.

[0764] For example, the above calculation of QoS includes calculating latency and / or calculating accuracy.

[0765] The fifth message includes at least one of the following:

[0766] ID of the first sub-policy (policy ID corresponding to the first sub-QoS information), ID of the second sub-policy (policy ID corresponding to the second sub-QoS information), first sub-QoS information, indication of the reason for updating the policy, second change amount, second configuration information, and second action indication;

[0767] The parameters of the first QoS information are a subset of the parameters of the second QoS information, including at least one of the following:

[0768] Calculation accuracy;

[0769] Calculate latency.

[0770] The parameters of the second sub-QoS information are a subset of the parameters of the third QoS information, including at least one of the following:

[0771] Reduced computation latency or increased computation latency;

[0772] Increased or decreased computational accuracy.

[0773] The second configuration information is configuration information used to achieve the second sub-QoS information, including at least one of the following:

[0774] The numerical representation of the input data during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0775] The numerical representation of the model parameters during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0776] The numerical representation of intermediate data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0777] The numerical representation of the result data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0778] The allocation of computing resources corresponds to an increase or decrease in computing latency.

[0779] The second action indication is an action indication used to achieve the second sub-QoS, including at least one of the following:

[0780] Increasing the number of bits used to represent the numerical value of the input data during calculation (point C, which is closer to point B) results in increased calculation precision.

[0781] Increasing the number of bits used to represent the numerical values ​​of model parameters during calculation (point C, which is closer to point B) results in increased calculation accuracy.

[0782] Increase the number of bits used to represent intermediate data during calculations (point C, which is closer to the square of point B), resulting in increased calculation precision.

[0783] Increase the number of bits used to represent the numerical result data during calculation (point C, which is closer to the square of point B), corresponding to increased calculation precision;

[0784] Reducing the number of bits used to represent the numerical value of the input data during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0785] Reducing the number of bits used to represent the numerical values ​​of model parameters during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0786] Reducing the number of bits used to represent intermediate data during calculations (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision.

[0787] Reducing the number of bits used to represent the numerical result data during calculation (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision;

[0788] Provide more computing resources, resulting in reduced computing latency;

[0789] It provides fewer computing resources, which corresponds to increased computing latency.

[0790] Step 3c-2 (optional):

[0791] After the sixth network element receives the fifth message, it sends a seventh message to the seventh network element. The seventh message is used to modify the configuration of the computation subtask. The seventh message includes at least one of the following:

[0792] ID of the first sub-policy (policy ID corresponding to the first sub-QoS information), ID of the second sub-policy (policy ID corresponding to the second sub-QoS information), first sub-QoS information, indication of the reason for updating the policy, second change amount, second configuration information, and second action indication;

[0793] The parameters of the first QoS information are a subset of the parameters of the second QoS information, including at least one of the following:

[0794] Calculation accuracy;

[0795] Calculate latency.

[0796] The parameters of the second sub-QoS information are a subset of the parameters of the third QoS information, including at least one of the following:

[0797] Reduced computation latency or increased computation latency;

[0798] Increased or decreased computational accuracy.

[0799] The second configuration information is configuration information used to achieve the second sub-QoS information, including at least one of the following:

[0800] The numerical representation of the input data during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0801] The numerical representation of the model parameters during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[0802] The numerical representation of intermediate data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0803] The numerical representation of the result data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[0804] The allocation of computing resources corresponds to an increase or decrease in computing latency.

[0805] The second action indication is an action indication used to achieve the second sub-QoS, including at least one of the following:

[0806] Increasing the number of bits used to represent the numerical value of the input data during calculation (point C, which is closer to point B) results in increased calculation precision.

[0807] Increasing the number of bits used to represent the numerical values ​​of model parameters during calculation (point C, which is closer to point B) results in increased calculation accuracy.

[0808] Increase the number of bits used to represent intermediate data during calculations (point C, which is closer to the square of point B), resulting in increased calculation precision.

[0809] Increase the number of bits used to represent the numerical result data during calculation (point C, which is closer to the square of point B), corresponding to increased calculation precision;

[0810] Reducing the number of bits used to represent the numerical value of the input data during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0811] Reducing the number of bits used to represent the numerical values ​​of model parameters during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[0812] Reducing the number of bits used to represent intermediate data during calculations (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision.

[0813] Reducing the number of bits used to represent the numerical result data during calculation (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision;

[0814] Provide more computing resources, resulting in reduced computing latency;

[0815] It provides fewer computing resources, which corresponds to increased computing latency.

[0816] Step 3c-3 (optional):

[0817] The seventh network element sends a seventh message in response to the sixth network element. This is used to indicate AI resources or computing power resources. It includes at least one of the following:

[0818] The numerical representation of the input data during calculation;

[0819] The numerical representation of model parameters during calculation;

[0820] The numerical representation of intermediate data during calculation;

[0821] The numerical representation of the result data during calculation;

[0822] Calculate the quantity or type of resources.

[0823] Steps 3c-4 (optional):

[0824] The sixth network element sends a sixth message to the first network element. This sixth message is a response to the fifth message and is used to indicate AI resources or computing power resources. It includes at least one of the following:

[0825] IP address;

[0826] ID information;

[0827] Fully Qualified Domain Name (FQDN);

[0828] The numerical representation of the input data during calculation;

[0829] The numerical representation of model parameters during calculation;

[0830] The numerical representation of intermediate data during calculation;

[0831] The numerical representation of the result data during calculation;

[0832] Calculate the quantity or type of resources.

[0833] Step 4: The first network element sends a task modification response to the terminal.

[0834] Example 5: Example 4 is a network-initiated subtask QoS modification, while this example is a UE-initiated task QoS modification.

[0835] like Figure 15 As shown, the scheme includes:

[0836] Step 1: The UE sends a second message to the first network element. The second message includes: task modification related information.

[0837] The task modification information includes at least one of the following:

[0838] The identifier of the first strategy, which includes the strategy corresponding to the second business requirement, where the second business requirement is the business requirement before modification.

[0839] Indicator of the reason for updating the strategy;

[0840] Instructions for modifying the second business requirement;

[0841] Instructions for the second business requirement;

[0842] The first business requirement is the modified business requirement.

[0843] The second message can be forwarded via, for example, AMF, or IMS, or sent directly to the first network element.

[0844] The reason for the update strategy includes at least one of the following:

[0845] The UE's computing power is reduced, resulting in reduced display accuracy (corresponding to situation 1 below);

[0846] The UE's computing power is reduced, but the display accuracy remains unchanged (corresponding to case 2 below);

[0847] Improved UE computing power leads to improved display accuracy (corresponding to case 3 below);

[0848] The UE's computing power is improved, but the display accuracy remains unchanged (corresponding to case 4 below);

[0849] With the UE computing power and display accuracy remaining unchanged, the computational requirements per unit time increase (corresponding to situation 5 below);

[0850] With the UE computing power and display accuracy remaining unchanged, the computational requirements per unit time are reduced (corresponding to case 6 below).

[0851] The modification instruction for the second business requirement is used to indicate at least one of the following:

[0852] The network needs to offload more or less computing power;

[0853] The network needs to provide lower or higher service accuracy.

[0854] Scenario 1: The UE's power level decreases, resulting in a decrease in the UE's computing power and display accuracy.

[0855] The network requirements (primary business requirements) include at least one of the following:

[0856] More computing power needs to be offloaded from the network.

[0857] The network needs to provide lower service accuracy.

[0858] Case 1-1: The second message only includes an indication of the reason for updating the strategy, for example:

[0859] The change in UE computing power, for example, the difference in computing power after switching from normal mode to power saving mode;

[0860] Display the change in display accuracy, such as the difference after switching from normal mode to power saving mode;

[0861] or

[0862] The UE's secondary computing power, for example, the computing power after switching from normal mode to power-saving mode;

[0863] The second display accuracy is the display accuracy after switching from normal mode to power saving mode;

[0864] or,

[0865] The change in the amount of computation required per unit time for the UE, for example, the change in computation after expanding the map with a new one.

