A dedicated wireless protocol stack and service orchestration method for video streaming

Abstract

The present invention discloses a dedicated wireless protocol stack and service orchestration method for video streaming, wherein the dedicated wireless protocol stack is constructed as a cloud-edge-end system architecture, including a core network, a service-based RAN and a terminal device: the core network is deployed in the cloud and connected to a video application server; the service-based RAN is deployed at the edge of the network to provide wireless access services for the terminal device, and the service-based RAN includes a service management network element and a local AI service; the terminal device is wirelessly connected to the service-based RAN and uploads the video stream data collected by its camera through the service-based RAN using different processing strategies according to different resolutions: for 4K video stream: when the service-based RAN discovers the local AI service, the 4K video stream collected by the terminal device is diverted to the local AI service deployed at the edge of the network for processing; for 480P video stream: the service-based RAN directly transmits the 480P video stream uplink through the core network to the video application server deployed in the cloud for intelligent analysis and processing.

Description

Technical Field

[0001] The present invention relates to a video stream data transmission processing technology, and in particular to a dedicated wireless protocol stack and service orchestration method for video streams. Background Art

[0002] With the rapid development of wireless communication technology, mobile network bandwidth and transmission capabilities have significantly increased, and the demand for transmitting high-definition video streams over mobile networks is becoming increasingly common. In particular, in areas such as video surveillance, intelligent security, and virtual reality, the demand for real-time transmission and intelligent analysis of high-definition video data is increasing. Traditional wireless network architectures primarily focus on the transmission of general service traffic, while relatively little research has been conducted on the efficient transmission and intelligent processing of high-definition video streams, especially those with resolutions of 4K and above, over wireless networks.

[0003] In the current technological landscape, cloud computing and edge computing have become crucial tools for intelligent analysis of video streams. Cloud computing is suitable for handling large-scale, complex video data analysis tasks. However, due to the large volume of HD video data, processing it in the cloud consumes significant network bandwidth and can result in high latency. To reduce network burden and service latency, edge computing has emerged, migrating intelligent analysis tasks for some video streams to network edge nodes. However, existing wireless protocol stacks struggle to adapt to these complex application scenarios. Therefore, the design of specialized wireless protocol stacks for HD video streaming has become a hot topic of research. Summary of the Invention

[0004] The purpose of the present invention is to provide a dedicated wireless protocol stack and service orchestration method for video streaming to solve the following problems existing in the prior art:

[0005] 1) Network bandwidth pressure: 4K video streams generate enormous amounts of data. Traditional RAN (Radio Access Network) networks transmit all video data to the cloud for processing. This approach not only consumes a large amount of backhaul link bandwidth but also increases data transmission latency, impacting real-time performance.

[0006] 2) Computing resource optimization: In practical applications, video analysis requirements vary across different scenarios. For example, intelligent analysis of 480P video streams can be directly processed by cloud-based video servers; however, due to the large data volume of 4K video streams, preliminary analysis must be performed on the edge to reduce the amount of data transmitted back. However, existing technologies lack an effective mechanism to dynamically discover and utilize edge computing resources to achieve optimal processing of video streams with different resolutions.

[0007] 3) Protocol stack adaptability: The existing RAN protocol stack is difficult to flexibly adapt to different video streaming scenarios and lacks dedicated support for large data volume scenarios such as 4K video streaming, resulting in low system processing efficiency.

[0008] To achieve the above objectives, the present invention provides a dedicated wireless protocol stack for video streaming, which is constructed as a cloud-edge-end system architecture, including a core network, a service-oriented RAN, and terminal devices, wherein:

[0009] The core network is deployed in the cloud and connected to the video application server also deployed in the cloud, responsible for processing and storing video stream data from the terminal;

[0010] The service-based RAN is deployed at the network edge to provide wireless access services for the terminal device. The service-based RAN includes a service management network element and a local AI service also deployed at the network edge. The local AI service runs on an AI server. The service-based RAN dynamically discovers and utilizes the local AI service through the service management network element to process intelligent analysis tasks for 4K video streams.

[0011] The terminal device is wirelessly connected to the service-oriented RAN and uploads the video stream data collected by its camera through the service-oriented RAN using different processing strategies according to different resolutions, specifically:

[0012] For 4K video streams: When the service-oriented RAN discovers the local AI service, the 4K video stream collected by the terminal device is offloaded to the local AI service deployed at the network edge for processing;

[0013] For 480P video streams: the service-oriented RAN directly transmits the 480P video stream uplink through the core network to the video application server deployed in the cloud for intelligent analysis and processing.

[0014] In one embodiment of the present invention, the service-oriented RAN adopts service-oriented and modular design principles to decompose traditional radio access network functions into independent service components, and further includes:

[0015] Service-oriented RAN control plane: including CUCP, responsible for radio resource management, session management, and mobility management and control functions;

[0016] Service-oriented RAN user plane: including CUUP, which is responsible for handling the transmission of user data, including data encapsulation, encryption and compression;

[0017] Data flow distribution service: responsible for distributing user-plane data to different processing paths;

[0018] Service orchestrator: responsible for dynamically orchestrating and optimizing service instances based on business needs and network policies;

[0019] Decision-making and control functions: Make decisions on starting, stopping, and scaling service instances based on network policies and business needs;

[0020] Digital object management function: responsible for managing the status and configuration of digital objects in the network;

[0021] Service scheduling function: Responsible for the specific scheduling and scaling of service instances based on the decisions made by the decision control function;

[0022] Network Function Service Interface: This defines the communication protocols and interfaces between services, supporting service interoperability and modularity.

[0023] Tunnel conversion service: responsible for establishing and maintaining data transmission tunnels between the user plane and edge computing nodes;

[0024] Video stream forwarding module: responsible for sending video stream data packets to the local AI server via UDP to achieve local diversion and processing of video streams;

[0025] Service-oriented RAN prototype verification system: used to test and verify the functions of the service-oriented RAN architecture and components, supporting end-to-end business processes, including a dedicated 4K video prototype service-oriented RAN system and a dedicated non-4K video prototype service-oriented RAN system.

[0026] In one embodiment of the present invention, the service-oriented RAN prototype verification system defines a radio bearer mechanism for intelligent analysis of large video streams, namely, VF DRB. The implementation of the radio bearer mechanism specifically includes:

[0027] DRB configuration and data forwarding settings: When a user's terminal device accesses the network, the CUUP of the service-based RAN checks whether the DRB to which the uploaded data packets belong is a VF DRB. If so, the packets are forwarded to the AI ​​server for processing before reaching the NGU of the service-based RAN.

