Communication method and apparatus, electronic device, and storage medium

Abstract

The application provides a communication method and device, electronic equipment and storage medium, and belongs to the technical field of communication. The method comprises the following steps: in response to receiving a calculation task request sent by an application layer, analyzing the calculation task request to obtain a first calculation task data packet and a first analysis parameter, wherein the first analysis parameter comprises a communication quality parameter set and a calculation quality parameter set corresponding to the calculation task request; adding an adaptation layer packet header to the first calculation task data packet according to the first analysis parameter to obtain a second calculation task data packet; mapping at least one second calculation task data packet to a corresponding quality of service data stream, and determining quality of service indication information of the quality of service data stream; and sending the quality of service data stream and the corresponding quality of service indication information to a service data adaptation protocol layer entity. The embodiment of the application can transmit calculation task data packets based on the calculation service quality requirement of the calculation task.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a communication method, communication device, electronic device, and computer-readable storage medium. Background Technology

[0002] In related technologies, the 5G (5th Generation) air interface user plane protocol stack, from top to bottom, includes the Service Data Adaptation Protocol (SDAP), Packet Data Convergence Protocol (PDCP), Radio Link Control (RLC), Medium Access Control (MAC), and Physical Layer (PHY). This protocol stack is mainly designed for traditional data transmission services (such as video services and web browsing services). Its Quality of Service (QoS) mapping mechanism only supports the mapping and bearing of communication QoS (such as bandwidth, latency, and packet loss rate), and does not consider the computational QoS requirements (such as computing power complexity, model accuracy, and inference execution latency) specific to computing tasks such as Artificial Intelligence (AI). The service effect for computing tasks needs to be improved. Summary of the Invention

[0003] To address this issue, the present invention provides a communication method, communication device, electronic device, and computer-readable storage medium to solve the problem in the prior art where the QoS mapping mechanism does not consider the specific QoS requirements of computing tasks, resulting in poor service performance for computing tasks.

[0004] To achieve the above objectives, a first aspect of the present invention provides a communication method applied to a computing task adaptation protocol layer entity, wherein the computing task adaptation protocol layer is located between the application layer and the service data adaptation protocol layer. The method includes: in response to receiving a computing task request sent by the application layer, parsing the computing task request to obtain a first computing task data packet and first parsing parameters, wherein the first parsing parameters include a communication quality parameter set and a computing quality parameter set corresponding to the computing task request; and adding an adaptation layer header to the first computing task data packet according to the first parsing parameters to obtain a second computing task data packet.

[0005] At least one of the second computing task data packets is mapped to a corresponding quality of service (QoS) data stream, and the QoS indication information of the QoS data stream is determined; the QoS data stream and the corresponding QoS indication information are sent to the service data adaptation protocol layer entity.

[0006] Furthermore, the first parsing parameter also includes model attribute information;

[0007] The step of adding an adaptation layer header to the first computing task data packet according to the first parsing parameters to obtain the second computing task data packet includes:

[0008] Based on the communication quality parameter set, determine the service quality level identifier and the communication service quality index parameter;

[0009] Based on the set of computational quality parameters and the model attribute information, determine the computational priority, computational service quality index parameters, and model version parameters;

[0010] The adaptation layer header is generated based on the service quality level identifier, the communication service quality index, the calculation priority, the calculation service quality index, and the model version parameter.

[0011] The adaptation layer header is added to the first computing task data packet to obtain the second computing task data packet;

[0012] The service quality level identifier is used to characterize the communication priority of the computing task request, the communication service quality index parameter is used to indicate the target communication service quality configuration template that matches the computing task request, and the computing service quality index parameter is used to indicate the target computing service quality configuration template that matches the computing task request.

[0013] Further, determining the service quality level identifier and communication service quality index parameters based on the communication quality parameter set includes:

[0014] Based on the first communication quality parameter value requirements corresponding to each service quality level, determine the target service quality level that matches the communication quality parameter set, and determine the identifier of the target service quality level as the service quality level identifier;

[0015] Based on the second communication quality parameter value requirements corresponding to each communication service quality configuration template, a target communication service quality configuration template matching the communication quality parameter set is determined, and the communication service quality index parameter is determined based on the identifier of the target communication service quality configuration template.

[0016] Further, determining the computation priority, computation service quality index parameter, and model version parameter based on the computation quality parameter set and the model attribute information includes:

[0017] Based on the preset quality priority mapping relationship, the calculation priority corresponding to the set of calculation quality parameters is determined;

[0018] Based on the computing quality parameter value requirements corresponding to each computing service quality configuration template, a target computing service quality configuration template matching the computing quality parameter set is determined, and the computing service quality index parameter is determined based on the identifier of the target computing service quality configuration template.

[0019] The model version parameters are determined based on the model attribute information.

[0020] Furthermore, the communication quality parameter set includes at least one of minimum data rate, maximum data rate, maximum end-to-end latency, and maximum jitter, and the computation quality parameter set includes at least one of computational complexity, maximum computational execution latency, model size, and model accuracy requirements.

[0021] Further, the step of mapping at least one of the second computing task data packets to a corresponding quality of service (QoS) data stream and determining the QoS indication information of the QoS data stream includes:

[0022] Map multiple second computing task data packets with the same set of communication quality parameters to the same quality of service data stream;

[0023] For any quality of service (QoS) data stream, determine the data radio bearer information and the corresponding set of QoS parameters used to carry the QoS data stream; obtain the QoS indication information of the QoS data stream based on the stream identifier of the QoS data stream, the data radio bearer information, and the set of QoS parameters.

[0024] The data wireless bearer information is used to indicate the computing node carrying the computing task request. The computing node includes end-side computing nodes, edge computing nodes, and cloud computing nodes. The service quality parameter set includes at least some quality parameters from the communication quality parameter set and the computing quality parameter set.

[0025] Furthermore, the method also includes:

[0026] In response to receiving a third computing task data packet sent by the service data adaptation protocol layer entity, the third computing task data packet is parsed to obtain a fourth computing task data packet and a second parsing parameter.

[0027] The fourth computing task data packet is sent to the corresponding target computing node according to the second parsing parameters, and the target computing node is used to execute the corresponding computing task according to the fourth computing task data packet.

[0028] Secondly, the present invention provides a communication device disposed in a computing task adaptation protocol layer entity, wherein the computing task adaptation protocol layer is located between the application layer and the service data adaptation protocol layer. The communication device includes: a parsing module, configured to, in response to receiving a computing task request sent by the application layer, parse the computing task request to obtain a first computing task data packet and a first parsing parameter, wherein the first parsing parameter includes a communication quality parameter set and a computing quality parameter set corresponding to the computing task request; an adding module, configured to, according to the first parsing parameter, add an adaptation layer header to the first computing task data packet to obtain a second computing task data packet; a mapping module, configured to, map at least one of the second computing task data packets to a corresponding service quality data stream, and determine service quality indication information of the service quality data stream; and a sending module, configured to, send the service quality data stream and the corresponding service quality indication information to the service data adaptation protocol layer entity.

[0029] Thirdly, the present invention provides an electronic device comprising: one or more processors;

[0030] A memory storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the aforementioned communication method; and one or more I / O interfaces connected between the processors and the memory, configured to enable information interaction between the processors and the memory.

[0031] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described communication method.