[0866] The first service requirement is determined by the network side, such as the first network element and / or the second network element, that is, the service requirement after the terminal switches from normal mode to power saving mode.

[0867] Scenario 1-2: The second message only includes the first business requirement, for example:

[0868] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from normal mode to power saving mode.

[0869] The first service precision required by the network is the service precision required after the terminal switches from normal mode to power saving mode.

[0870] or,

[0871] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from normal mode to power saving mode.

[0872] The service accuracy change value provided by the network is required. This service accuracy change value is the difference in service accuracy after the terminal switches from normal mode to power saving mode.

[0873] or,

[0874] The change in computing power that needs to be unloaded by the network is the value of computing power that needs to be unloaded by the network after the terminal expands the map with new data.

[0875] Scenario 1-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0876] Scenario 2: The UE's power level decreases, resulting in a decrease in the UE's computing power, while the display accuracy remains unchanged.

[0877] Network requirements (primary business requirement):

[0878] More computing power needs to be offloaded from the network.

[0879] Case 2-1: The fifth message only includes the reason for updating the strategy, for example:

[0880] The change in UE computing power, for example, the difference in computing power after switching from normal mode to power-saving mode.

[0881] or,

[0882] The UE's secondary computing power, for example, the computing power after switching from normal mode to power-saving mode;

[0883] The first service requirement is determined by the network side, such as the first network element and / or the second network element, that is, the service requirement after the terminal switches from normal mode to power saving mode.

[0884] Scenario 2-2: The second message only includes the first business requirement, for example:

[0885] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from normal mode to power saving mode.

[0886] or,

[0887] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from normal mode to power saving mode.

[0888] Scenario 2-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0889] Scenario 3: Increased head-mounted display power in the UE leads to improved computing power and display accuracy.

[0890] The network requirements (primary business requirements) include at least one of the following:

[0891] It requires less computing power to be offloaded by the network.

[0892] The network needs to provide higher service accuracy.

[0893] Case 3-1: The second message only includes an indication of the reason for updating the strategy, for example:

[0894] The change in UE computing power (the difference in computing power after switching from power saving mode to normal mode), and the change in display accuracy (the change in display accuracy after switching from power saving mode to normal mode);

[0895] or,

[0896] The UE's second computing power (computing power after switching from power saving mode to normal mode), and second display accuracy (display accuracy after switching from power saving mode to normal mode);

[0897] The first service requirement (the service requirement after switching from power saving mode to normal mode) is determined by the network side, such as the first network element and / or the second network element.

[0898] Scenario 3-2: The second message only includes the first business requirement, for example:

[0899] The third computing power that needs to be offloaded via the network is the computing power that the terminal needs to be offloaded via the network after switching from power saving mode to normal mode.

[0900] The first service precision required by the network is the service precision required after the terminal switches from power saving mode to normal mode.

[0901] or,

[0902] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from power saving mode to normal mode.

[0903] The service accuracy change value provided by the network is required. This service accuracy change value is the difference after the terminal switches from power saving mode to normal mode.

[0904] Scenario 3-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0905] Scenario 4: The UE's head-mounted display battery level increases, resulting in increased UE computing power, while the display accuracy remains unchanged.

[0906] Network requirements (primary business requirement):

[0907] It requires less computing power to be offloaded by the network.

[0908] Case 4-1: The second message only includes an indication of the reason for updating the strategy, for example:

[0909] The change in UE computing power, for example, the difference in computing power after switching from power saving mode to normal mode;

[0910] or,

[0911] The UE's secondary computing power, for example, the computing power after switching from power saving mode to normal mode;

[0912] The first service requirement is determined by the network side, such as the first network element and / or the second network element, which is the service requirement after the terminal switches from power saving mode to normal mode.

[0913] Scenario 4-2: The fifth message only includes the first business requirement, for example:

[0914] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from power saving mode to normal mode.

[0915] or,

[0916] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from power saving mode to normal mode.

[0917] Scenario 4-3: The fifth message includes both the reason for updating the strategy and the primary business requirement.

[0918] Scenario 5: The UE has downloaded a new map, which increases the amount of computation required per unit time.

[0919] The network requirements (primary business requirements) include:

[0920] More computing power needs to be offloaded from the network.

[0921] Scenario 5-1: The fifth message only includes the first business requirement, for example:

[0922] (After the new map is expanded) the change in computing power required for network unloading.

[0923] Case 6: The UE has deleted some areas of the map, which reduces the amount of computation that needs to be processed per unit time;

[0924] The network requirements (primary business requirements) include:

[0925] It requires less computing power to be offloaded by the network.

[0926] Scenario 6-1: The second message only includes the first business requirement, for example:

[0927] (After deleting part of the map) the change in computing power required for network unloading.

[0928] Step 2-1: The first network element decides to adjust or modify the policy and sends a third message to the second network element. The third message is used to request a policy update. The third message includes at least one of the following: the ID of the first policy, an indication of the reason for updating the policy, first QoS information, and second QoS information that needs to be updated;

[0929] In this embodiment, the reason for updating the strategy includes at least one of the following:

[0930] Improved UE computing power;

[0931] The reduction in UE computing power;

[0932] The display accuracy of the UE is improved;

[0933] The display accuracy of the UE is reduced;

[0934] The amount of computation required per unit time for the UE increases;

[0935] The amount of computation required per unit time for the UE is reduced.

[0936] The first QoS information is QoS information related to the reason for the update policy.

[0937] In this embodiment, the first QoS information includes at least one of the following:

[0938] Calculation accuracy;

[0939] Calculate latency;

[0940] Calculation speed.

[0941] In this embodiment, the second QoS information includes at least one of the following:

[0942] Algorithm accuracy, model accuracy, computational accuracy, data accuracy, inference latency, computational latency.

[0943] Step 2-2: Adjust or modify the strategy of the second network element.

[0944] In this embodiment, the second QoS information includes at least one of the following:

[0945] Algorithm accuracy, model accuracy, computational accuracy, data accuracy, inference latency, computational latency

[0946] Scenario 1: The UE's power level decreases, resulting in a decrease in the UE's computing power and display accuracy.

[0947] Case 1-1: The second message only includes the reason for updating the strategy, for example:

[0948] The change in UE computing power, for example, the difference in computing power after switching from normal mode to power saving mode;

[0949] Display the change in display accuracy, such as the difference after switching from normal mode to power saving mode;

[0950] or

[0951] The UE's secondary computing power, for example, the computing power after switching from normal mode to power-saving mode;

[0952] The second display accuracy is the display accuracy after switching from normal mode to power saving mode;

[0953] Scenario 1-2: The second message only includes the first business requirement, for example:

[0954] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from normal mode to power saving mode.

[0955] The first service precision required by the network is the service precision required after the terminal switches from normal mode to power saving mode.

[0956] or,

[0957] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from normal mode to power saving mode.

[0958] The service accuracy change value provided by the network is required. This service accuracy change value is the difference in service accuracy after the terminal switches from normal mode to power saving mode.

[0959] Scenario 1-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0960] Based on the above, the second network element determines the modification strategy.

[0961] Modification Strategy 1: Increase computation speed, reduce computation precision;

[0962] Modification Strategy 2: Increase computation speed, reduce algorithm precision;

[0963] Modification Strategy 3: Increase computation speed, reduce data precision;

[0964] Modification Strategy 4: Increase computing speed and communication speed, reduce computing precision;

[0965] Modification Strategy 5: Increase computation speed and communication speed, but reduce algorithm precision;

[0966] Modification Strategy 6: Increase computation speed, increase communication speed, and reduce algorithm precision.

[0967] Scenario 2: The UE's power level decreases, resulting in a decrease in the UE's computing power, while the display accuracy remains unchanged.