[0028] Video stream forwarding module settings: The video stream forwarding module is integrated into the CUUP uplink transmission processing component. Its main functions include: sending 4K video stream packets submitted by the CUUP UL thread to the AI ​​server via UDP; configuring the VF DRB identifier through the Telnet protocol; and querying the NRF to obtain the network address of the AI ​​server.

[0029] A data adaptation module is also added to the AI ​​server to process the video stream data transmitted from the video stream forwarding module. The data adaptation module is an independent process that runs on the AI ​​server together with the video server. The main functions of the data adaptation module include: creating a virtual network card on the AI ​​server so that the video analysis application can monitor and process the data received through the virtual network card; adapting the video stream data sent by the video stream forwarding module into data that the AI ​​server can parse and submitting it to the virtual network card; and

[0030] Define the interaction rules between the video stream forwarding module and CUUP UL thread (task, video stream forwarding module and NRF, and video stream forwarding module and data adaptation module.

[0031] In one embodiment of the present invention, the interaction rules between the video stream forwarding module and the CUUP UL thread task are specifically as follows:

[0032] The video stream forwarding module configures the VF DRB ID to the CUUP UL thread task. The CUUP UL thread task decides whether to submit the processed data packet to the task of the video stream forwarding module or the UPF of the core network based on the configuration.

[0033] In one embodiment of the present invention, the interaction rules between the video stream forwarding module and the NRF are specifically as follows:

[0034] The video stream forwarding module queries NRF for the IP address of the AI ​​server and the port number of the data adaptation application through HTTP requests.

[0035] In one embodiment of the present invention, the interaction rules between the video stream forwarding module and the data adaptation module are specifically as follows:

[0036] The data adaptation module receives UDP tunnel data packets from the physical network card, i.e., data packets obtained by the service-oriented RAN through tunnel conversion services after the video stream data packets uploaded by the terminal device are diverted;

[0037] The data adaptation module obtains the original video stream data packet;

[0038] The data adapter module writes the data into the virtual network card;

[0039] The video server reads the video stream data from the virtual network card.

[0040] In one embodiment of the present invention, the specific process of the service-based RAN discovering the local AI service includes:

[0041] Step S101: The tunnel switching service in the service-oriented RAN registers services with the RNRF module in the service-oriented RAN via the HTTP interface. Specifically, the tunnel switching service sends a service registration request to the RNRF module, and the RNRF module returns a service registration response to the tunnel switching service. The RNRF module is a network service discovery and subscription module that supports service discovery and subscription.

[0042] Step S102: After the local AI service is started, the local AI service registers services with the RNRF module in the service-oriented RAN through the HTTP interface;

[0043] Step S103: The service-oriented RAN obtains the service name, IP address / port information of the local AI service and responds with a service registration response;

[0044] Step S104: The service-oriented RAN queries the health status of the local AI service according to a preset timer;

[0045] Step S105: When the device terminal initiates a query for a local AI service, it is sent to the core network via the service-based RAN, and the core network returns the query result to the device terminal 103 via the service-based RAN.

[0046] In one embodiment of the present invention, a dedicated non-4K video prototype service-oriented RAN system is used to offload 480P video stream data. The non-4K video stream collected by the terminal device is uploaded to the cloud core network through the service-oriented RAN. The specific process is as follows:

[0047] Step S201: The service-oriented RAN queries the local AI service and adopts the service-oriented RAN protocol stack for non-4K video streams;

[0048] Step S202: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services;

[0049] Step S203: The camera of the terminal device collects video stream data;

[0050] Step S204: The terminal device does not find the local AI service, and the terminal device sends the collected non-4K video stream data to the service-based RAN;

[0051] Step S205: The service-oriented RAN receives the uplink video stream data service from the terminal device, and pushes the video stream data to the video application server through the core network.

[0052] In one embodiment of the present invention, a dedicated 4K video prototype service-oriented RAN system is used to offload 4K video stream data. The specific process is as follows:

[0053] Step S301: The service-based RAN discovers the local AI service, queries the metadata of the local AI service, including the service name and IP address / port information, and switches to the service-based RAN protocol stack dedicated to 4K video streaming;

[0054] Step S302: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services;

[0055] Step S303: The camera of the terminal device collects video stream data;

[0056] Step S304: After the terminal device queries the local AI service, the terminal device sends the collected video stream data to the service-based RAN;

[0057] Step S305: The service-oriented RAN receives the uplink video stream data service from the terminal device, identifies the local offload identifier, and then offloads the uplink video stream data of the terminal device to the local IP network. Specifically, the service-oriented RAN pushes the video stream data to the tunnel conversion service via the UDP transport protocol.

[0058] Step S306: The tunnel conversion service performs tunnel relationship adaptation on the diverted video stream data and pushes the video stream data to the target local AI service;

[0059] Step S307: The local AI service performs corresponding intelligent analysis on the uplink video stream data.

[0060] The present invention also provides a video stream-oriented service orchestration method, which is implemented by the aforementioned dedicated wireless protocol stack and specifically includes:

[0061] S1. The decision control function of the service-oriented RAN initiates a request to the computing plane control function to launch the AI ​​analysis service for edge video stream data.

[0062] S2. The computing surface control function feeds back the computing resource scheduling results to the decision control function;

[0063] S3. The decision control function initiates a request to the service scheduling function to schedule a service-based RAN protocol stack dedicated to 4K video streaming and AI analysis services for edge video stream data.

[0064] S4. The service dispatch function launches a service-oriented RAN protocol stack function instance dedicated to 4K video streaming;

[0065] S5. The service scheduling function launches the AI ​​analysis service for edge video stream data.

[0066] The dedicated wireless protocol stack and service orchestration method for video streaming disclosed in the present invention have at least the following advantages over the prior art:

[0067] 1) Improved network resource utilization efficiency: Through a cloud-edge-end collaborative system architecture, the present invention can process and analyze 4K video stream data in real time at edge nodes, significantly reducing the backhaul bandwidth required for video stream transmission and alleviating the burden on the core network. This rational allocation and optimization of resources effectively improves the utilization efficiency of the entire network.