[0032] The present invention has the following advantages:

[0033] This embodiment adds a CTAP layer to the existing 5G air interface user plane protocol stack, realizing a two-dimensional mapping of communication QoS and computing QoS. It can filter and map computing task data packets with the same communication QoS into a single QoS flow. Computing tasks corresponding to different computing QoS can be distinguished by fields in the adapter layer header. Furthermore, it sets QoS indication information corresponding to the QoS flow, realizing differentiated air interface bearing of computing tasks. Based on this, after the CTAP entity on the receiving end receives the data packet, it can restore the QoS flow association information by parsing the header and QoS indication information, thereby achieving accurate distribution of computing tasks.

[0034] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0035] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. The above and other features and advantages will become more apparent to those skilled in the art from the detailed description of exemplary embodiments with reference to the accompanying drawings, in which:

[0036] Figure 1 A flowchart illustrating a communication method provided in an embodiment of this disclosure.

[0037] Figure 2 This is a schematic diagram illustrating the processing steps of a communication method provided in an embodiment of this disclosure.

[0038] Figure 3 This is a block diagram of a communication device provided in an embodiment of the present disclosure.

[0039] Figure 4 This is a block diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0040] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0041] As used in this invention, the term "and / or" includes any and all combinations of one or more of the associated enumerated entries.

[0042] The terminology used in this invention is for describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms "a" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0043] When the terms “comprising” and / or “made of” are used in this invention, the presence of the said feature, integral, step, operation, element and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or groups thereof is not excluded.

[0044] Unless otherwise specified, all terms used in this invention (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in common dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined by the invention.

[0045] In existing technologies, AI inference requests enter the air interface protocol stack as ordinary application data. The network side cannot identify whether the data packet carries "AI computing tasks" or "ordinary business data." This may result in the inability to allocate differentiated air interface resources based on the complexity of the AI ​​model, and the inability to select the corresponding execution path (edge, local, or cloud) based on inference latency requirements. In the highly dynamic scenarios of connected vehicles, communication and computing resources are scheduled in a fragmented manner, making it difficult to meet stringent latency constraints. Specifically, this manifests in the following aspects:

[0046] (1) Fragmentation of general computing protocols - the network "cannot see" AI computing tasks: The existing 5G air interface protocol stack lacks protocol layer design for AI computing tasks. AI inference requests need to go through cross-layer adaptation from the application layer to the SDAP layer. The network side only forwards data packets according to communication QoS and cannot perceive the attributes of the computing tasks corresponding to the data packets (such as model version, computing priority, and accuracy requirements), resulting in the disconnect between air interface resource scheduling and AI inference needs.

[0047] (2) Convergence of vehicle networking scenarios - lack of fine-grained task differentiation: the existing single vehicle modeling method treats all vehicle networking data streams equally; for example, remote driving control commands and batch fault detection logs may be mapped to the same data radio bearer (DRB), which cannot make fine-grained differentiation and differentiated bearing for AI inference task types, making it difficult to meet the differentiated real-time requirements of computing tasks such as remote driving and image recognition.

[0048] (3) Lack of computation task distribution - the receiving end is "unclear" about which computing task to distribute to: the existing protocol stack only identifies the communication data flow (flow) through the QoS Flow Identifier (QFI) of the SDAP layer at the receiving end (such as base station, edge node), and cannot further identify which computing node (such as local computing node, edge computing node or cloud computing node) should distribute different computing tasks in the communication data flow. As a result, the AI ​​inference data needs to be parsed twice in the application layer message before a decision can be made, which increases the end-to-end processing latency.

[0049] In summary, the existing technology has at least the following technical problems:

[0050] (1) The existing 5G air interface user plane protocol stack only supports communication QoS mapping and lacks protocol layer design for AI computing tasks, resulting in the separation of communication protocol and computing protocol, and making it impossible to achieve integrated collaborative management and control of QoS for computing task execution and connection transmission.

[0051] (2) Multiple AI inference tasks in the Internet of Vehicles (remote driving, image recognition, fault detection) share a single air interface for transmission. The existing protocol stack cannot carry out differentiated transmission based on the task calculation characteristics, resulting in high-priority, low-latency tasks being squeezed out by low-priority, high-volume tasks.

[0052] (3) The existing protocol stack receiver cannot directly distribute AI computing tasks to computing resource nodes based on the air interface packet header. It needs to rely on secondary parsing at the application layer, which increases end-to-end latency.

[0053] Based on this, embodiments of the present disclosure provide a communication method, a communication device, an electronic device, and a computer-readable storage medium.

[0054] In a first aspect, embodiments of the present invention provide a communication method applicable to a Compute Task Adaptation Protocol (CTAP) entity, which is located between the application layer and the service data adaptation protocol layer. The application layer can be used to support AI inference services, while CTAP is a newly added protocol layer in the existing 5G air interface user plane protocol stack, used to implement QoS mapping for computing tasks, quality of service data stream mapping, and mapping of computing resources and connection bearers.

[0055] Figure 1 A flowchart illustrating a communication method provided in an embodiment of this disclosure. (Refer to...) Figure 1 The communication method may include the following steps:

[0056] In step S11, in response to receiving a computing task request sent by the application layer, the computing task request is parsed to obtain a first computing task data packet and a first parsing parameter.

[0057] The first set of parsed parameters includes the communication quality parameter set and the computation quality parameter set corresponding to the computation task request.

[0058] In step S12, an adaptation layer header is added to the first computation task data packet according to the first parsing parameters to obtain the second computation task data packet.

[0059] In step S13, at least one second computing task data packet is mapped to the corresponding quality of service data stream, and the quality of service indication information of the quality of service data stream is determined.

[0060] In step S14, the quality of service data stream and the corresponding quality of service indication information are sent to the service data adaptation protocol layer entity.

[0061] Therefore, this embodiment adds a CTAP layer to the existing 5G air interface user plane protocol stack, realizing a two-dimensional mapping of communication QoS and computing QoS. It can filter and map computing task data packets with the same communication QoS into a single QoS data stream. Computing tasks corresponding to different computing QoS can be distinguished by fields in the adapter layer header. Furthermore, QoS indication information corresponding to the QoS data stream is set to achieve differentiated air interface bearing of computing tasks. Based on this, after the CTAP entity on the receiving end receives the data packet, it can restore the QoS data stream association information by parsing the header and QoS indication information, thereby achieving accurate distribution of computing tasks.

[0062] Furthermore, as a newly added independent protocol layer, the CTAP layer can interact with the SDAP layer through the standard SDAP service access point (SAP) without modifying the existing PDCP, RLC, MAC and PHY layer processing procedures, and has high compatibility with existing protocols.

[0063] The communication method of the present disclosure embodiment will be described in detail below.

[0064] In some optional embodiments, computing task requests include requests corresponding to AI inference tasks, requests corresponding to various computing tasks in the vehicle-to-everything (V2X) scenario, etc.

[0065] In some optional embodiments, a computing task request is a request message initiated by the requesting end to trigger the execution of a target computing task. The message may carry data packets required for task execution as well as various parameter information related to task execution.

[0066] For example, computing task requests may include remote driving reasoning task requests, image recognition task requests, sensor data reporting requests, etc., and this disclosure does not limit these.

[0067] In some optional embodiments, the CTAP entity is a logical functional entity located at the computing task adaptation protocol layer, possessing capabilities such as data transmission and reception and message parsing. It can be deployed in network devices, computing nodes, or gateway devices to implement data processing and interaction functions of the CTAP layer.

[0068] In some optional embodiments, after receiving a computing task request, the CTAP entity can parse the computing task request to obtain a first computing task data packet corresponding to the computing task and corresponding first parsing parameters. Parsing refers to the process by which the CTAP entity disassembles and extracts fields from the message structure of the computing task request according to a preset message format specification. The first parsing parameters include a communication quality parameter set and a computing quality parameter set corresponding to the computing task request. The communication quality parameter set is a set of parameters used to characterize the communication status of the computing task request transmission link and can reflect the link performance indicators during data transmission. The computing quality parameter set is a set of parameters used to characterize the performance requirements required for the execution of the computing task and can reflect the execution standards of the computing task in terms of computing power, latency, accuracy, and other dimensions.