[0968] Case 2-1: The fifth message only includes the reason for updating the strategy, for example:

[0969] The change in UE computing power, for example, the difference in computing power after switching from normal mode to power-saving mode.

[0970] or,

[0971] The UE's secondary computing power, for example, the computing power after switching from normal mode to power-saving mode;

[0972] Scenario 2-2: The second message only includes the first business requirement, for example:

[0973] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from normal mode to power saving mode.

[0974] or,

[0975] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from normal mode to power saving mode.

[0976] Scenario 2-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0977] Based on the above, the second network element determines the modification strategy.

[0978] Modify Strategy 1: Improve computation speed;

[0979] Modify Strategy 2: Improve computing speed and communication speed;

[0980] Scenario 3: Increased head-mounted display power in the UE leads to improved computing power and display accuracy.

[0981] Case 3-1: The second message only includes an indication of the reason for updating the strategy, for example:

[0982] The change in UE computing power (the difference in computing power after switching from power saving mode to normal mode), and the change in display accuracy (the change in display accuracy after switching from power saving mode to normal mode);

[0983] or,

[0984] The UE's second computing power (computing power after switching from power saving mode to normal mode), and second display accuracy (display accuracy after switching from power saving mode to normal mode);

[0985] The first service requirement (the service requirement after switching from power saving mode to normal mode) is determined by the network side, such as the first network element and / or the second network element.

[0986] Scenario 3-2: The second message only includes the first business requirement, for example:

[0987] The third computing power that needs to be offloaded via the network is the computing power that the terminal needs to be offloaded via the network after switching from power saving mode to normal mode.

[0988] The first service precision required by the network is the service precision required after the terminal switches from power saving mode to normal mode.

[0989] or,

[0990] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from power saving mode to normal mode.

[0991] The service accuracy change value provided by the network is required. This service accuracy change value is the difference after the terminal switches from power saving mode to normal mode.

[0992] Scenario 3-3: The second message includes both the reason for updating the strategy and the primary business requirement.

[0993] Based on the above, the second network element determines the modification strategy.

[0994] Modification Strategy 1: Reduce computation speed and improve computation accuracy;

[0995] Modification Strategy 2: Reduce computation speed and improve algorithm accuracy;

[0996] Modification Strategy 3: Reduce computation speed and improve data accuracy;

[0997] Modification Strategy 4: Reduce computation speed, reduce communication speed, and improve computation accuracy;

[0998] Modification Strategy 5: Reduce computation speed, reduce communication speed, and improve algorithm accuracy;

[0999] Modification Strategy 6: Reduce computation speed, reduce communication speed, and improve algorithm accuracy.

[1000] Scenario 4: The UE's head-mounted display battery level increases, resulting in increased UE computing power, while the display accuracy remains unchanged.

[1001] Network requirements (primary business requirement):

[1002] It requires less computing power to be offloaded by the network.

[1003] Case 4-1: The second message only includes an indication of the reason for updating the strategy, for example:

[1004] The change in UE computing power, for example, the difference in computing power after switching from power saving mode to normal mode;

[1005] or,

[1006] The UE's secondary computing power, for example, the computing power after switching from power saving mode to normal mode;

[1007] The first service requirement is determined by the network side, such as the first network element and / or the second network element, which is the service requirement after the terminal switches from power saving mode to normal mode.

[1008] Scenario 4-2: The fifth message only includes the first business requirement, for example:

[1009] The third computing power that needs to be offloaded by the network is the computing power that needs to be offloaded by the network after the terminal switches from power saving mode to normal mode.

[1010] or,

[1011] The computing power change value that needs to be offloaded by the network is the difference in computing power after the terminal switches from power saving mode to normal mode.

[1012] Scenario 4-3: The fifth message includes both the reason for updating the strategy and the primary business requirement.

[1013] Based on the above, the second network element determines the modification strategy.

[1014] Modification Strategy 1: Reduce computation speed;

[1015] Modified Strategy 2: Reduce computing speed and reduce communication speed;

[1016] Scenario 5: The UE has downloaded a new map, which increases the amount of computation required per unit time.

[1017] The network requirements (primary business requirements) include:

[1018] More computing power needs to be offloaded from the network.

[1019] Scenario 5-1: The fifth message only includes the first business requirement, for example:

[1020] (After the new map is expanded) the change in computing power required for network unloading.

[1021] Based on the above, the second network element determines the modification strategy.

[1022] Modify Strategy 1: Improve computation speed;

[1023] Modified Strategy 2: Improve computing speed and communication speed.

[1024] Case 6: The UE has deleted some areas of the map, which reduces the amount of computation that needs to be processed per unit time.

[1025] The network requirements (primary business requirements) include:

[1026] It requires less computing power to be offloaded by the network.

[1027] Scenario 6-1: The second message only includes the first business requirement, for example:

[1028] (After deleting part of the map) the change in computing power required for network unloading.

[1029] Based on the above, the second network element determines the modification strategy.

[1030] Modification Strategy 1: Reduce computation speed;

[1031] Modified Strategy 2: Reduce computing speed and reduce communication speed.

[1032] Steps 2-3: The first network element receives the fourth message sent by the second network element. The fourth message is used to respond to a policy modification request. The fourth message includes at least one of the following:

[1033] The ID of the first policy, the ID of the second policy, the second QoS information, the reason for updating the policy, the first change, the first configuration information, and the first action indication.

[1034] in,

[1035] The second QoS information includes at least one of the following:

[1036] Computational latency, model accuracy, data accuracy, computational accuracy;

[1037] The first change includes at least one of the following:

[1038] Increased or decreased communication rate;

[1039] Increased or decreased computing speed;

[1040] Reduced or increased computational latency;

[1041] Improve or reduce model accuracy;

[1042] Improve or reduce algorithm accuracy;

[1043] Increase or decrease data precision;

[1044] Increase or decrease the accuracy of calculations.

[1045] The first action indication is an action indication used to achieve the third QoS information, including at least one of the following:

[1046] Increasing the number of bits used to represent the numerical values ​​of the input data during calculation (point C, which is closer to point B) results in increased calculation precision.

[1047] Improving the numerical representation of model parameters during calculation (point C, which is closer to point B) results in increased calculation accuracy.

[1048] Increasing the number of bits used to represent intermediate data during calculations (point C, which is closer to the square of point B) results in increased calculation precision.

[1049] The numerical representation of the result data during calculation is increased (point C, which is closer to the square of point B), corresponding to an increase in calculation precision.

[1050] Reducing the number of bits used to represent the numerical values ​​of the input data during calculation (point C, which is smaller than point B) corresponds to a reduction in calculation precision.

[1051] Reducing the number of bits used to represent the model parameters during computation (point C, which is smaller than point B) corresponds to a reduction in computational precision.

[1052] Reducing the number of bits used to represent intermediate data during calculations (point C, which is less than the square of point B) corresponds to a reduction in calculation precision.

[1053] Reducing the number of bits used to represent the numerical result data during calculation (point C, which is smaller than the square of point B) corresponds to a reduction in calculation precision.

[1054] Using an AI model with higher inference accuracy (e.g., reducing pruning or using a more complex model) corresponds to increased model accuracy.

[1055] Using an AI model with lower inference accuracy (e.g., increasing pruning or using a simpler model) corresponds to a reduction in model accuracy.

[1056] Using an algorithm with higher precision corresponds to increased algorithm precision.

[1057] Using an algorithm with lower precision corresponds to a reduction in algorithm precision.

[1058] Higher precision input data is collected on the network side, corresponding to increased data accuracy.

[1059] The network side collects input data with lower precision, which corresponds to a reduction in data accuracy.

[1060] Providing more computing resources results in reduced computing latency and / or reduced computing speed.

[1061] Providing fewer computing resources corresponds to increased computing latency and / or reduced computing speed.