[0068] 2) Reduced video stream processing latency: By introducing a dedicated 4K video streaming protocol stack into the service-oriented RAN, the present invention enables intelligent traffic diversion and processing on the RAN side of the data transmission path, reducing the latency of video stream data on the transmission link. In particular, the design of the video stream forwarding module (VSS) enables rapid processing of video streams in the access network, significantly shortening the time from acquisition to analysis and improving the real-time performance of the system.

[0069] 3) Enhanced system flexibility and adaptability: Through dynamic service orchestration, the present invention can flexibly switch between cloud-based analysis of 480P video streams and edge analysis of 4K video streams to meet the needs of different application scenarios. This flexibility not only adapts to changing network environments but also ensures efficient system operation and service quality.

[0070] 4) Efficient intelligent video stream analysis: This invention utilizes edge AI services to perform local analysis and processing of 4K video streams. Combined with the intelligent offload technology of service-oriented RAN, it greatly improves the processing efficiency of video stream data. By reducing the bottlenecks caused by long-distance transmission and centralized processing, the system can more quickly perform intelligent analysis of video streams, thereby improving the response speed and accuracy of applications.

[0071] 5) Improved system scalability: The proposed dynamic discovery and interaction mechanism between the service-oriented RAN and local AI services enables the system to scale computing resources and service instances up or down as needed, enhancing system scalability. Whether in small-scale deployments or large-scale applications, the system can dynamically adapt to varying loads and processing requirements.

[0072] 6) Optimized video data transmission and processing: Through the collaborative work of the data adapter module (DA) and the VSS module, the present invention realizes efficient transmission of video stream data from the physical network card to the virtual network card in the data transmission path, ensuring the efficiency and stability of video data in the processing link. This not only improves the processing capability of the system, but also further reduces the data packet loss rate and delay in network transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 A schematic diagram of a system architecture according to an embodiment of the present invention;

[0075] Figure 2 Used to illustrate the data flow direction of local forwarding of video stream data of the present invention;

[0076] Figure 3 Schematic diagram of the interaction between the video stream forwarding module task and the CUUP UL thread task in one embodiment of the present invention;

[0077] Figure 4 Schematic diagram of the interaction between the video stream forwarding module task and NRF in one embodiment of the present invention;

[0078] Figure 5 Schematic diagram of the interaction between the video stream forwarding module and the data adaptation module in one embodiment of the present invention;

[0079] Figure 6 A schematic diagram of a format for splitting video stream data packets uploaded by a terminal device in one embodiment of the present invention;

[0080] Figure 7 This is a schematic diagram of the format of a video stream data packet pushed by a service-oriented RAN to an AI server in one embodiment of the present invention;

[0081] Figure 8 This is a schematic diagram of the process of discovering local AI services by a service-oriented RAN in one embodiment of the present invention;

[0082] Figure 9 Schematic diagram of the processing flow of 480P video stream data in one embodiment of the present invention;

[0083] Figure 10 4K video stream data offloading process diagram according to an embodiment of the present invention;

[0084] Figure 11 A schematic diagram of service arrangement for video scenarios according to an embodiment of the present invention;

[0085] Figure 12 This is a schematic diagram of RNRF module service discovery in one embodiment of the present invention;

[0086] Figure 13 Schematic diagram of the RNRF module service discovery and subscription process in one embodiment of the present invention;

[0087] Figure 14 This is a schematic diagram of the RNRF module service registration process in one embodiment of the present invention;

[0088] Figure 15 This is a schematic diagram of the RNRF module service subscription / query process in one embodiment of the present invention.

[0089] Explanation of reference numerals: 101 - core network; 102 - service-oriented RAN; 103 - terminal device; 104 - video application server; 105 - local AI service; 106 - tunnel switching service; 107 - RNRF module. DETAILED DESCRIPTION

[0090] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without creative work are within the scope of protection of the present invention.

[0091] Before describing the embodiments of the present invention, the relevant key technologies involved in the present invention are first described as follows:

[0092] (1) Service-Oriented RAN: Service-Oriented RAN technology modularizes and service-oriented radio access network functions, dynamically orchestrating network functions based on application scenarios to deliver customized network services tailored to different business needs. This invention utilizes service-oriented RAN to flexibly distribute and process video streams, particularly by moving intelligent analysis of 4K video streams to edge nodes, thereby reducing the burden on the backhaul network.

[0093] (2) Edge Computing: Edge computing deploys computing and storage resources at the edge of the network close to the data source, enabling data processing close to the user, thereby reducing latency and bandwidth consumption. In this paper, edge computing is used to perform real-time AI intelligent analysis of 4K video streams, significantly improving the system's response speed and resource utilization efficiency.

[0094] (3) Dedicated wireless protocol stack: To meet the different processing requirements of 4K and 480P video streams, the present invention designs two dedicated wireless protocol stacks. The traditional cloud-based analysis mode is used for 480P video streams, while the edge analysis mode is performed for 4K video streams through a dedicated wireless protocol stack, thereby optimizing the overall performance of the system;

[0095] (4) Video Stream Intelligent Analysis: Video stream intelligent analysis technology involves real-time processing and analysis of high-definition video content, using AI algorithms to identify and process key events in the video. By deploying such analysis tasks at the edge nodes, the present invention can significantly reduce data transmission volume and service latency.

[0096] (5) Dynamic service orchestration: The dynamic service orchestration function in the present invention enables the system to flexibly schedule and switch different protocol stacks and service instances according to real-time network and business requirements, thereby achieving optimized processing of video streams with different resolutions.

[0097] Figure 1 FIG. 1 is a schematic diagram of a system architecture according to an embodiment of the present invention. Figure 1 As shown, this embodiment provides a dedicated wireless protocol stack for video streaming, which is constructed as a cloud-edge-end system architecture to support intelligent analysis of video streams in high-definition video surveillance scenarios. The system architecture includes a core network 101, a service-oriented RAN 102, and a terminal device 103, wherein:

[0098] The core network 101 is deployed in the cloud and connected to the video application server 104 (DN server in the figure), which is also deployed in the cloud. It is responsible for processing and storing video stream data from the terminal. The video server deployed in the cloud is mainly used to handle intelligent analysis requests from 480P video streams.