[0069] In related technologies, computing task requests typically only carry a set of communication quality parameters in addition to data packets, and do not include computing service quality parameters related to task scheduling and computation execution. Therefore, downstream transmission nodes cannot allocate differentiated air interface resources to computing task requests based on computing service quality parameters, nor can they allocate them to computing nodes that match the computing service requirements for execution. Based on this, in this embodiment, the computing task request includes both a set of communication quality parameters and a set of computing quality parameters, satisfying both communication quality and computing quality requirements.

[0070] Therefore, the CTAP entity can perform parsing operations on the received computing task requests. Based on extracting the data packets corresponding to the computing tasks, it can simultaneously obtain parsing parameters containing two dimensions: communication quality and computing quality. This completes the multi-dimensional extraction of the information carried by the computing task requests, providing parameter support for the subsequent scheduling, distribution, and execution of computing tasks.

[0071] In some optional embodiments, the first parsing parameter further includes model attribute information; correspondingly, according to the first parsing parameter, an adaptation layer header is added to the first computing task data packet to obtain a second computing task data packet, including: determining a service quality level identifier and a communication service quality index parameter according to a communication quality parameter set; determining a computing priority, a computing service quality index parameter, and a model version parameter according to a computing quality parameter set and model attribute information; generating an adaptation layer header according to the service quality level identifier, the communication service quality index, the computing priority, the computing service quality index, and the model version parameter; adding the adaptation layer header to the first computing task data packet to obtain the second computing task data packet; wherein, the service quality level identifier is used to characterize the communication priority of the computing task request, the communication service quality index parameter is used to indicate the target communication service quality configuration template matching the computing task request, and the computing service quality index parameter is used to indicate the target computing service quality configuration template matching the computing task request.

[0072] The model attribute information includes attribute descriptions related to the computing model called by the computing task, which may include the model's name, version, type, runtime dependencies, and other related content. The adaptation layer header refers to the structured message header attached to the outer layer of the computing task data packet, used to carry multi-dimensional scheduling and execution parameters, and can be identified and read by the corresponding functional entities on the transmission path. The second computing task data packet is the data packet formed after adding the adaptation layer header, which contains the original computing task data and structured multi-dimensional parameter information.

[0073] Furthermore, the quality of service level identifier is an identifier field used to characterize the communication priority of computing task requests, which can be identified by network transmission nodes to execute the corresponding transmission scheduling strategy; computing priority is a parameter that can be used to characterize the execution order of computing tasks and can reflect the execution priority of computing tasks.

[0074] For example, the quality of service (QoS) level identifier can include 5, 2, and 1. 5 corresponds to Ultra-Reliable Low Latency Communication (URLLC) level, which is mainly for service scenarios with low latency and high reliability requirements. 2 corresponds to Enhanced Mobile Broadband (eMBB) level, which is mainly for service scenarios with high bandwidth and high speed requirements. 1 corresponds to Massive Machine Type Communication (mMTC) level, which is mainly for service scenarios with massive terminal access and low power consumption requirements.

[0075] Furthermore, corresponding communication service quality configuration templates can be pre-configured for commonly used and important communication service quality, and corresponding identifiers can be configured for each communication service quality configuration template to accurately and quickly distinguish between them; similarly, corresponding computing service quality configuration templates can be pre-configured for commonly used and important computing service quality, and corresponding identifiers can be configured for each computing service quality configuration template to accurately and quickly distinguish between them.

[0076] Based on this, the communication service quality index parameter is used to index the corresponding communication service quality configuration template. Through this parameter, the matching communication service quality configuration template (i.e., the target communication service quality configuration template) can be quickly located, and the corresponding communication service quality configuration rules can be determined. The computing service quality index parameter is used to index the corresponding computing service quality configuration template. Through this parameter, the matching computing service quality configuration template (i.e., the target computing service quality configuration template) can be quickly located, and the corresponding computing service quality configuration rules can be determined.

[0077] Through the above processing, the identifier and index parameters of the communication dimension can be generated based on the communication quality parameter set, and the priority, index parameters and model version parameters of the computing dimension can be generated based on the computing quality parameter set and model attribute information. Then, the adaptation layer packet header is generated from the above parameters and encapsulated into the first computing task data packet to obtain the second computing task data packet, realizing the structured integration of the three types of information of transmission, scheduling and model in the header of the data packet.

[0078] Therefore, by adding an adaptation layer header to the outer layer of the first computation task data packet, parameters of the three dimensions of communication, computation, and model can be uniformly and structurally encapsulated, enabling differentiated resource allocation for computation tasks and improving the processing efficiency and effectiveness of task transmission and scheduling. Using parameter formats where index parameters correspond to configuration templates reduces the field length of the adaptation layer header, lowering the transmission overhead caused by the header, while ensuring unified identification of service quality configuration rules by each node, improving the consistency of parameter parsing and the accuracy of scheduling execution. Including model version parameters in the adaptation layer header allows computation nodes to quickly match the corresponding model resources after receiving the data packet, shortening the preparation latency for task loading and improving the overall execution efficiency of computation tasks.

[0079] In some optional embodiments, the communication quality parameter set includes at least one of minimum data rate, maximum data rate, maximum end-to-end latency, and maximum jitter, and the computation quality parameter set includes at least one of computational complexity, maximum computational execution latency, model size, and model accuracy requirements.

[0080] For example, the communication QoS parameter set Q_c = {R_min, R_max, D_max, J_max} is used to calculate the QoS parameter set Q_comp = {C_req, T_exec, M_size, A_model}, where R_min / R_max are the minimum / maximum data rates, D_max is the maximum end-to-end latency, J_max is the maximum jitter, C_req is the required computational complexity (FLOPs), T_exec is the maximum computational execution latency, M_size is the AI ​​model size, and A_model is the model accuracy requirement.

[0081] It should be noted that the above examples of communication quality parameter sets and computational quality parameter sets are merely illustrative and this disclosure does not impose any limitations on them.

[0082] In some optional embodiments, determining the service quality level identifier and communication service quality index parameter based on the communication quality parameter set includes: determining a target service quality level matching the communication quality parameter set based on the first communication quality parameter value requirements corresponding to each service quality level, and determining the identifier of the target service quality level as the service quality level identifier; determining a target communication service quality configuration template matching the communication quality parameter set based on the second communication quality parameter value requirements corresponding to each communication service quality configuration template, and determining the communication service quality index parameter based on the identifier of the target communication service quality configuration template.

[0083] For example, a corresponding first communication quality parameter value requirement is preset for each service quality level. After determining the communication quality parameter set, the service quality level that matches it can be determined according to the parameters therein, which is taken as the target service quality level, and the identifier of the target service quality level is taken as the service quality level identifier.

[0084] For example, corresponding communication quality parameter value requirements are pre-configured for each communication service quality configuration template. Based on this, after determining the communication quality parameter set, the matching communication service quality configuration template can be determined according to the parameters therein, and it can be used as the target communication service quality configuration template. The identifier of the target communication service quality configuration template is used as the communication service quality index parameter.

[0085] The first and second communication quality parameters can be set using a single-parameter threshold method. Each service quality level corresponds to a range of values ​​for a basic communication quality parameter. For example, transmission delay can be used as the criterion, with different service quality levels corresponding to different delay intervals to determine the target service quality level. This setting method has a simple judgment logic, fast matching speed, and is suitable for application scenarios with a single parameter dimension and few level division dimensions.