[1062] The first configuration information is configuration information used to achieve the third QoS information, including at least one of the following:

[1063] The numerical representation of the input data during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[1064] The numerical representation of the model parameters during calculation (point C, which is closer to point B) corresponds to an increase or decrease in calculation precision.

[1065] The numerical representation of intermediate data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[1066] The numerical representation of the result data during calculation (point C, which is closer to the square of point B) corresponds to an increase or decrease in calculation precision.

[1067] AI model indicators for inference (such as indicators of pruning level or model identifiers) correspond to increases or decreases in model accuracy.

[1068] Algorithm indication, corresponding to increasing or decreasing algorithm precision.

[1069] The network side configures the input data collection, which can increase or decrease the data precision accordingly.

[1070] The allocation of computing resources corresponds to an increase or decrease in computing latency.

[1071] Step 3: Same as step 3 in Example 4.

[1072] The difference lies in the fact that existing technologies can be reused to modify the QoS of communication.

[1073] Step 4: The first network element sends an eighth message to the UE. The eighth message is used in response to the second message. The eighth message includes at least one of the following:

[1074] First task ID and / or first task session identifier, where the first task is the task before modification;

[1075] The second task ID and / or the second task session identifier, wherein the second task is the task corresponding to the first business requirement;

[1076] The second subtask identifier and / or the second subtask session identifier, wherein the second subtask is a subtask of the second task;

[1077] The second subtask execution function identifier, represented in formats including but not limited to ID, IP address, and FQDN;

[1078] The second subtask management function identifier, represented in formats including but not limited to ID, IP address, and FQDN;

[1079] Second task strategy / QoS;

[1080] Second subtask policy / QoS;

[1081] The execution order of the second subtask;

[1082] Second task slice and / or second task DNN;

[1083] Second subtask slice and / or second subtask DNN;

[1084] If the eighth message is a rejection message, it may optionally carry the rejection reason, including but not limited to: the number of task requests exceeds the maximum value; task requests are not allowed; the UE is illegitimate.

[1085] In this embodiment, the first network element acquires a first service requirement of the target service; the first network element adjusts first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability. In the above scheme, the first information is adjusted so that the adjusted first information can meet the first service accuracy in the first service requirement, thus meeting the user's service accuracy when the wireless transmission channel changes or user needs change, without occupying a large amount of storage space to store the inference model or algorithm.

[1086] The information adjustment method provided in this application can be executed by an information adjustment device. This application uses an information adjustment device executing the information adjustment method as an example to illustrate the information adjustment device provided in this application.

[1087] This application provides an information adjustment device. As an example, the information adjustment device may be a communication device or a component within a communication device, such as a chip. The communication device may be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal may include, but is not limited to, the type of terminal 11 listed above, and the network-side device may include, but is not limited to, the type of network-side device 12 listed above. This application does not impose specific limitations.

[1088] The information adjustment device includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware. When implemented in hardware, the processing module can be implemented by a processor. For example, the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc. The receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.

[1089] See Figure 16 When the information adjustment device is a network-side device or a component of a network-side device, the information adjustment device 1600 includes a first receiving module 1601 for acquiring a first service requirement of the target service; and a first processing module 1602 for adjusting the first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability information.

[1090] In this embodiment of the application, the model accuracy is adjusted by a first parameter; the first parameter includes at least one of the following: the quantization accuracy of the model parameters, the data type of the model parameters, the number of model parameters, the structure of the model, and the degree of pruning of the model structure.

[1091] And / or, the calculation precision is adjusted by a second parameter; the second parameter includes at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the model parameters configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation.

[1092] And / or, the data precision is adjusted by a third parameter; the third parameter includes at least one of the following: the number of bits used to represent the numerical value during data acquisition, the number of bits used to represent the numerical value of the data to be transmitted, data integrity information, and data timeliness information;

[1093] And / or, the AI ​​inference accuracy is adjusted by a fourth parameter; the fourth parameter includes at least one of the following: the data accuracy, the model accuracy, and the computational accuracy;

[1094] And / or, the AI ​​processing accuracy is adjusted by a fifth parameter; the fifth parameter includes at least one of the following: the model accuracy, the calculation accuracy, and the data accuracy.

[1095] In this embodiment of the application, the first service requirement further includes a first service latency, and the first processing module is used for:

[1096] Based on the first service delay, at least one of the communication delay, computation delay, and data acquisition delay is adjusted.

[1097] In this embodiment of the application, the first processing module is used for:

[1098] The first information is adjusted based on the service QoS information or task QoS information corresponding to the first service accuracy.

[1099] In this embodiment of the application, the first processing module is used for:

[1100] Receive a first message, the first message including a notification of changes in first Quality of Service (QoS) information; adjust the first information according to the first message and the first service requirement;

[1101] The first QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS;

[1102] The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information;

[1103] The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy;

[1104] The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

[1105] In this embodiment of the application, the first receiving module is used for:

[1106] Receive a second message, the second message including: the first business requirement, or task modification related information, the task modification related information being used to determine the first business requirement;

[1107] Based on the second message, the first business requirement is determined;

[1108] The task modification information includes at least one of the following:

[1109] The identifier of the first strategy, which includes the strategy corresponding to the second business requirement, where the second business requirement is the business requirement before modification.

[1110] Indicator of the reason for updating the strategy;

[1111] Instructions for modifying the second business requirement;

[1112] Instructions for the second business requirement;

[1113] The first business requirement is the modified business requirement.

[1114] In this embodiment of the application, the reason for the modification instruction of the second service requirement or the update strategy in the second message includes at least one of the following:

[1115] Information on changes in the terminal's computing power;

[1116] Information on changes in the terminal's display resolution;

[1117] Information on changes in the terminal's computing requirements.

[1118] In this embodiment of the application, the first processing module is used for:

[1119] The first information of the AI ​​model or algorithm is updated according to the first business requirement to obtain the second information;

[1120] The second information includes at least one of the following:

[1121] The second QoS information that needs to be updated includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows.

[1122] The updated third QoS information includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows.

[1123] The first change amount is used to indicate the change amount between the third QoS information and the QoS information corresponding to the first information;

[1124] First configuration information, which is used to indicate parameter configuration information corresponding to the third QoS information, wherein the parameter configuration information includes configuration information of at least one of the first parameter, the second parameter and the third parameter;

[1125] A first action instruction is used to increase or decrease the parameter corresponding to the second QoS information, wherein the parameter corresponding to the second QoS information includes at least one of a first parameter, a second parameter, a third parameter, a fourth parameter, and a fifth parameter;

[1126] Indicator of the reason for updating the strategy;

[1127] The first strategy is an identifier, wherein the first strategy is a strategy corresponding to the first service requirement requested by the terminal before receiving the first message, or the first strategy is a strategy corresponding to the second service requirement requested by the terminal before receiving the second message.

[1128] The identifier of the second policy, which is the policy corresponding to the third QoS information.

[1129] In this embodiment of the application, the first processing module is used for:

[1130] A third message is sent to the second network element based on the first service accuracy, the third message being used to request an update to the first information;

[1131] Obtain the fourth message sent by the second network element, wherein the fourth message is a response message to the third message and includes the second information;

[1132] Based on the fourth message, obtain the second information;

[1133] The third message includes at least one of the following:

[1134] The identifier of the first strategy;

[1135] Indicator of the reason for updating the strategy;

[1136] The first QoS information;

[1137] The second QoS information needs to be updated.

[1138] In this embodiment of the application, it also includes:

[1139] The first sending module is used to send a fifth message to the third network element. The fifth message is used to request modification of the first sub-QoS information. The parameters in the first sub-QoS information are a subset of the parameters in the second QoS information. The first sub-QoS information includes model QoS information, data QoS information, computing power QoS information, and communication QoS information.