[0099] The service-oriented RAN 102 is deployed at the network edge and provides wireless access services for terminal devices 103. The service-oriented RAN 102 includes a service management network element (SME) and a local AI service 105, also deployed at the network edge. The local AI service 105 runs on an AI server. The service-oriented RAN 102 dynamically discovers and utilizes the local AI service 105 through the SME to handle intelligent analysis of 4K video streams. The local AI service 105, deployed at the network edge, receives 4K video streams from terminal devices 103 via UDP (User Datagram Protocol), processes and analyzes them in real time, and transmits the processing results back to the video application server 104 in the cloud via the service-oriented RAN 102.

[0100] Among them, the network element (NE) refers to a device or software entity with specific functions and independent operation capabilities in the communication network. They are the basic units that constitute the communication network and are responsible for completing various communication tasks. The network element can be a physical device, such as a switch, router, base station, etc., or it can be a software entity, such as a protocol stack, application, etc. The service management network element of this embodiment is a key component (or module) in the service-based RAN 102, which is used to discover, call, and manage various services in the service-based RAN 102. The service management network element enables the system to flexibly select the most appropriate service and protocol stack based on the resolution of the video stream (such as 4K or 480P) and business requirements, so as to achieve efficient video stream processing and intelligent analysis. This dynamic service orchestration function is a key component of the service-based RAN 102 architecture. It improves the flexibility and adaptability of the network while optimizing the efficiency of resource utilization.

[0101] In this embodiment, the Service-Based Radio Access Network (RAN) is a new network architecture that adopts service-oriented and modular design principles, decomposing traditional RAN functions into independent service components to improve network flexibility, scalability, and automation. The following is an overview of some key functions and modules in the Service-Based Radio Access Network (RAN):

[0102] Service-based RAN control plane (Control-plane): including CUCP (Central Unit Control-plane): responsible for control functions such as radio resource management, session management, and mobility management;

[0103] Service management network element: responsible for service registration, discovery, subscription and lifecycle management;

[0104] Service-oriented RAN user plane (User-plane): including CUUP (Central Unit User-plane): handles the transmission of user data, including data encapsulation, encryption, and compression;

[0105] Data flow distribution service: responsible for distributing user plane data to different processing paths, such as the core network or edge computing nodes;

[0106] Edge Computing: This includes local AI services that perform intelligent analysis of video streams, data processing, and other tasks at the edge of the network to reduce latency and bandwidth consumption.

[0107] Service orchestrator: Responsible for dynamically orchestrating and optimizing service instances based on business needs and network policies, such as intelligent analysis tasks for video streams;

[0108] Policy and Control Function (PCF): Makes decisions about starting, stopping, scaling, and other aspects of service instances based on network policies and business needs.

[0109] Digital Object Management Function (DOMF): manages the status and configuration of digital objects (such as cells and user equipment) in the network;

[0110] Service Scheduling Function (SSF): Responsible for the specific scheduling and scaling of service instances based on PCF decisions.

[0111] Service-based Interfaces (SBIs): define the communication protocols and interfaces between services, supporting service interoperability and modularity.

[0112] Tunnel conversion service: responsible for establishing and maintaining data transmission tunnels between the user plane and edge computing nodes;

[0113] Video Stream Steering (VSS): This module sends video stream data packets to the AI ​​server via UDP, enabling local distribution and processing of video streams.

[0114] Data Adaptation Module (DA): On the AI ​​server, adapts the video stream data transmitted from the VSS module into a format that the server can parse;

[0115] Service-oriented RAN prototype verification system: used to test and verify the functions of the service-oriented RAN architecture and components, supporting end-to-end business processes, including a dedicated 4K video prototype service-oriented RAN system and a dedicated non-4K video prototype service-oriented RAN system.

[0116] The terminal device 103 is wirelessly connected to the service-oriented RAN 102 and uploads the video stream data collected by its camera through the service-oriented RAN 102 using different processing strategies according to different resolutions, specifically:

[0117] For 4K video streams: When the service-oriented RAN 102 discovers the local AI service 105, the 4K video stream captured by the terminal device 103 is offloaded to the local AI service 105 deployed at the network edge for processing, thereby reducing backhaul link bandwidth usage and latency. The local AI service 105 includes an application server dedicated to processing 4K video streams.

[0118] For 480P video streams: the service-oriented RAN 102 directly transmits the 480P video stream uplink through the core network 101 to the video application server 104 deployed in the cloud for intelligent analysis and processing.

[0119] The terminal device 103 uploads the video stream data collected by the camera through its built-in camera video acquisition application, and the service-oriented RAN 102 determines the resolution of the video stream data:

[0120] If the video stream data is a 480P video stream, the service-oriented RAN 102 schedules its service-oriented RAN protocol stack dedicated to non-4K video streams and directly transmits the 480P video stream uplink to the cloud.

[0121] If the video stream data is a 4K video stream, the service-oriented RAN 102 queries the access network system through its service management network element whether there is an available local AI service 105. If so, the service-oriented RAN 102 schedules its service-oriented RAN protocol stack dedicated to 4K video streams and pushes the 4K video stream uploaded by the terminal device 103 to the local AI service 105 for processing. If not, the service-oriented RAN 102 schedules its service-oriented RAN protocol stack dedicated to 4K video streams and pushes the 4K video stream uploaded by the terminal device 103 to the cloud for processing.

[0122] The purpose of the local AI service 105 in this embodiment is to directly process video streams locally. The bandwidth required to transmit AI analysis results is much higher than directly transmitting the video. The difference between the processing of 480P and 4K video streams is that 480P video streams can be pushed directly to the cloud, while 4K video streams are prioritized for local processing, first searching for available local AI services. If so, they are processed locally. If not, they need to be uploaded to the cloud server for processing. The 4K video stream-specific protocol component is designed to achieve intelligent offload of large video stream data on the RAN side, ensuring that 4K video streams can be efficiently transmitted and processed.

[0123] Figure 2 The data flow used to illustrate the local forwarding of the video stream data of the present invention is as follows: Figure 2 As shown, the data streams uploaded by the terminal device through the wireless network include DRB#1 (used to indicate 480P video stream data) and DRB#2 (used to indicate 4K video stream data). The DRB#1 data stream is uploaded to the cloud through the edge service-oriented RAN, and the cloud returns the processing results, and forwards the DRB#2 data stream to the edge service for processing.