[0086] In addition, the requirements for the values ​​of the first and second communication quality parameters can also adopt a multi-parameter joint threshold setting method. For example, each service quality level corresponds to a combined range of values ​​for multiple communication quality parameters, such as combining the thresholds of transmission latency, packet loss rate, and available bandwidth for joint judgment. This setting method provides a more comprehensive dimension for level determination, can more accurately reflect the overall communication quality of the link, and improves the rationality of level classification and template configuration.

[0087] Furthermore, taking the determination of the target quality of service (QoS) level as an example, a step-by-step comparison and matching method can be adopted. For instance, according to the QoS levels from high to low, the set of communication quality parameters is compared with the first communication quality parameter value requirement of the corresponding level, and the QoS level that first meets the value requirement is determined as the target QoS level. The method for determining the target communication QoS configuration template is similar and will not be described in detail here. This matching method can prioritize matching higher-level communication QoS, which helps to ensure the task transmission performance when the link condition is good.

[0088] Alternatively, the target service quality level can be determined using an optimal fit matching method. For example, the degree of fit between the communication quality parameter set and the required values ​​for each service quality level can be evaluated, and the level with the highest degree of fit can be determined as the target service quality level. The method for determining the target communication service quality configuration template is similar and will not be described in detail here. This matching method is suitable for scenarios where the parameters are within the critical range of the level, improving the accuracy of level determination and template matching, and reducing frequent level switching and frequent changes to the communication service quality configuration template caused by fluctuations in the critical value.

[0089] In summary, this embodiment of the present disclosure, by setting two sets of communication quality parameter value requirements, can decouple service quality level determination from communication service quality configuration template matching, supporting adaptation to multiple different configuration templates under the same service quality level, further improving the flexibility and scenario adaptability of transmission scheduling strategies. Furthermore, using standardized parameter value requirements as the matching basis provides a clear execution standard for the matching process between level determination and communication service quality configuration templates, improving the consistency and predictability of processing results and ensuring the stable execution of subsequent transmission scheduling strategies. In addition, the matching process is based on preset value requirements, eliminating the need for complex real-time calculations, reducing the processing overhead of CTAP entities, and improving the processing efficiency of parameter determination. Moreover, since both sets of value requirements can be flexibly configured and adjusted according to business needs without altering the overall matching process architecture, it possesses good maintainability and scalability.

[0090] In some optional embodiments, determining the computation priority, computation service quality index parameter, and model version parameter based on the computation quality parameter set and the model attribute information includes: determining the computation priority corresponding to the computation quality parameter set according to a preset quality priority mapping relationship; determining a target computation service quality configuration template matching the computation quality parameter set according to the computation quality parameter value requirements corresponding to each computation service quality configuration template; determining the computation service quality index parameter based on the identifier of the target computation service quality configuration template; and determining the model version parameter based on the model attribute information.

[0091] For example, the correspondence between different computing service quality parameters and computing priority parameter value requirements can be pre-set. After determining the set of computing quality parameters, the matching computing priority can be determined according to the actual parameter values ​​and parameter value requirements.

[0092] For example, corresponding computing quality parameter value requirements are pre-configured for each computing service quality configuration template. Based on this, after determining the computing quality parameter set, the matching computing service quality configuration template can be determined according to the parameters therein, and it can be used as the target computing service quality configuration template. The identifier of the target computing service quality configuration template is used as the computing service quality index parameter.

[0093] For example, the model attribute information may include template version information, based on which the corresponding model version parameters can be determined.

[0094] The parameter values ​​for computation priority and computation quality can be set using a single-parameter threshold method. For example, each computation priority corresponds to a range of values ​​for a basic computation quality parameter, such as using the maximum computation execution latency parameter as the criterion. Different computation priorities correspond to different latency intervals, thereby determining the matching computation priority. This setting method has a simple judgment logic, fast matching processing speed, and is suitable for application scenarios with a single parameter dimension and few level division dimensions.

[0095] Furthermore, the parameter values ​​for computation priority and computation quality can also be set using a multi-parameter joint threshold method. For example, each computation priority can correspond to a combined range of values ​​for multiple computation quality parameters, such as combining the thresholds of maximum computation execution latency and computation complexity for joint judgment. This setting method provides a more comprehensive dimension for level determination, can more accurately reflect the overall computation service quality of the link, and improves the rationality of level division and template configuration.

[0096] Furthermore, taking the determination of computational priority as an example, a step-by-step comparison and matching method can be adopted. For instance, the set of computational quality parameters is compared with the corresponding level of computational quality parameter value requirements in descending order of computational priority, and the first computational priority that meets the value requirements is determined as the matched computational priority. The method for determining the target computational service quality configuration template is similar and will not be described in detail here. This matching method can prioritize matching higher levels of computational service quality, which helps to ensure the configuration requirements of task execution resources.

[0097] Alternatively, computational priorities can be determined using an optimal fit matching method. For example, the degree of fit between the computational quality parameter set and the parameter value requirements of each computational priority can be evaluated, and the priority with the highest degree of fit can be determined as the matched computational priority. The method for determining the target computational service quality configuration template is similar and will not be described in detail here. This matching method is suitable for scenarios where parameters are in the critical range of the level, which can improve the accuracy of priority determination and template matching, and reduce frequent priority switching and frequent changes to the computational service quality configuration template caused by fluctuations in the critical value.

[0098] In summary, this embodiment of the present disclosure, by separately setting requirements for the values ​​of computation priority parameters and computation quality parameters, decouples the determination of computation priority from the matching of computation service quality configuration templates. It supports adaptation to multiple different configuration templates under the same computation priority, further enhancing the flexibility and scenario adaptability of computation service strategies. Furthermore, using standardized parameter value requirements as the matching basis provides a clear execution standard for the matching process between priority determination and computation service quality configuration templates, improving the consistency and predictability of processing results and ensuring the stability of subsequent computation service quality. In addition, the matching process is executed based on preset value requirements, eliminating the need for complex real-time calculations, reducing the processing overhead of CTAP entities, and improving the efficiency of parameter determination. Moreover, since both sets of value requirements can be flexibly configured and adjusted according to business needs without altering the overall matching process architecture, it possesses good maintainability and scalability.

[0099] In some optional embodiments, after obtaining the second computing task data packet, the second computing task data packet can be mapped to the corresponding quality of service data stream, the quality of service indication information of the quality of service data stream can be determined, and then the quality of service data stream and the corresponding quality of service indication information can be sent to the service data adaptation protocol layer entity to perform further data transmission.

[0100] In some optional embodiments, mapping at least one second computing task data packet to a corresponding quality of service (QoS) data stream and determining QoS indication information for the QoS data stream includes: mapping multiple second computing task data packets with the same set of communication quality parameters to the same QoS data stream; for any QoS data stream, determining data radio bearer information and a corresponding set of QoS parameters used to carry the QoS data stream; obtaining QoS indication information for the QoS data stream based on the stream identifier, data radio bearer information, and QoS parameter set; wherein the data radio bearer information is used to indicate the computing node carrying the computing task request, the computing node includes end-side computing nodes, edge computing nodes, and cloud computing nodes, and the QoS parameter set includes at least some quality parameters from the communication quality parameter set and the computing quality parameter set.