[1140] The second receiving module is used to acquire the sixth message sent by the third network element, wherein the sixth message is a response message to the fifth message;

[1141] The model QoS information includes at least one of the following: model accuracy, model inference latency, model training latency, and model monitoring latency; the data QoS information includes data accuracy, data acquisition latency, and data processing latency; the computing power QoS information includes computing latency or computing accuracy; and the communication QoS information includes at least one of the following: communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

[1142] In this embodiment of the application, the fifth message includes at least one of the following:

[1143] First sub-QoS information;

[1144] The updated second sub-QoS information is determined based on the third QoS information;

[1145] The second change amount is used to indicate the change amount between the second sub-QoS information and the sub-QoS information corresponding to the first information; the sub-QoS information corresponding to the first information is determined based on the QoS information corresponding to the first information.

[1146] The second configuration information is used to indicate the parameter configuration information corresponding to the second sub-QoS information;

[1147] The second action instruction is used to increase or decrease the parameter corresponding to the first sub-QoS information, and the parameter corresponding to the first sub-QoS information includes at least one of the first parameter, the second parameter, the third parameter, the fourth parameter, and the fifth parameter.

[1148] Indicator of the reason for updating the strategy;

[1149] The identifier of the first sub-policy, where the first sub-policy is the policy corresponding to the first sub-QoS information;

[1150] The identifier of the second sub-policy, which is the policy corresponding to the second sub-QoS information.

[1151] In this embodiment of the application, when the first sub-QoS information includes the model QoS information, the third network element includes AI-related functional nodes;

[1152] Alternatively, if the first sub-QoS information includes the data QoS information, the third network element includes a data-related functional node;

[1153] Alternatively, if the first sub-QoS information includes the computing power QoS information, the third network element includes a functional node related to computing power.

[1154] In this embodiment of the application, upon receiving the first message, the reason indication for the update strategy includes at least one of the following: change information of communication QoS, change information of computing quality QoS, and change information of data QoS.

[1155] In this embodiment of the application, the first processing module is used for:

[1156] When the communication QoS change notification indicates a decrease in transmission rate, and the communication reliability information indicates that the communication reliability remains unchanged, the adjustment of the first information includes: reducing the calculation latency;

[1157] When the communication QoS change notification indicates a decrease in transmission rate and a decrease in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy;

[1158] When the communication QoS change notification indicates a decrease in transmission rate and an increase in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy;

[1159] When the communication QoS change notification indicates that the transmission rate remains unchanged but the communication reliability is reduced, the adjustment of the first information includes at least one of the following: improving model accuracy; improving data accuracy; improving calculation accuracy;

[1160] When the communication QoS change notification indicates that the transmission rate remains unchanged and the communication reliability is improved, the adjustment to the first information includes at least one of the following: reducing model accuracy; reducing data accuracy; reducing calculation accuracy;

[1161] When the communication QoS change notification indicates an increase in transmission rate and the communication reliability remains unchanged, the adjustment to the first information includes: increasing the calculation delay;

[1162] When the communication QoS change notification indicates an increase in transmission rate and an increase in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy;

[1163] When the communication QoS change notification indicates an increase in transmission rate and a decrease in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy.

[1164] In this embodiment of the application, the first processing module is configured to perform at least one of the following:

[1165] When the second message indicates that the terminal's computing power has decreased and the display accuracy has decreased, the adjustment to the first message includes at least one of the following: increasing the computing speed, decreasing the computing accuracy, decreasing the algorithm accuracy, decreasing the data accuracy, and increasing the communication speed;

[1166] When the second message indicates that the terminal's computing power is reduced but the display accuracy remains unchanged, the adjustment to the first information includes at least one of the following: increasing the computing speed and increasing the communication speed;

[1167] When the second message indicates that the terminal's computing power and display accuracy have increased, the adjustment to the first information includes at least one of the following: reducing the computing speed, increasing the computing accuracy, increasing the algorithm accuracy, increasing the data accuracy, and reducing the communication speed;

[1168] When the computing power of the second message terminal is increased while the display accuracy remains unchanged, the adjustment of the first information includes at least one of the following: reducing the computing rate and reducing the communication rate;

[1169] When the second message indicates that the terminal's computational demand per unit time increases, the adjustment to the first message includes at least one of the following: increasing the computational rate and increasing the communication rate;

[1170] When the second message indicates that the terminal's computational requirements per unit time are reduced, the adjustment to the first message includes at least one of the following: reducing the computational rate and reducing the communication rate.

[1171] In the embodiments of this application,

[1172] When the AI ​​unit is provided by a network-side device, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the network-side device, the model structure of the AI ​​unit from the network-side device, and the degree of pruning of the model structure of the AI ​​unit from the network-side device.

[1173] When the AI ​​unit is provided by the terminal, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the terminal, the model structure of the AI ​​unit from the terminal, and the degree of pruning of the model structure of the AI ​​unit from the terminal.

[1174] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the terminal is represented by the number of bits, and the numerical value of the first data is represented by the number of bits;

[1175] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the network-side device, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the network-side device is represented by the number of bits, and the numerical value of the first data is represented by the number of bits;

[1176] When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal and the network side device, the data accuracy is adjusted by at least one of the following: the value of the data collected by the network side device is represented by the number of bits, the value of the data collected by the terminal is represented by the number of bits, and the value of the first data is represented by the number of bits.

[1177] When the service precision is set to AI-specific service precision, the calculation precision is adjusted by at least one of the following: the number of bits used to represent the numerical values ​​of the input data configured during calculation, the number of bits used to represent the numerical values ​​of the model parameters configured during calculation, the number of bits used to represent the numerical values ​​of the intermediate variables configured during calculation, and the number of bits used to represent the numerical values ​​of the calculation results configured during calculation.

[1178] When the service precision is for non-AI services, the calculation precision is adjusted by at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation.

[1179] The first data includes data to be transmitted that the network sends to the terminal.

[1180] In this embodiment, the first network element acquires the first service requirement of the target service; the first network element adjusts the first information according to the first service requirement, wherein the first service requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability. In the above scheme, the first information is adjusted so that the adjusted first information can meet the first service accuracy in the first service requirement, thus meeting the user's service accuracy when the wireless transmission channel changes or user needs change, without occupying a large amount of storage space to store the inference model or algorithm.

[1181] The information adjustment device provided in this application embodiment can achieve... Figures 2 to 15 The various processes implemented in the method embodiments achieve the same technical effect, and will not be described again here to avoid repetition.

[1182] like Figure 17 As shown in the illustration, this application also provides a communication device 1700, including a processor 1701 and a memory 1702. The memory 1702 stores a program or instructions that can run on the processor 1701. For example, when the communication device 1700 is a network-side device, the program or instructions, when executed by the processor 1701, implement the various steps of the above-described information adjustment method embodiments and achieve the same technical effect. To avoid repetition, further details are omitted here.

[1183] This application embodiment also provides a network-side device, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement, for example... Figure 2 The steps of the method embodiment shown are illustrated. This network-side device embodiment corresponds to the above-described network-side device method embodiment. All implementation processes and methods of the above-described method embodiments can be applied to this network-side device embodiment and can achieve the same technical effect.

[1184] Specifically, embodiments of this application also provide a network-side device, which may be... Figure 16 The information adjustment device shown. (As shown) Figure 18 As shown, the network-side device 1800 includes: an antenna 181, a radio frequency (RF) device 182, a baseband device 183, a processor 184, and a memory 185. The antenna 181 is connected to the RF device 182. In the uplink direction, the RF device 182 receives information through the antenna 181 and transmits the received information to the baseband device 183 for processing. In the downlink direction, the baseband device 183 processes the information to be transmitted and sends it to the RF device 182. The RF device 182 processes the received information and transmits it through the antenna 181.

[1185] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 183, which includes a baseband processor.

[1186] Baseband device 183 may include, for example, at least one baseband board on which multiple chips are disposed, such as Figure 18 As shown, one of the chips is, for example, a baseband processor, which is connected to the memory 185 via a bus interface to call the program in the memory 185 and execute the network device operation shown in the above method embodiment.