[0124] In this embodiment, a radio bearer mechanism for intelligent analysis of large video streams, namely VF DRB (Video Forwarding Data Radio Bearer), is defined in the service-oriented RAN prototype verification system. When a user equipment (UE) accesses the network, a dedicated VF DRB is created for 4K video streams to directly forward all 4K video stream data packets to the AI ​​server for processing. The implementation of this radio bearer mechanism specifically includes:

[0125] DRB configuration and data forwarding settings: When a user equipment (UE) accesses the network, the CUUP (Central Unit User-plane) of the Serving RAN 102 checks whether the DRB (Data Radio Bearer) to which the data packets uploaded by the UE belong is a VF DRB. If so, these packets are forwarded to the AI ​​server for processing before reaching the NGU (Next Generation User-plane) of the Serving RAN 102. This ensures that 4K video stream data can be processed in a timely manner on the RAN side, thereby reducing the burden on the core network.

[0126] Among them, CUUP is a component in the service-based RAN, responsible for handling the transmission of user-plane data. In the service-based RAN architecture, CUUP, as part of the user plane, performs data encapsulation, transmission, and related protocol functions to ensure that data can flow effectively in the wireless network. NGU is a key component for user-plane data transmission in 5G and future networks, responsible for handling user data forwarding, including data encapsulation, transmission, and related protocol functions. NGU is integrated into the service-based RAN and represents the function of user-plane data transmission. In the actual system architecture, it may be integrated into other user-plane processing units, such as CUUP or other user-plane functional entities. NGU usually works in conjunction with control-plane network elements (such as CUCP, Central Unit Control-plane) to ensure efficient and correct data transmission. In the service-based RAN architecture, the role of NGU is to achieve fast forwarding of user-plane data and support a variety of services and application scenarios, such as the transmission of high-definition video streams. It may include caching, processing, and optimization of data to meet the Quality of Service (QoS) requirements of different business needs. That is, the NGU usually does not appear as a separate physical entity in the system architecture, but is integrated into user-plane network elements such as CUUP as part of the user-plane data processing and transmission functions to support efficient and flexible data transmission services.

[0127] Video Stream Steering (VSS) module settings: The VSS module is integrated into the CUUP uplink processing component. Its main functions include: sending 4K video stream packets submitted by the CUUP UL thread (uplink) to the AI ​​server via UDP; configuring the VF DRB identifier through the Telnet protocol; and querying the Network Repository Function (NRF) to obtain the network address of the AI ​​server.

[0128] The VSS module is a component specially designed to process high-definition video streams, especially the intelligent diversion and forwarding of 4K video streams. The VSS module is integrated into the uplink transmission processing component of CUUP (Central Unit User-plane). Its main function is to send video stream data to the AI ​​server through the UDP protocol for real-time processing and analysis. NRF is a key component in the 5G network architecture. It is used as part of the service-oriented architecture to support service discovery and registration of network functions (NFs). The main functions of NRF include: 1) Service registration: Each service instance in the network can register its own service information on NRF, including the type, location, interface, etc. of the service. 2) Service discovery: When a service in the network needs to interact with other services, it can query the required service information through NRF to realize service discovery. 3) Service management: NRF is also responsible for managing the life cycle of the service, including service registration, update, and deregistration. In the service-oriented RAN architecture of this embodiment, NRF enables each network function to flexibly discover and access other services, supporting the dynamic service orchestration and network slicing in the service-oriented network. Through NRF, the network can respond more flexibly to changes in business needs and achieve optimal configuration and utilization of resources.

[0129] A data adaptation (DA) module is also added to the AI ​​server to process the video stream data transmitted from the video stream forwarding module. The data adaptation module is an independent process that runs on the AI ​​server together with the video server. The main functions of the data adaptation module include: creating a virtual network card on the AI ​​server so that the video analysis application can monitor and process the data received through the virtual network card; adapting the video stream data sent by the video stream forwarding module into data that the AI ​​server can parse and submitting it to the virtual network card; and

[0130] Define the interaction rules between the video stream forwarding module and the CUUP UL thread (UL Task), the video stream forwarding module and the NRF, and the video stream forwarding module and the data adaptation module.

[0131] Figure 3 FIG. 1 is a schematic diagram of the interaction between the video stream forwarding module task and the CUUP UL thread task in one embodiment of the present invention. Figure 3 As shown, in this embodiment, the interaction rules between the video stream forwarding (VSS) module and the CUUP UL thread (UL Task) task are specifically as follows:

[0132] The video stream forwarding module configures the VF DRB ID to the CUUP UL thread task. The CUUP UL thread task decides whether to submit the processed data packet to the task of the video stream forwarding module or the UPF (UserPlane Function) of the core network based on the configuration.

[0133] Figure 4 FIG. 1 is a schematic diagram of the interaction between the video stream forwarding module task and NRF in one embodiment of the present invention. Figure 4 As shown, in this embodiment, the interaction rules between the video stream forwarding (VSS) module and the NRF are specifically as follows:

[0134] The video stream forwarding module queries NRF through HTTP requests for the IP address of the AI ​​server and the port number of the data adaptation (DA) application. The service name registered by the data adaptation application is "videoAnalyzingService".

[0135] Figure 5 FIG. 1 is a schematic diagram of the interaction between the video stream forwarding module task and the data adaptation module in one embodiment of the present invention. Figure 5 As shown, in this embodiment, the interaction rules between the video stream forwarding (VSS) module and the data adaptation (DA) module are specifically as follows:

[0136] The data adaptation module receives data packets of the UDP tunnel from the physical network card (pNIC), i.e., data packets obtained after the service-oriented RAN diverts the video stream data packets uploaded by the terminal device 103 through the tunnel conversion service;

[0137] The data adaptation module obtains the original video stream data packet, that is, the data packet with the UDP+IP header of VSS-DA removed;

[0138] The data adapter module writes the data to the virtual network card (vNIC);

[0139] The video server reads the video stream data from the virtual network card, that is, obtains the video stream data sent to it by the UE.

[0140] Figure 6 FIG. 1 is a schematic diagram of a format for splitting video stream data packets uploaded by a terminal device in one embodiment of the present invention. Figure 7FIG. 1 is a schematic diagram of the format of a video stream data packet pushed by a service-oriented RAN to an AI server in one embodiment of the present invention. Figure 6 and Figure 7 As shown, the data adapter module receives the UDP tunnel data packet from the physical network card in the following format: Figure 6 Shown include:

[0141] The uplink video stream data portion (video stream data uploaded by the terminal device), which includes RTSP video data + TCP header (20-40 bytes) or UDP header (8 bytes) + IP header (20-40 bytes). The source IP and destination IP constituted by the TCP header or UDP header + IP header are the terminal device IP and the target AI server IP, respectively; and

[0142] The service-oriented RAN adds an 8-byte UDP header and an 20-60-byte IP header. The source and destination IP addresses are the service-oriented RAN IP and the tunnel switching service IP, respectively.