[0101] Therefore, it can be seen that data packets from second computing tasks with the same communication quality requirements can be mapped to a single Quality of Service (QoS) data stream for unified transmission. Within this QoS data stream, different computing tasks can be identified through certain fields. The QoS indication information can include the stream identifier of the QoS data stream, data radio bearer information, and a set of QoS parameters, thus clearly defining the QoS data stream, the corresponding data radio bearer, and the required QoS parameters.

[0102] For example, a Quality of Service (QoS) data stream is a logical data stream formed by aggregating computing task data packets with the same communication QoS requirements, and can serve as the basic unit for transmission scheduling and resource allocation. QoS indication information is comprehensive indication information that can be used to characterize the identity attributes, bearer method, and QoS requirements of the QoS data stream, and can be identified and read by network nodes and computing nodes. Data wireless bearer information can indicate the executing entity carrying the computing task. A computing node is a functional node with the capability to execute computing tasks, and can be deployed at different network layers, including end-side computing nodes, edge computing nodes, and cloud computing nodes. End-side computing nodes are computing nodes deployed on the terminal device side, which can process lightweight computing tasks initiated by the terminal nearby. Edge computing nodes are computing nodes deployed at the network edge, which can provide low-latency computing services close to the terminal. Cloud computing nodes are computing nodes deployed in cloud data centers, which can provide large-scale, high-performance computing services.

[0103] Furthermore, when all parameters in the communication quality parameter set of multiple second computing task data packets have the same or similar values, they can be mapped to the same quality of service data stream. Based on this, the communication attributes of all data packets in the same data stream can be kept highly consistent, which facilitates the network side to execute a unified transmission scheduling strategy and improves the compatibility between the scheduling strategy and the data packet attributes.

[0104] In summary, by aggregating second computing task data packets with consistent communication quality attributes into a quality of service (QoS) data stream, the basic unit of transmission scheduling can be elevated from a single data packet to the data stream level. This reduces the frequency of scheduling processing, lowers the overhead of flow management and resource allocation for network nodes, and improves the processing efficiency of transmission scheduling. The QoS indication information integrates three types of content: flow identifier, bearer node information, and quality parameters. Network nodes and computing nodes can obtain the identity, bearer target, and quality requirements of the data stream at once by reading this indication information, without the need for deep parsing of the data packet payload content, thus improving the processing efficiency of cross-node task distribution and resource adaptation. The wireless data bearer information can distinguish between three types of bearer entities: end-side computing nodes, edge computing nodes, and cloud computing nodes. This supports matching computing tasks with different levels of computing resources according to quality requirements, adapting to collaborative scheduling scenarios of multi-level heterogeneous computing architectures, and improving the flexibility of task bearing. The QoS parameter set can be flexibly selected from the communication quality parameter set and the computing quality parameter set. Therefore, the parameter dimensions of the indication information can be adjusted according to scenario requirements, achieving a balance between information completeness and transmission overhead, and adapting to diverse computing business scenarios.

[0105] For example, the CTAP entity can perform filtering mapping to filter and map second computation task data packets with the same communication QoS parameter set Q_c into a Quality of Service (QoS) data flow. The corresponding mapping rule is: Flow_k ={pkt_i | Q_c(pkt_i) = Q_c(k), ∀i}, where Flow_k represents the k-th QoS data flow, k is the sequence index or identifier of the data flow, pkt_i represents the i-th second computation task data packet, Q_c(pkt_i) represents the communication QoS parameter set carried by the i-th second computation task data packet pkt_i, and Q_c(k) is the k-th communication QoS configuration template.

[0106] Furthermore, for computation tasks with different computational QoS requirements (i.e., different sets of computational quality parameters Q_comp) within the same QoS Flow, the CTAP entity can distinguish them through the Comp_QoS_Idx subfield in the adaptation layer header and send the QoS indication information to the lower-level SDAP entity: I_QoS = {Flow_ID, DRB_ID, QoS_Profile}, where QoS_Profile is a set of service quality parameters, which may contain some or all of the parameters in Q_c and Q_comp.

[0107] For example, the CTAP entity receives an AI computing task request from the application layer, parses it to obtain a first computing task data packet and a first parsing parameter, the first parsing parameter including:

[0108] The communication QoS parameter set Q_c = {R_min, R_max, D_max, J_max};

[0109] Calculate the QoS parameter set Q_comp = {C_req, T_exec, M_size, A_model};

[0110] Model attribute information: Model version parameter;

[0111] Where: R_min / R_max are the minimum / maximum data rates, D_max is the maximum end-to-end latency, J_max is the maximum jitter, C_req is the required computational complexity (FLOPs), T_exec is the maximum computational execution latency, M_size is the AI ​​model size, and A_model is the model accuracy requirement.

[0112] Based on this, the CTAP entity can generate a CTAP header for the first computation task data packet according to Q_c and Q_comp, filling it with fields such as Task ID, Quality of Service Class (QoS) Class, Comp Priority (Comp_Priority), Communication Quality of Service Index (CRM_QoS_Idx), Computation Quality of Service Index (Comp_QoS_Idx), and Model Version (Model_Version). Table 1 shows a field diagram of one type of CTAP header.

[0113] Table 1. Field diagram of CTAP header

[0114]

[0115] The CTAP entity performs filtering mapping, filtering and mapping second computation task packets with the same communication QoS parameter set Q_c into a QoS Flow, with corresponding service quality indication information I_QoS = {Flow_ID, DRB_ID, QoS_Profile}. The CTAP entity sends the QoS Flow and its I_QoS to SDAP for further transmission.

[0116] In some alternative embodiments, the CTAP entity can perform the aforementioned data processing at the sending end as well as at the receiving end.

[0117] In some optional embodiments, the communication method may further include: in response to receiving a third computing task data packet sent by a service data adaptation protocol layer entity, parsing the third computing task data packet to obtain a fourth computing task data packet and a second parsing parameter; and sending the fourth computing task data packet to a corresponding target computing node according to the second parsing parameter, wherein the target computing node is used to execute a corresponding computing task according to the fourth computing task data packet.

[0118] For example, the second parsing parameters may include fields from the service quality indication information to restore the association of the service quality data stream. Based on this, the adapter layer packet header is parsed to determine the computing service quality index and model version parameters, so that the fourth computing task data packet can be accurately and directly distributed to the corresponding computing node.

[0119] Furthermore, on the receiving side of the computing task transmission, the relevant technologies typically require the application layer to parse the received data packets and determine the corresponding computing node. In this embodiment of the disclosure, the receiving side can distribute the fourth computing task data packet to the accurate computing node through parsing.

[0120] Therefore, the embodiments disclosed herein support task distribution to computing nodes at different levels, such as the end side, edge side, and cloud side, which can adapt to the access requirements of multi-level heterogeneous computing architectures, improve the flexibility of task scheduling on the receiving side and the ability to adapt to different scenarios; the CTAP entity has both sending end encapsulation and receiving end parsing processing capabilities, which can form a symmetrical protocol processing architecture for sending and receiving, which facilitates the unified design, deployment and maintenance of the protocol system and reduces the overall implementation and operation and maintenance costs of the system.

[0121] For example, the SDAP entity can map the QoS Flow to the corresponding Data Radio Bearer (DRB) based on I_QoS, add an SDAP header index (containing a QFI field), and transmit it sequentially through the PDCP, RLC, MAC, and PHY layers. At the network-side receiver, the CTAP header is parsed through a reverse process: the QFI field in the SDAP header is parsed to restore the Flow association; the Comp_QoS_Idx and Model_Version fields in the CTAP header are further parsed; and the data packets are directly distributed to the corresponding computing resource nodes (such as edge NPUs, edge GPUs, and cloud TPUs) to execute AI inference, thereby achieving end-to-end mapping from computing data to computing connection bearers.