[1187] The network-side device may also include a network interface 186, such as a Common Public Radio Interface (CPRI).

[1188] Specifically, the network-side device 1800 in this application embodiment further includes: instructions or programs stored in memory 185 and executable on processor 184, wherein processor 184 calls the instructions or programs in memory 185 to execute. Figure 16 The methods executed by each module shown achieve the same technical effect, and to avoid repetition, they will not be described in detail here.

[1189] Specifically, embodiments of this application also provide a network-side device. For example... Figure 19 As shown, the network-side device 1900 includes: a processor 1901, a network interface 1902, and a memory 1903. This network-side device can be... Figure 16 The information adjustment device shown. The network interface 1902 is, for example, a common public radio interface (CPRI).

[1190] Specifically, the network-side device 1900 in this application embodiment further includes: instructions or programs stored in memory 1903 and executable on processor 1901, wherein processor 1901 calls the instructions or programs in memory 1903 to execute. Figure 16 The methods executed by each module shown achieve the same technical effect, and to avoid repetition, they will not be described in detail here.

[1191] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described information adjustment method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[1192] The processor mentioned above is the processor in the terminal described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[1193] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described information adjustment method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[1194] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[1195] This application also provides a computer program / program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described information adjustment method embodiments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[1196] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[1197] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[1198] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

1. An information adjustment method, characterized in that, The method includes: The first network element obtains the first business requirement of the target service; The first network element adjusts the first information according to the first service requirement, wherein the first service requirement includes the first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability information.

2. The method according to claim 1, characterized in that, The model accuracy is adjusted by a first parameter; the first parameter includes at least one of the following: the quantization accuracy of the model parameters, the data type of the model parameters, the number of model parameters, the structure of the model, and the degree of pruning of the model structure. And / or, the calculation precision is adjusted by a second parameter; the second parameter includes at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the model parameters configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation. And / or, the data precision is adjusted by a third parameter; the third parameter includes at least one of the following: the number of bits used to represent the numerical value during data acquisition, the number of bits used to represent the numerical value of the data to be transmitted, data integrity information, and data timeliness information; And / or, the AI ​​inference accuracy is adjusted by a fourth parameter; the fourth parameter includes at least one of the following: the data accuracy, the model accuracy, and the computational accuracy; And / or, the AI ​​processing accuracy is adjusted by a fifth parameter; the fifth parameter includes at least one of the following: the model accuracy, the calculation accuracy, and the data accuracy.

3. The method according to claim 1 or 2, characterized in that, The first service requirement also includes a first service latency, and the method further includes: The first network element adjusts at least one of the communication delay, computing delay, and data acquisition delay based on the first service delay.

4. The method according to any one of claims 1 to 3, characterized in that, The first network element adjusts the first information according to the first service requirement, including: The first network element adjusts the first information based on the service QoS information or task QoS information corresponding to the first service precision.

5. The method according to any one of claims 1 to 4, characterized in that, The first network element adjusts the first information according to the first service requirement, including: The first network element receives a first message, which includes a notification of changes in first Quality of Service (QoS) information; the first network element adjusts the first information according to the first message and the first service requirement. The first QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS; The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information; The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy; The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

6. The method according to any one of claims 1 to 4, characterized in that, The first network element acquires the first service requirement of the target service, including: The first network element receives a second message, which includes: the first service requirement, or task modification related information, wherein the task modification related information is used to determine the first service requirement; The first network element determines the first service requirement based on the second message; The task modification information includes at least one of the following: The identifier of the first strategy, which includes the strategy corresponding to the second business requirement, where the second business requirement is the business requirement before modification. Indicator of the reason for updating the strategy; Instructions for modifying the second business requirement; Indications of the second business requirement; The first business requirement is the modified business requirement.

7. The method according to claim 6, characterized in that, The reason for the modification instruction or update strategy of the second business requirement in the second message includes at least one of the following: Information on changes in the terminal's computing power; Information on changes in the terminal's display resolution; Information on changes in the terminal's computing requirements.

8. The method according to any one of claims 5 to 7, characterized in that, Adjustments were made to the first piece of information, including: The first network element updates the first information of the AI ​​model or algorithm according to the first service requirement to obtain the second information; The second information includes at least one of the following: The second QoS information that needs to be updated includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows. The updated third QoS information includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows. The first change amount is used to indicate the change amount between the third QoS information and the QoS information corresponding to the first information; First configuration information, which is used to indicate parameter configuration information corresponding to the third QoS information, wherein the parameter configuration information includes configuration information of at least one of the first parameter, the second parameter and the third parameter; A first action instruction is used to increase or decrease the parameter corresponding to the second QoS information, wherein the parameter corresponding to the second QoS information includes at least one of a first parameter, a second parameter, a third parameter, a fourth parameter, and a fifth parameter; Indicator of the reason for updating the strategy; The first strategy is an identifier, wherein the first strategy is a strategy corresponding to the first service requirement requested by the terminal before receiving the first message, or the first strategy is a strategy corresponding to the second service requirement requested by the terminal before receiving the second message. The identifier of the second policy, which is the policy corresponding to the third QoS information.

9. The method according to claim 8, characterized in that, The first network element updates the first information according to the first service requirement to obtain the second information, including: The first network element sends a third message to the second network element based on the first service accuracy, the third message being used to request an update to the first information; The first network element obtains the fourth message sent by the second network element, the fourth message being a response message to the third message, and the fourth message including the second information; The first network element obtains the second information based on the fourth message; The third message includes at least one of the following: The identifier of the first strategy; Indicator of the reason for updating the strategy; The first QoS information; The second QoS information.

10. The method according to claim 8 or 9, characterized in that, Also includes: The first network element sends a fifth message to the third network element. The fifth message is used to request modification of the first sub-QoS information. The parameters in the first sub-QoS information are a subset of the parameters in the second QoS information. The first sub-QoS information includes model QoS information, data QoS information, computing power QoS information, and communication QoS information. The first network element obtains the sixth message sent by the third network element, wherein the sixth message is a response message to the fifth message; The model QoS information includes at least one of the following: model accuracy, model inference latency, model training latency, and model monitoring latency; the data QoS information includes data accuracy, data acquisition latency, and data processing latency; the computing power QoS information includes computing latency or computing accuracy; and the communication QoS information includes at least one of the following: communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

11. The method according to claim 10, characterized in that, The fifth message includes at least one of the following: First sub-QoS information; The updated second sub-QoS information is determined based on the third QoS information; The second change amount is used to indicate the change amount between the second sub-QoS information and the sub-QoS information corresponding to the first information; The sub-QoS information corresponding to the first information is determined based on the QoS information corresponding to the first information; The second configuration information is used to indicate the parameter configuration information corresponding to the second sub-QoS information; The second action instruction is used to increase or decrease the parameter corresponding to the first sub-QoS information, and the parameter corresponding to the first sub-QoS information includes at least one of the first parameter, the second parameter, the third parameter, the fourth parameter, and the fifth parameter. Indicator of the reason for updating the strategy; The identifier of the first sub-policy, where the first sub-policy is the policy corresponding to the first sub-QoS information; The identifier of the second sub-policy, which is the policy corresponding to the second sub-QoS information.

12. The method according to claim 10 or 11, characterized in that, In the case where the first sub-QoS information includes the model QoS information, the third network element includes AI-related functional nodes; Alternatively, if the first sub-QoS information includes the data QoS information, the third network element includes a data-related functional node; Alternatively, if the first sub-QoS information includes the computing power QoS information, the third network element includes a functional node related to computing power.

13. The method according to any one of claims 8 to 12, characterized in that, Upon receiving the first message, the reason indication for the update strategy includes at least one of the following: change information of communication QoS, change information of computing quality QoS, and change information of data QoS.