[0143] The format of the video data packet pushed by the service-based RAN to the AI ​​server at the edge is as follows: Figure 7 As shown, only the uplink video stream data portion is included, and the header portion of the service-oriented RAN IP and the tunnel switching service IP is removed.

[0144] Figure 8 This is a flow chart of a service-oriented RAN discovering a local AI service in one embodiment of the present invention. According to the above description, the local AI service 105 is discoverable. After the service-oriented RAN discovers the AI ​​service, it performs local network offload on the 4K streaming video. Figure 8 As shown, in this embodiment, the specific process of the service-oriented RAN 102 discovering the local AI service 105 includes:

[0145] Step S101: The tunnel switching service 106 in the service-based RAN 102 registers services with the RNRF module (a network service discovery and subscription module) in the service-based RAN 102 through the HTTP interface. Specifically, the tunnel switching service 106 sends a service registration request to the RNRF module, and the RNRF module returns a service registration response to the tunnel switching service 106.

[0146] Step S102: After the local AI service 105 is started, the local AI service 105 registers services with the RNRF module in the service-based RAN through the HTTP interface;

[0147] Step S103: The service-oriented RAN 102 obtains the service name and IP address / port information of the local AI service 105 and responds with a service registration response. In this embodiment, the service registration creates a <service name>-<IP&port> The mapping between them is similar to DNS mapping, so I will not go into details;

[0148] Step S104: the service-oriented RAN 102 queries the health status of the local AI service 105 according to a preset timer;

[0149] Step S105: When the device terminal 103 initiates a query for the local AI service 105, the query is sent to the core network 101 via the service-based RAN 102. The core network then returns the query result to the device terminal 103 via the service-based RAN 102.

[0150] Among them, under the premise of distributed service deployment (i.e., each service instance runs as an independent process), the address of the service instance may change dynamically during the process of service instance startup, deletion, migration, etc. In order to ensure mutual communication between services, a centralized service registration center is established to provide service discovery functions for other services. The RNRF module provided by the present invention is a network service discovery and subscription module. The RNRF module (Radio Network Repository Function) is used to support service discovery and subscription, and the RNRF module is implemented based on the open source software Consul.

[0151] Figure 12 FIG. 1 is a schematic diagram of RNRF module service discovery in one embodiment of the present invention. Figure 12 As shown in the figure, service instances 1 and 2 are both clients of the service registry and register services with the service registry. Service instance 2 queries the service registry for the network address of service instance 1 and then initiates a business request to service instance 1. The specific process is as follows:

[0152] Figure 13 FIG. 1 is a schematic diagram of the RNRF module service discovery and subscription process in one embodiment of the present invention. Figure 13 As shown in the figure, the service discovery and subscription process of the RNRF module is as follows:

[0153] Step S1101: serv1 (service instance 1) initiates a registration request to serv_reg (service registration center). serv_reg (service registration center) maintains metadata of serv1 (service instance 1) and starts health monitoring.

[0154] Step S1102: serv_reg (service registration center) replies to serv1 (service instance 1) with the registration result and the heartbeat port;

[0155] Step S1103: serv2 (service instance 2) initiates a registration request to serv_reg (service registration center). serv_reg (service registration center) maintains metadata of serv2 (service instance 2) and starts health monitoring.

[0156] Step S1104: serv_reg (service registration center) replies to serv2 (service instance 2) with the registration result and the heartbeat port;

[0157] Step S1105: serv2 (service instance 2) subscribes to the service of serv1 (service instance 1) from serv_reg (service registration center) and sends serv_name (service instance name). serv_reg (service registration center) maintains the subscription information of serv2 (service instance 2) and pushes it to serv2 (service instance 2) when the data subscribed by serv2 (service instance 2) changes.

[0158] Step S1106: serv_reg (service registration center) pushes the IP address and port of serv1 (service instance 1) to serv2 (service instance 2) for the first time;

[0159] Step S1107: serv2 (service instance 2) sends a service request to serv1 (service instance 1);

[0160] Step S1108: serv1 (service instance 1) feeds back a service response to serv2 (service instance 2).

[0161] Figure 14 FIG. 1 is a schematic diagram of the RNRF module service registration process in one embodiment of the present invention. Figure 14 As shown, the service registration process of the RNRF module is as follows:

[0162] After serv_reg (service registration center) is started and the interface is initialized,

[0163] Step S1201: Any service instance (serv) sends a registration request, including information such as serv_name (service name), ip_addr (IP address), and serv_port (port);

[0164] Step S1202: serv_reg (service registration center) receives the request and gives a response whether the registration is successful:

[0165] If the registration is successful, serv_reg (service registration center) feeds back the heartbeat signal interface information to the service instance (serv);

[0166] If registration fails, serv_reg (service registration center) will feedback the reason for registration failure to the service instance (serv).

[0167] Figure 15 FIG. 1 is a schematic diagram of a RNRF module service subscription / query process in an embodiment of the present invention. Figure 15 As shown, the service subscription and query process of the RNRF module is as follows:

[0168] Step S1301: Any service instance subscribes to the network address of serv_name (service name) from the service registration center:

[0169] If the service registration center has subscribed service information, the service instance is added to the subscription list of serv_name (service name);

[0170] If the service registration center does not have the subscribed service information, the feedback query fails;

[0171] Step S1302: When the service registration center receives the address information modification request of serv_name (service name), it traverses the subscription list of serv_name (service name) to modify it;

[0172] Step S1303: The service registration center publishes the address information change of serv_name (service name).