[0122] In summary, this application provides corresponding solutions to the aforementioned technical problems from the following perspectives:

[0123] (1) On the basis of the existing 5G air interface user plane protocol stack, the CTAP layer is introduced to complete the QoS mapping and update of AI computing tasks, realize the joint mapping of AI computing execution QoS and AI connection QoS to computing resources and connection bearers, and meet the connection and distributed computing fusion control requirements of vehicle network AI inference.

[0124] (2) By using the dual-dimensional QoS indexing mechanism and QoS Flow filtering mapping rules of the CTAP layer, AI inference data packets with the same communication QoS are grouped into a quality of service data stream, and different computing tasks within the same quality of service data stream are distinguished by calculating the QoS index subfield, so as to realize differentiated air interface bearing of AI inference tasks in the Internet of Vehicles.

[0125] (3) By using fields such as Comp_QoS_Idx and Model_Version in the CTAP packet header (i.e., the adaptation layer packet header), the CTAP entity on the receiving end can directly identify the computing task type and the required model version after the air interface is decapsulated, and complete the accurate distribution of data packets to computing resource nodes, supporting the edge-cloud collaborative AI inference of the Internet of Vehicles.

[0126] Figure 2 This is a schematic diagram illustrating the processing steps of a communication method provided in an embodiment of this disclosure. (Refer to...) Figure 2 The sending end corresponds to the application scenario layer, application layer, CTAP layer, SDAP layer, PDCP layer, RLC layer, MAC layer, and PHY layer, while the receiving end corresponds to the computing layer, CTAP layer, SDAP layer, PDCP layer, RLC layer, MAC layer, and PHY layer. The application scenario layer can correspond to vehicle networking, supporting remote driving control, real-time image recognition, fault detection, and safety warnings, with latency requirements of 10ms, 5ms, and 500ms-1s, respectively. The application layer corresponds to AI inference services. At the sending end, the application layer can encapsulate AI inference requests and several QoS parameters, while at the receiving end, the application layer can execute inference and return inference results. The corresponding computing nodes can include edge computing nodes, cloud computing nodes, and end-user computing nodes.

[0127] At the sending end, the application layer generates a computing task request and sends it to the CTAP layer. After parsing, adding packet headers, data flow mapping, and service quality indicators, the service quality data stream and its service quality indicator information are sent to the SDAP layer. This is then processed sequentially by the PDCP layer, RLC layer, MAC layer, and PHY layer before being transmitted wirelessly to the receiving end's PHY layer. After further transmission through the MAC layer, RLC layer, PDCP layer, and SDAP layer, it is sent to the CTAP layer. The CTAP entity parses the QFI field in the SDAP packet header, reconstructs the flow association, and further parses fields such as Comp_QoS_Idx and Model_Version in the CTAP packet header. The data packet is then directly distributed to the corresponding computing resource nodes (such as edge computing nodes, device computing nodes, and cloud computing nodes) to perform AI inference and obtain the inference result. After obtaining the inference result, it can be sent back to the request initiator in the reverse direction.

[0128] In some optional embodiments, federated learning can be used to assist CTAP model version management. For example, based on a federated learning mechanism, models can be trained locally on each vehicle side, with only parameters uploaded to the edge base station for aggregation. The Model_Version field in the CTAP layer can be expanded to a federated model version identifier. The base station synchronizes the global model version number through CTAP-SDAP interaction, achieving air interface consistency management of distributed AI models in the vehicle-to-everything (V2X) network.

[0129] In some optional embodiments, the CTAP layer supports dynamic model version degradation. For example, when preset switching conditions are met, the CTAP layer can dynamically modify the Comp_QoS_Idx field to instruct the receiver to switch model versions. For instance, when the quality of the vehicular network channel deteriorates (e.g., in NLOS scenarios) or edge computing resources are scarce, the CTAP layer can dynamically modify the Comp_QoS_Idx field to instruct the receiver to switch to a lightweight AI model version (e.g., from ResNet-50 to MobileNet-V2), trading accuracy for latency guarantees, thus avoiding the need to rebuild the air interface bearer.

[0130] In some optional embodiments, the CTAP layer can extend the zero-knowledge authentication field to support trusted vehicle-to-everything (V2X) access. For example, a privacy protection identifier can be extended to the Reserved field in the CTAP header, combined with a zero-knowledge authentication algorithm based on secondary residues, to achieve anonymized transmission of V2X AI inference task data packets, ensuring that "the prover knows the secret but does not reveal it," thus supporting trusted V2X access.

[0131] The communication method of this disclosure embodiment is illustrated below using several application scenarios as examples.

[0132] Table 2. Application Scenarios

[0133]

[0134] For remote driving control scenarios, the corresponding scenario description is as follows: The vehicle is traveling at a speed of 120 km / h on the highway. It needs to upload the perception data of the vehicle camera and radar to the edge base station in real time through the V2I link, generate control commands through the edge AI inference service and send them back to the vehicle. The end-to-end latency is required to be ≤10ms.

[0135] Accordingly, the processing flow of CTAP entities includes task parsing, CTAP header generation, Flow mapping, SDAP carrying, etc., as detailed below:

[0136] (1) Task parsing: The vehicle AI application generates a remote driving inference task request and sends it to the CTAP layer entity. The CTAP layer entity obtains the first computation task data packet and the first parsing parameters through request parsing. The first parsing parameters include Q_c={R_min=1Mbps,R_max=5Mbps,D_max=10ms,J_max=1ms}, Q_comp={C_req=5GFLOPs,T_exec=3ms,M_size=50MB,A_model=99.9%}, task identifier 0x00A1 (i.e., the unique identifier of the remote driving inference task), model version 0x03 (i.e., the corresponding safe driving model v3.2), etc.

[0137] (2) CTAP header generation: The CTAP layer entity generates the CTAP header according to the first parsing parameter mentioned above, which may include the following fields: Task_ID=0x00A1, QoS_Class=5 (URLLC extension), Comp_Priority=7 (highest calculation priority), Comm_QoS_Idx=0x01 (10ms low latency communication template), Comp_QoS_Idx=0x10 (corresponding to lightweight control model), Model_Version=0x03 (v3.2 safe driving model).

[0138] (3) Flow mapping: The Q_c of the entity recognition of the remote driving inference task in the CTAP layer is different from that of other vehicle tasks (such as entertainment video Q_c). Therefore, it is mapped as a separate quality of service data flow Flow_RemoteDrive. If the same vehicle reports multiple remote driving sensor data transmission tasks at the same time (all corresponding to the same Q_c), these tasks and the current task can be included in the Flow_RemoteDrive.

[0139] (4) SDAP bearer: The CTAP entity determines the quality of service indication information I_QoS={Flow_ID,DRB_ID,QoS_Profile} of the quality of service data flow Flow_RemoteDrive and sends the quality of service data flow Flow_RemoteDrive and its I_QoS to the SDAP entity. Based on this, the SDAP entity can map the quality of service data flow Flow_RemoteDrive to a dedicated low latency DRB (corresponding to DRB_ID) and configure a short TTI and high reliability HARQ retransmission strategy.

[0140] (5) Distribution at the receiving end: The base station side parses Task_ID=0x00A1 and Comp_QoS_Idx=0x10 for CTAP entity parsing, and directly distributes the data packet to the edge GPU computing node. The edge GPU computing node loads the safe driving model of Model_Version=0x03, skips the application layer secondary parsing, and can save 2 to 3ms of processing latency.

[0141] (6) Result feedback: The reasoning result (such as the form of control command) is encapsulated by the base station CTAP layer. Comp_Priority=7 can ensure that it is sent first in downlink scheduling and meets the 10ms delay constraint.