14. The method according to any one of claims 8 to 13, characterized in that, Based on the first business requirement, the first information is adjusted, including at least one of the following: When the communication QoS change notification indicates a decrease in transmission rate, and the communication reliability information indicates that the communication reliability remains unchanged, the adjustment to the first information includes: reducing the calculation latency; When the communication QoS change notification indicates a decrease in transmission rate and a decrease in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy; When the communication QoS change notification indicates a decrease in transmission rate and an increase in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy; When the communication QoS change notification indicates that the transmission rate remains unchanged but the communication reliability is reduced, the adjustment of the first information includes at least one of the following: improving model accuracy; improving data accuracy; improving calculation accuracy; When the communication QoS change notification indicates that the transmission rate remains unchanged and the communication reliability is improved, the adjustment to the first information includes at least one of the following: reducing model accuracy; reducing data accuracy; reducing calculation accuracy; When the communication QoS change notification indicates an increase in transmission rate and the communication reliability remains unchanged, the adjustment to the first information includes: increasing the calculation delay; When the communication QoS change notification indicates an increase in transmission rate and an increase in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy; When the communication QoS change notification indicates an increase in transmission rate and a decrease in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy.

15. The method according to any one of claims 8 to 12, characterized in that, Based on the first business requirement, the first information is adjusted, including at least one of the following: When the second message indicates that the terminal's computing power has decreased and the display accuracy has decreased, the adjustment to the first message includes at least one of the following: increasing the computing speed, decreasing the computing accuracy, decreasing the algorithm accuracy, decreasing the data accuracy, and increasing the communication speed; When the second message indicates that the terminal's computing power is reduced but the display accuracy remains unchanged, the adjustment to the first information includes at least one of the following: increasing the computing speed and increasing the communication speed; When the second message indicates that the terminal's computing power and display accuracy have increased, the adjustment to the first information includes at least one of the following: reducing the computing speed, increasing the computing accuracy, increasing the algorithm accuracy, increasing the data accuracy, and reducing the communication speed; When the computing power of the second message terminal is increased while the display accuracy remains unchanged, the adjustment of the first information includes at least one of the following: reducing the computing speed and reducing the communication speed; When the second message indicates that the terminal's computational demand per unit time increases, the adjustment to the first message includes at least one of the following: increasing the computational rate and increasing the communication rate; When the second message indicates that the terminal's computational requirements per unit time are reduced, the adjustment to the first message includes at least one of the following: reducing the computational rate and reducing the communication rate.

16. The method according to any one of claims 1 to 15, characterized in that, When the AI ​​unit is provided by a network-side device, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the network-side device, the model structure of the AI ​​unit from the network-side device, and the degree of pruning of the model structure of the AI ​​unit from the network-side device. When the AI ​​unit is provided by the terminal, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the terminal, the model structure of the AI ​​unit from the terminal, and the degree of pruning of the model structure of the AI ​​unit from the terminal. When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the terminal is represented by the number of bits, and the numerical value of the first data is represented by the number of bits; When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the network-side device, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the network-side device is represented by the number of bits, and the numerical value of the first data is represented by the number of bits; When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal and the network side device, the data accuracy is adjusted by at least one of the following: the value of the data collected by the network side device is represented by the number of bits, the value of the data collected by the terminal is represented by the number of bits, and the value of the first data is represented by the number of bits. When the service precision is set to AI-specific service precision, the calculation precision is adjusted by at least one of the following: the number of bits used to represent the numerical values ​​of the input data configured during calculation, the number of bits used to represent the numerical values ​​of the model parameters configured during calculation, the number of bits used to represent the numerical values ​​of the intermediate variables configured during calculation, and the number of bits used to represent the numerical values ​​of the calculation results configured during calculation. When the service precision is for non-AI services, the calculation precision is adjusted by at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation. The first data includes data to be transmitted that the network sends to the terminal.

17. An information adjustment device, characterized in that, include: The first receiving module is used to obtain the first business requirements of the target service. The first processing module is used to adjust the first information according to the first business requirement, wherein the first business requirement includes a first service accuracy, and the first information includes at least one of the following: model accuracy, calculation accuracy, data accuracy, artificial intelligence (AI) processing accuracy, AI inference accuracy, and communication reliability information.

18. The apparatus according to claim 17, characterized in that, The model accuracy is adjusted by a first parameter; the first parameter includes at least one of the following: the quantization accuracy of the model parameters, the data type of the model parameters, the number of model parameters, the structure of the model, and the degree of pruning of the model structure. And / or, the calculation precision is adjusted by a second parameter; the second parameter includes at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the model parameters configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation. And / or, the data precision is adjusted by a third parameter; the third parameter includes at least one of the following: the number of bits used to represent the numerical value during data acquisition, the number of bits used to represent the numerical value of the data to be transmitted, data integrity information, and data timeliness information; And / or, the AI ​​inference accuracy is adjusted by a fourth parameter; the fourth parameter includes at least one of the following: the data accuracy, the model accuracy, and the computational accuracy; And / or, the AI ​​processing accuracy is adjusted by a fifth parameter; the fifth parameter includes at least one of the following: the model accuracy, the calculation accuracy, and the data accuracy.

19. The apparatus according to claim 17 or 18, characterized in that, The first business requirement also includes a first service latency, and the first processing module is used for: Based on the first service delay, at least one of the communication delay, computation delay, and data acquisition delay is adjusted.

20. The apparatus according to any one of claims 17 to 19, characterized in that, The first processing module is used for: The first information is adjusted based on the service QoS information or task QoS information corresponding to the first service accuracy.

21. The apparatus according to any one of claims 17 to 20, characterized in that, The first processing module is used for: Receive a first message, the first message including a notification of changes in first Quality of Service (QoS) information; adjust the first information according to the first message and the first service requirement; The first QoS information includes at least one of the following: communication QoS, computing quality QoS, and data QoS; The communication QoS includes at least one of the following: communication transmission rate, communication latency, and communication reliability information; The computational quality of service (QoS) includes at least one of the following: computational latency and computational accuracy; The data QoS includes at least one of the following: data accuracy, data acquisition latency, and data processing latency.

22. The apparatus according to any one of claims 17 to 20, characterized in that, The first receiving module is used for: Receive a second message, the second message including: the first business requirement, or task modification related information, the task modification related information being used to determine the first business requirement; Based on the second message, the first business requirement is determined; The task modification information includes at least one of the following: The identifier of the first strategy, which includes the strategy corresponding to the second business requirement, where the second business requirement is the business requirement before modification. Indicator of the reason for updating the strategy; Instructions for modifying the second business requirement; Indications of the second business requirement; The first business requirement is the modified business requirement.

23. The apparatus according to claim 22, characterized in that, The reason for the modification instruction or update strategy of the second business requirement in the second message includes at least one of the following: Information on changes in the terminal's computing power; Information on changes in the terminal's display resolution; Information on changes in the terminal's computing requirements.

24. The apparatus according to any one of claims 21 to 23, characterized in that, The first processing module is used for: The first information of the AI ​​model or algorithm is updated according to the first business requirement to obtain the second information; The second information includes at least one of the following: The second QoS information that needs to be updated includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows. The updated third QoS information includes at least one of the following: computation latency, model accuracy, data accuracy, computational accuracy, model inference latency, model training latency, model monitoring latency, data acquisition latency, data processing latency, communication reliability information, priority level, resource type, communication transmission rate, and communication latency; wherein, the priority level is used to determine the priority of resource scheduling for QoS flows. The first change amount is used to indicate the change amount between the third QoS information and the QoS information corresponding to the first information; First configuration information, which is used to indicate parameter configuration information corresponding to the third QoS information, wherein the parameter configuration information includes configuration information of at least one of the first parameter, the second parameter and the third parameter; A first action instruction is used to increase or decrease the parameter corresponding to the second QoS information, wherein the parameter corresponding to the second QoS information includes at least one of a first parameter, a second parameter, a third parameter, a fourth parameter, and a fifth parameter; Indicator of the reason for updating the strategy; The first strategy is an identifier, wherein the first strategy is a strategy corresponding to the first service requirement requested by the terminal before receiving the first message, or the first strategy is a strategy corresponding to the second service requirement requested by the terminal before receiving the second message. The identifier of the second policy, which is the policy corresponding to the third QoS information.