[0173] Figure 9 This is a schematic diagram of the processing flow of 480P video stream data in one embodiment of the present invention. Since 480P video stream data is mainly analyzed and processed by cloud servers, the present invention uses a dedicated non-4K video prototype service-oriented RAN system for 480P video stream data, and uploads the non-4K video stream collected by the terminal device to the cloud core network through the service-oriented RAN. This method does not discover local AI services, such as Figure 9 As shown, in this embodiment, the 480P video stream data offloading process is specifically as follows:

[0174] Step S201: The service-oriented RAN queries the local AI service and adopts the service-oriented RAN protocol stack for non-4K video streams;

[0175] Step S202: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services;

[0176] Step S203: The camera of the terminal device collects video stream data;

[0177] Step S204: The terminal device does not find the local AI service, and the terminal device sends the collected non-4K video stream data (such as 480P) to the service-based RAN;

[0178] Step S205: The service-oriented RAN receives the uplink video stream data service from the terminal device, and pushes the video stream data to the video application server through the core network.

[0179] Figure 10 This is a schematic diagram of the offload processing flow of 4K video stream data in one embodiment of the present invention. For 4K video stream data services, the present invention adopts a dedicated 4K video prototype service-oriented RAN system. The 4K video stream data collected by the terminal, the service-oriented RAN matches the video data that meets the local AI service characteristics during the service transmission process, and pushes it to the local AI service, such as Figure 10 As shown, in this embodiment, the 4K video stream data offloading process is specifically as follows:

[0180] Step S301: The service-based RAN discovers the local AI service in the system, queries the metadata of the local AI service, including the service name and IP address / port information, and switches to the service-based RAN protocol stack dedicated to 4K video streaming;

[0181] Step S302: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services;

[0182] Step S303: The camera of the terminal device collects video stream data;

[0183] Step S304: After the terminal device queries the local AI service, the terminal device sends the collected video stream data to the service-based RAN;

[0184] Step S305: The service-oriented RAN receives the uplink video stream data service from the terminal device and identifies the local offload identifier (e.g., VF DRB dedicated to 4K video streams). It then offloads the uplink video stream data of the terminal device to the local IP network, specifically by pushing the video stream data to the tunnel conversion service via the UDP transport protocol.

[0185] Step S306: The tunnel conversion service performs tunnel relationship adaptation on the diverted video stream data and pushes the video stream data to the target local AI service;

[0186] Step S307: The local AI service performs corresponding intelligent analysis on the uplink video stream data.

[0187] Figure 11This is a service orchestration diagram for video scenarios in one embodiment of the present invention. The present invention performs service orchestration for two application scenarios: 480P video stream cloud analysis and 4K video stream edge analysis. Assuming a scenario where a 480P dedicated protocol stack switches to a 4K dedicated protocol stack, and assuming that the wireless network in a certain area only needs to serve the 480P video cloud AI analysis scenario, at this time, according to the instructions of the network operator or network service manager, the wireless network in this area needs to serve the 4K video AI analysis scenario. In order to reduce the bandwidth pressure of the backhaul network, a 4K dedicated wireless protocol stack needs to be orchestrated. Figure 11 As shown, this embodiment provides a service orchestration method for video streaming, which is implemented through the aforementioned dedicated wireless protocol stack and specifically includes:

[0188] S1. The service-oriented RAN's decision control function (PCF) initiates a request to the compute plane control function (implemented by the PCF in the service-oriented RAN) to initiate an AI analysis service for edge video stream data.

[0189] S2. The computing surface control function feeds back the computing resource scheduling results to the decision control function;

[0190] S3. The decision control function initiates a request to the service scheduling function (SSF) to schedule a service-based RAN protocol stack dedicated to 4K video streaming and AI analysis services for edge video streaming data.

[0191] S4. The service scheduling function pulls up a service-oriented RAN protocol stack function instance dedicated to 4K video streaming (service-oriented RAN control plane and user plane services);

[0192] S5. The service scheduling function launches the AI ​​analysis service for edge video stream data.

[0193] In summary, the present invention designs a service-oriented RAN protocol stack specifically for 4K video streams, including VideoForwarding DRB (VF DRB) and a video stream forwarding module (VSS). VF DRB effectively diverts and processes 4K video stream data on the RAN side, and the VSS module is responsible for sending the video stream data to the AI ​​server through the UDP protocol to achieve real-time analysis and processing. The present invention proposes an interaction mechanism between service-oriented RAN and local AI services. Through the service management network element, the service-oriented RAN can automatically discover and register local AI services, and flexibly select processing paths in different scenarios to ensure the real-time performance and processing efficiency of the system. In response to the needs of video streams with different resolutions, the present invention provides a dynamically switching service orchestration method that can switch between 480P video stream cloud analysis and 4K video stream edge analysis. The system starts or shuts down the corresponding service instance according to actual needs to achieve optimal resource configuration. The present invention can identify and process 4K video streams by designing a dedicated 4K video stream local diversion processing process, reducing long-distance data transmission and reducing latency. The collaboration between the data adapter module (DA) and the VSS module ensures the efficiency and stability of the video data transmission process from the physical network card to the virtual network card.

[0194] Those skilled in the art will appreciate that the accompanying drawings are merely schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0195] Those skilled in the art will appreciate that the modules in the apparatuses of the embodiments may be distributed in the apparatuses of the embodiments as described in the embodiments, or may be located in one or more apparatuses different from the embodiments with corresponding changes. The modules in the above embodiments may be combined into one module or further divided into multiple sub-modules.

[0196] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dedicated wireless protocol stack for video streaming, built as a cloud-edge-end system architecture, characterized by: It includes the core network, service-oriented RAN, and terminal equipment, including: The core network is deployed in the cloud and connected to the video application server also deployed in the cloud, responsible for processing and storing video stream data from the terminal; The service-based RAN is deployed at the network edge to provide wireless access services for the terminal device. The service-based RAN includes a service management network element and a local AI service also deployed at the network edge. The local AI service runs on an AI server. The service-based RAN dynamically discovers and utilizes the local AI service through the service management network element to process intelligent analysis tasks for 4K video streams. The terminal device is wirelessly connected to the service-oriented RAN and uploads the video stream data collected by its camera through the service-oriented RAN using different processing strategies according to different resolutions, specifically: For 4K video streams: When the service-oriented RAN discovers the local AI service, the 4K video stream collected by the terminal device is offloaded to the local AI service deployed at the network edge for processing; For 480P video streams: the service-oriented RAN directly transmits the 480P video stream uplink through the core network to the video application server deployed in the cloud for intelligent analysis and processing.