[0142] Technical effect: Compared with the communication method in related technologies where all data is mapped to the same DRB and then distributed after parsing at the application layer of the receiving end, the introduction of the CTAP layer reduces the end-to-end latency of remote driving control commands from 12-15ms to 8-10ms, meeting the stringent latency requirements of remote driving.

[0143] For real-time image recognition scenarios, the corresponding scenario description is as follows: At urban intersections, vehicles need to perform real-time image recognition of pedestrians and obstacles, with a recognition latency of ≤5ms and high computational complexity.

[0144] Accordingly, the processing flow for CTAP entities is as follows:

[0145] (1) Task Analysis: The in-vehicle AI application initiates an image recognition task, Q_comp={C_req=50GFLOPs,T_exec=5ms,M_size=200MB,A_model=95%}. Since the edge-side NPU only supports 20GFLOPs, edge-side collaborative inference needs to be used.

[0146] (2) CTAP header generation: Task_ID=0x00B2, QoS_Class=2, Comp_Priority=6, Comm_QoS_Idx=0x02 (corresponding to high bandwidth low jitter template), Comp_QoS_Idx=0x21 (corresponding to split inference-front-end model), Model_Version=0x05 (ResNet-front-end v5).

[0147] (3) Flow Mapping and Differentiation: The vehicle system simultaneously reports ordinary navigation video and real-time recognition images (both have the same Q_c but different Q_comp). The CTAP entity maps both to the same QoS Flow (sharing high-bandwidth communication resources), but distinguishes the two tasks through Comp_QoS_Idx:

[0148] Navigation video task: Comp_QoS_Idx=0x00 (Normal transmission, local processing on the device side)

[0149] Image recognition task: Comp_QoS_Idx=0x21 (split inference, intermediate features need to be uploaded)

[0150] (4) End-side execution: The on-board CTAP entity distributes the data packet with Comp_QoS_Idx=0x21 in the image data packet to the local NPU to perform the first L_1 layer inference of the neural network and generate intermediate feature tensors.

[0151] (5) Feature upload: The intermediate feature tensor is used as a new AI computing task. The CTAP entity regenerates the packet header (Task_ID=0x00B2-FEAT,Comp_QoS_Idx=0x22 ​​(corresponding to split inference-backend layer model)) and uploads it to the base station through the Uu interface.

[0152] (6) Edge execution: The base station side CTAP entity parsing Comp_QoS_Idx=0x22 ​​directly distributes the feature tensor to the edge GPU to execute the remaining L_2 layer inference and outputs the recognition result.

[0153] Technical effects: The CTAP layer achieves accurate differentiation of different computing tasks within the same Flow through Comp_QoS_Idx, enabling edge-to-edge collaborative split streams to share air interface resources with ordinary video streams but with differentiated processing. The end-to-end latency of image recognition is reduced from 8ms to 4.5ms, and the utilization rate of air interface resources is improved by about 30%.

[0154] For fault detection and safety warnings, the corresponding scenario is described as follows: The vehicle periodically reports sensor data for fault detection. The latency requirement is lenient (500ms-1s), but the data volume is large, requiring high-accuracy full-scale model inference. The CTAP side needs to perform batch task aggregation verification.

[0155] Accordingly, the processing flow for CTAP entities is as follows:

[0156] (1) Task analysis: Q_c={R_min=0.1Mbps, R_max=2Mbps, D_max=1s, J_max=100ms}, Q_comp={C_req=200GFLOPs, T_exec=500ms, M_size=1GB, A_model=99.5%}.

[0157] (2) CTAP header generation: Task_ID=0x00C3, QoS_Class=1, Comp_Priority=3, Comm_QoS_Idx=0x03 (corresponding to best-effort transmission template), Comp_QoS_Idx=0x30 (corresponding to full diagnostic model), Model_Version=0x0A (corresponding to fault detection large model v10).

[0158] (3) Batch aggregation: The vehicle-mounted CTAP entity aggregates 10 fault detection task data packets generated within 5 seconds (with the same Q_c and the same Comp_QoS_Idx) and maps them into a single QoS Flow, thereby reducing the DRB establishment overhead of the SDAP layer.

[0159] (4) Cloud offloading: The base station side CTAP entity parsing Comp_QoS_Idx=0x30 and Model_Version=0x0A indicates that the task requires the use of the full diagnostic model, while the edge cache only stores the high-frequency lightweight model. Therefore, the data packets can be directly distributed to the cloud TPU to perform complex diagnostic inference.

[0160] (5) Result return: After the cloud inference is completed, the inference result is transmitted back through the core network, base station and vehicle. The CTAP layer matches the result to the original computing task request through Task_ID=0x00C3.

[0161] Technical effects: The CTAP layer achieves full lifecycle tracking of tasks through Task_ID, accurately identifies large model requirements through Comp_QoS_Idx and triggers cloud unloading, avoiding model retrieval latency caused by missing cache at edge nodes (saving approximately 200ms), and the end-to-end latency of fault detection is stable at 600-800ms.

[0162] In summary, the innovative aspects of this disclosure include at least the following:

[0163] (1) Design the CTAP protocol layer to realize native air interface perception computing tasks: On the basis of the existing 5G air interface user plane protocol stack, a CTAP layer is added between the application layer and the SDAP layer. For the first time, the computing QoS dimension is introduced into the air interface protocol stack, so as to realize: QoS mapping and update of computing tasks; dual-dimensional joint index of communication QoS and computing QoS; accurate distribution of computing tasks by the receiver based on the air interface packet header.

[0164] (2) Standardized interaction between CTAP and SDAP, which is compatible with the existing protocol stack: The CTAP layer can interact with the SDAP layer through the standard SAP. The CTAP packet header is encapsulated in the SDAP service data unit. There is no need to modify the existing PDCP / RLC / MAC / PHY layer, ensuring compatibility with the current communication protocol and subsequent evolution versions, and has the ability to evolve smoothly.

[0165] (3) Differentiated delivery of AI inference in the Internet of Vehicles, based on the differentiation of computing tasks within the same Flow: By using the Comp_QoS_Idx subfield in the CTAP packet header, the granularity limitation of the SDAP layer in only distinguishing Flows through QFI is broken, and the differentiated identification and distribution of different inference tasks are encapsulated within the same QoS Flow. This eliminates the secondary parsing overhead of the application layer, and Comp_Priority can guarantee scheduling priority.

[0166] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but without changing the core design of the algorithm and process, are also within the scope of protection of this patent.

[0167] Secondly, embodiments of the present invention provide a communication device that can be disposed in a computing task adaptation protocol layer entity, wherein the computing task adaptation protocol layer is located between the application layer and the service data adaptation protocol layer.

[0168] Figure 3 This is a block diagram of a communication device provided according to an embodiment of this disclosure. See also... Figure 3 The communication device 300 may include the following modules:

[0169] The parsing module 301 is used to respond to a computing task request sent by the application layer, parse the computing task request, and obtain a first computing task data packet and a first parsing parameter. The first parsing parameter includes a communication quality parameter set and a computing quality parameter set corresponding to the computing task request.

[0170] Add module 302, which is used to add an adaptation layer header to the first computing task data packet according to the first parsing parameters, so as to obtain the second computing task data packet;

[0171] The mapping module 303 is used to map at least one second computing task data packet to a corresponding quality of service data stream, and to determine the quality of service indication information of the quality of service data stream;

[0172] The sending module 304 is used to send the service quality data stream and the corresponding service quality indication information to the service data adaptation protocol layer entity.