25. The apparatus according to claim 24, characterized in that, The first processing module is used for: A third message is sent to the second network element based on the first service accuracy, the third message being used to request an update to the first information; Obtain the fourth message sent by the second network element, wherein the fourth message is a response message to the third message and includes the second information; Based on the fourth message, obtain the second information; The third message includes at least one of the following: The identifier of the first strategy; Indicator of the reason for updating the strategy; The first QoS information; The second QoS information.

26. The apparatus according to claim 24 or 25, characterized in that, Also includes: The first sending module is used to send a fifth message to the third network element. The fifth message is used to request modification of the first sub-QoS information. The parameters in the first sub-QoS information are a subset of the parameters in the second QoS information. The first sub-QoS information includes model QoS information, data QoS information, computing power QoS information, and communication QoS information. The second receiving module is used to acquire the sixth message sent by the third network element, wherein the sixth message is a response message to the fifth message; The model QoS information includes at least one of the following: model accuracy, model inference latency, model training latency, and model monitoring latency; the data QoS information includes data accuracy, data acquisition latency, and data processing latency; the computing power QoS information includes computing latency or computing accuracy; and the communication QoS information includes at least one of the following: communication reliability information, priority level, resource type, communication transmission rate, and communication latency.

27. The apparatus according to claim 26, characterized in that, The fifth message includes at least one of the following: First sub-QoS information; The updated second sub-QoS information is determined based on the third QoS information; The second change amount is used to indicate the change amount between the second sub-QoS information and the sub-QoS information corresponding to the first information; The sub-QoS information corresponding to the first information is determined based on the QoS information corresponding to the first information; The second configuration information is used to indicate the parameter configuration information corresponding to the second sub-QoS information; The second action instruction is used to increase or decrease the parameter corresponding to the first sub-QoS information, and the parameter corresponding to the first sub-QoS information includes at least one of the first parameter, the second parameter, the third parameter, the fourth parameter, and the fifth parameter. Indicator of the reason for updating the strategy; The identifier of the first sub-policy, where the first sub-policy is the policy corresponding to the first sub-QoS information; The identifier of the second sub-policy, which is the policy corresponding to the second sub-QoS information.

28. The apparatus according to claim 26 or 27, characterized in that, In the case where the first sub-QoS information includes the model QoS information, the third network element includes AI-related functional nodes; Alternatively, if the first sub-QoS information includes the data QoS information, the third network element includes a data-related functional node; Alternatively, if the first sub-QoS information includes the computing power QoS information, the third network element includes a functional node related to computing power.

29. The apparatus according to any one of claims 24 to 28, characterized in that, Upon receiving the first message, the reason indication for the update strategy includes at least one of the following: change information of communication QoS, change information of computing quality QoS, and change information of data QoS.

30. The apparatus according to any one of claims 24 to 29, characterized in that, The first processing module is used for: When the communication QoS change notification indicates a decrease in transmission rate, and the communication reliability information indicates that the communication reliability remains unchanged, the adjustment to the first information includes: reducing the calculation latency; When the communication QoS change notification indicates a decrease in transmission rate and a decrease in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy; When the communication QoS change notification indicates a decrease in transmission rate and an increase in communication reliability, the adjustment of the first information includes at least one of the following: reducing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy; When the communication QoS change notification indicates that the transmission rate remains unchanged but the communication reliability is reduced, the adjustment of the first information includes at least one of the following: improving model accuracy; improving data accuracy; improving calculation accuracy; When the communication QoS change notification indicates that the transmission rate remains unchanged and the communication reliability is improved, the adjustment to the first information includes at least one of the following: reducing model accuracy; reducing data accuracy; reducing calculation accuracy; When the communication QoS change notification indicates an increase in transmission rate and the communication reliability remains unchanged, the adjustment to the first information includes: increasing the calculation delay; When the communication QoS change notification indicates an increase in transmission rate and an increase in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, reducing model accuracy, reducing data accuracy, and reducing computational accuracy; When the communication QoS change notification indicates an increase in transmission rate and a decrease in communication reliability, the adjustment to the first information includes at least one of the following: increasing computational latency, improving model accuracy, improving data accuracy, and improving computational accuracy.

31. The apparatus according to any one of claims 24 to 28, characterized in that, The first processing module is configured to perform at least one of the following: When the second message indicates that the terminal's computing power has decreased and the display accuracy has decreased, the adjustment to the first message includes at least one of the following: increasing the computing speed, decreasing the computing accuracy, decreasing the algorithm accuracy, decreasing the data accuracy, and increasing the communication speed; When the second message indicates that the terminal's computing power is reduced but the display accuracy remains unchanged, the adjustment to the first information includes at least one of the following: increasing the computing speed and increasing the communication speed; When the second message indicates that the terminal's computing power and display accuracy have increased, the adjustment to the first information includes at least one of the following: reducing the computing speed, increasing the computing accuracy, increasing the algorithm accuracy, increasing the data accuracy, and reducing the communication speed; When the computing power of the second message terminal is increased while the display accuracy remains unchanged, the adjustment of the first information includes at least one of the following: reducing the computing speed and reducing the communication speed; When the second message indicates that the terminal's computational demand per unit time increases, the adjustment to the first message includes at least one of the following: increasing the computational rate and increasing the communication rate; When the second message indicates that the terminal's computational requirements per unit time are reduced, the adjustment to the first message includes at least one of the following: reducing the computational rate and reducing the communication rate.

32. The apparatus according to any one of claims 17 to 31, characterized in that, The first information satisfies at least one of the following: When the AI ​​unit is provided by a network-side device, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the network-side device, the model structure of the AI ​​unit from the network-side device, and the degree of pruning of the model structure of the AI ​​unit from the network-side device. When the AI ​​unit is provided by the terminal, the model accuracy is adjusted by at least one of the following: the data type of the model parameters of the AI ​​unit from the terminal, the model structure of the AI ​​unit from the terminal, and the degree of pruning of the model structure of the AI ​​unit from the terminal. When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the terminal is represented by the number of bits, and the numerical value of the first data is represented by the number of bits; When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the network-side device, the data precision is adjusted by at least one of the following: the numerical value of the data collected by the network-side device is represented by the number of bits, and the numerical value of the first data is represented by the number of bits; When the input data of the AI ​​unit or the input data of the calculation algorithm are provided by the terminal and the network side device, the data accuracy is adjusted by at least one of the following: the value of the data collected by the network side device is represented by the number of bits, the value of the data collected by the terminal is represented by the number of bits, and the value of the first data is represented by the number of bits. When the service precision is set to AI-specific service precision, the calculation precision is adjusted by at least one of the following: the number of bits used to represent the numerical values ​​of the input data configured during calculation, the number of bits used to represent the numerical values ​​of the model parameters configured during calculation, the number of bits used to represent the numerical values ​​of the intermediate variables configured during calculation, and the number of bits used to represent the numerical values ​​of the calculation results configured during calculation. When the service precision is for non-AI services, the calculation precision is adjusted by at least one of the following: the number of bits in the numerical representation of the input data configured during calculation, the number of bits in the numerical representation of the intermediate variables configured during calculation, the number of bits in the numerical representation of the calculation result configured during calculation, and the calculation algorithm configured during calculation. The first data includes data to be transmitted that the network sends to the terminal.

33. A network-side device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the information adjustment method as described in any one of claims 1 to 16.

34. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the information adjustment method as described in any one of claims 1 to 16.

35. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the information adjustment method as described in any one of claims 1 to 16.