[0173] In summary, the communication device of this embodiment adds a CTAP layer to the existing 5G air interface user plane protocol stack, realizing a two-dimensional mapping of communication QoS and computing QoS. It can filter and map computing task data packets with the same communication QoS into a single QoS flow. Computing tasks corresponding to different computing QoS can be distinguished by fields in the adapter layer header. Furthermore, it sets QoS indication information corresponding to the QoS flow, realizing differentiated air interface bearing of computing tasks. Based on this, after the CTAP entity on the receiving end receives the data packet, it can restore the QoS flow association information by parsing the header and QoS indication information, thereby achieving accurate distribution of computing tasks.

[0174] The apparatus provided in the embodiments of the present invention has functions or includes modules that can be used to perform the methods described in the first aspect of the method embodiments above. The specific implementation and technical effects can be referred to the description of the method embodiments above. For the sake of brevity, they will not be repeated here.

[0175] It should be noted that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by this invention; however, this does not mean that other units are absent from this embodiment.

[0176] Reference Figure 4 This disclosure provides an electronic device, which includes:

[0177] One or more processors 401;

[0178] The memory 402 stores one or more programs that, when executed by one or more processors, enable the one or more processors to implement any of the communication methods described above.

[0179] One or more I / O interfaces 403 are configured to enable information exchange between the processor and memory.

[0180] Among them, processor 401 is a device with data processing capabilities, including but not limited to central processing unit (CPU); memory 402 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); I / O interface (read-write interface) 403 is connected between processor 401 and memory 402, and can realize information interaction between processor 401 and memory 402, including but not limited to data bus (Bus).

[0181] In some embodiments, the processor 401, memory 402, and I / O interface 403 are interconnected via a bus, and thus connected to other components of the computing device.

[0182] This embodiment also provides a computer-readable medium having a computer program stored thereon. When the program is executed by a processor, it implements the communication method provided in this embodiment. To avoid repetition, the specific steps of the communication method will not be repeated here.

[0183] This invention also provides a computer program product, including computer-readable code, which, when executed in a processor of an electronic device, causes the processor in the electronic device to perform the aforementioned communication method.

[0184] Those skilled in the art will understand that all or some of the steps, systems, or apparatuses in the methods, systems, and apparatuses described above can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0185] 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. Unless otherwise specified, 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.

[0186] Those skilled in the art will understand that although some embodiments described herein include certain features that are included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this embodiment and form different embodiments.

[0187] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A communication method, characterized in that, The method, applied to the computing task adaptation protocol layer entity, which is located between the application layer and the service data adaptation protocol layer, includes: In response to receiving a computing task request sent by the application layer, the computing task request is parsed to obtain a first computing task data packet and a first parsing parameter, wherein the first parsing parameter includes a communication quality parameter set and a computing quality parameter set corresponding to the computing task request; Based on the first parsing parameters, an adaptation layer header is added to the first computing task data packet to obtain the second computing task data packet; Map at least one of the second computing task data packets to a corresponding quality of service data stream, and determine the quality of service indication information of the quality of service data stream; The service quality data stream and the corresponding service quality indication information are sent to the service data adaptation protocol layer entity.

2. The method according to claim 1, characterized in that, The first parsing parameter also includes model attribute information; The step of adding an adaptation layer header to the first computing task data packet according to the first parsing parameters to obtain the second computing task data packet includes: Based on the communication quality parameter set, determine the service quality level identifier and the communication service quality index parameter; Based on the set of computational quality parameters and the model attribute information, determine the computational priority, computational service quality index parameters, and model version parameters; The adaptation layer header is generated based on the service quality level identifier, the communication service quality index, the calculation priority, the calculation service quality index, and the model version parameter. The adaptation layer header is added to the first computing task data packet to obtain the second computing task data packet; The service quality level identifier is used to characterize the communication priority of the computing task request, the communication service quality index parameter is used to indicate the target communication service quality configuration template that matches the computing task request, and the computing service quality index parameter is used to indicate the target computing service quality configuration template that matches the computing task request.

3. The method according to claim 2, characterized in that, The step of determining the service quality level identifier and communication service quality index parameters based on the communication quality parameter set includes: Based on the first communication quality parameter value requirements corresponding to each service quality level, determine the target service quality level that matches the communication quality parameter set, and determine the identifier of the target service quality level as the service quality level identifier; Based on the second communication quality parameter value requirements corresponding to each communication service quality configuration template, a target communication service quality configuration template matching the communication quality parameter set is determined, and the communication service quality index parameter is determined based on the identifier of the target communication service quality configuration template.

4. The method according to claim 2, characterized in that, The step of determining the computation priority, computation service quality index parameter, and model version parameter based on the computation quality parameter set and the model attribute information includes: Based on the preset quality priority mapping relationship, the calculation priority corresponding to the set of calculation quality parameters is determined; Based on the computing quality parameter value requirements corresponding to each computing service quality configuration template, a target computing service quality configuration template matching the computing quality parameter set is determined, and the computing service quality index parameter is determined based on the identifier of the target computing service quality configuration template. The model version parameters are determined based on the model attribute information.

5. The method according to any one of claims 1-4, characterized in that, The communication quality parameter set includes at least one of minimum data rate, maximum data rate, maximum end-to-end latency, and maximum jitter, and the computation quality parameter set includes at least one of computational complexity, maximum computational execution latency, model size, and model accuracy requirements.

6. The method according to claim 1, characterized in that, The step of mapping at least one of the second computing task data packets to a corresponding quality of service (QoS) data stream and determining the QoS indication information of the QoS data stream includes: Map multiple second computing task data packets with the same set of communication quality parameters to the same quality of service data stream; For any quality of service (QoS) data stream, determine the data radio bearer information and the corresponding set of QoS parameters used to carry the QoS data stream; obtain the QoS indication information of the QoS data stream based on the stream identifier of the QoS data stream, the data radio bearer information, and the set of QoS parameters. The data wireless bearer information is used to indicate the computing node carrying the computing task request. The computing node includes end-side computing nodes, edge computing nodes, and cloud computing nodes. The service quality parameter set includes at least some quality parameters from the communication quality parameter set and the computing quality parameter set.

7. The method according to any one of claims 1-4, characterized in that, The method further includes: In response to receiving a third computing task data packet sent by the service data adaptation protocol layer entity, the third computing task data packet is parsed to obtain a fourth computing task data packet and a second parsing parameter. The fourth computing task data packet is sent to the corresponding target computing node according to the second parsing parameters, and the target computing node is used to execute the corresponding computing task according to the fourth computing task data packet.

8. A communication device, characterized in that, The device, which is configured in the computing task adaptation protocol layer entity and is located between the application layer and the service data adaptation protocol layer, includes: The parsing module is used to respond to receiving a computing task request sent by the application layer, parse the computing task request, and obtain a first computing task data packet and a first parsing parameter. The first parsing parameter includes a communication quality parameter set and a computing quality parameter set corresponding to the computing task request. An add module is used to add an adaptation layer header to the first computing task data packet according to the first parsed parameters to obtain the second computing task data packet; A mapping module is used to map at least one of the second computing task data packets to a corresponding quality of service data stream, and to determine the quality of service indication information of the quality of service data stream; The sending module is used to send the quality of service data stream and the corresponding quality of service indication information to the service data adaptation protocol layer entity.

9. An electronic device, characterized in that, include: One or more processors; A memory having stored one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the communication method according to any one of claims 1-7; One or more I / O interfaces are connected between the processor and the memory and configured to enable information interaction between the processor and the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the communication method as described in any one of claims 1-7.

Images