2. The dedicated wireless protocol stack for video streaming according to claim 1, wherein: The Service-Based RAN adopts service-oriented and modular design principles, decomposing traditional radio access network functions into independent service components. It also includes: Service-oriented RAN control plane: including CUCP, responsible for radio resource management, session management, and mobility management and control functions; Service-oriented RAN user plane: including CUUP, which is responsible for handling the transmission of user data, including data encapsulation, encryption and compression; Data flow distribution service: responsible for distributing user-plane data to different processing paths; Service orchestrator: responsible for dynamically orchestrating and optimizing service instances based on business needs and network policies; Decision-making and control functions: Make decisions on starting, stopping, and scaling service instances based on network policies and business needs; Digital object management function: responsible for managing the status and configuration of digital objects in the network; Service scheduling function: Responsible for the specific scheduling and scaling of service instances based on the decisions made by the decision control function; Network Function Service Interface: This defines the communication protocols and interfaces between services, supporting service interoperability and modularity. Tunnel conversion service: responsible for establishing and maintaining data transmission tunnels between the user plane and edge computing nodes; Video stream forwarding module: responsible for sending video stream data packets to the local AI server via UDP to achieve local diversion and processing of video streams; Service-oriented RAN prototype verification system: used to test and verify the functions of the service-oriented RAN architecture and components, supporting end-to-end business processes, including a dedicated 4K video prototype service-oriented RAN system and a dedicated non-4K video prototype service-oriented RAN system.

3. The dedicated wireless protocol stack for video streaming according to claim 2, wherein: The service-oriented RAN prototype verification system defines a radio bearer mechanism for intelligent analysis of large video streams, namely VF DRB. The implementation of this radio bearer mechanism specifically includes: DRB configuration and data forwarding settings: When a user's terminal device accesses the network, the CUUP of the service-based RAN checks whether the DRB to which the uploaded data packets belong is a VF DRB. If so, the packets are forwarded to the AI ​​server for processing before reaching the NGU of the service-based RAN. Video stream forwarding module settings: The video stream forwarding module is integrated into the CUUP uplink transmission processing component. Its main functions include: sending 4K video stream packets submitted by the CUUP UL thread to the AI ​​server via UDP; configuring the VF DRB identifier through the Telnet protocol; and querying the NRF to obtain the network address of the AI ​​server. A data adaptation module is also added to the AI ​​server to process the video stream data transmitted from the video stream forwarding module. The data adaptation module is an independent process that runs on the AI ​​server together with the video server. The main functions of the data adaptation module include: creating a virtual network card on the AI ​​server so that the video analysis application can monitor and process the data received through the virtual network card; adapting the video stream data sent by the video stream forwarding module into data that the AI ​​server can parse and submitting it to the virtual network card; and Define the interaction rules between the video stream forwarding module and CUUP UL thread tasks, the video stream forwarding module and NRF, and the video stream forwarding module and data adaptation module.

4. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: The interaction rules between the video stream forwarding module and the CUUP UL thread task are as follows: The video stream forwarding module configures the VF DRB ID to the CUUP UL thread task. The CUUP UL thread task decides whether to submit the processed data packet to the task of the video stream forwarding module or the UPF of the core network based on the configuration.

5. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: The interaction rules between the video stream forwarding module and NRF are as follows: The video stream forwarding module queries NRF for the IP address of the AI ​​server and the port number of the data adaptation application through HTTP requests.

6. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: The interaction rules between the video stream forwarding module and the data adaptation module are as follows: The data adaptation module receives UDP tunnel data packets from the physical network card, i.e., data packets obtained by the service-oriented RAN through tunnel conversion services after the video stream data packets uploaded by the terminal device are diverted; The data adaptation module obtains the original video stream data packet; The data adapter module writes the data into the virtual network card; The video server reads the video stream data from the virtual network card.

7. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: The specific process of service-oriented RAN discovering local AI services includes: Step S101: The tunnel switching service in the service-oriented RAN registers services with the RNRF module in the service-oriented RAN via the HTTP interface. Specifically, the tunnel switching service sends a service registration request to the RNRF module, and the RNRF module returns a service registration response to the tunnel switching service. The RNRF module is a network service discovery and subscription module that supports service discovery and subscription. Step S102: After the local AI service is started, the local AI service registers services with the RNRF module in the service-oriented RAN through the HTTP interface; Step S103: The service-oriented RAN obtains the service name, IP address / port information of the local AI service and responds with a service registration response; Step S104: The service-oriented RAN queries the health status of the local AI service according to a preset timer; Step S105: When the device terminal initiates a query for a local AI service, it is sent to the core network via the service-based RAN, and the core network returns the query result to the device terminal 103 via the service-based RAN.

8. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: A dedicated non-4K video prototype service-oriented RAN system is used to offload 480P video stream data. The non-4K video streams collected by terminal devices are uploaded to the cloud core network through the service-oriented RAN. The specific process is as follows: Step S201: The service-oriented RAN queries the local AI service and adopts the service-oriented RAN protocol stack for non-4K video streams; Step S202: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services; Step S203: The camera of the terminal device collects video stream data; Step S204: The terminal device does not find the local AI service, and the terminal device sends the collected non-4K video stream data to the service-based RAN; Step S205: The service-oriented RAN receives the uplink video stream data service from the terminal device, and pushes the video stream data to the video application server through the core network.

9. The dedicated wireless protocol stack for video streaming according to claim 3, wherein: A dedicated 4K video prototype service-oriented RAN system is used to offload 4K video stream data. The specific process is as follows: Step S301: The service-based RAN discovers the local AI service, queries the metadata of the local AI service, including the service name and IP address / port information, and switches to the service-based RAN protocol stack dedicated to 4K video streaming; Step S302: After the terminal device accesses the wireless network, the terminal device queries the network for local AI services; Step S303: The camera of the terminal device collects video stream data; Step S304: After the terminal device queries the local AI service, the terminal device sends the collected video stream data to the service-based RAN; Step S305: The service-oriented RAN receives the uplink video stream data service from the terminal device, identifies the local offload identifier, and then offloads the uplink video stream data of the terminal device to the local IP network. Specifically, the service-oriented RAN pushes the video stream data to the tunnel conversion service via the UDP transport protocol. Step S306: The tunnel conversion service performs tunnel relationship adaptation on the diverted video stream data and pushes the video stream data to the target local AI service; Step S307: The local AI service performs corresponding intelligent analysis on the uplink video stream data.

10. A video streaming-oriented service orchestration method, implemented by the dedicated wireless protocol stack according to any one of claims 3 to 9, characterized in that: Specifically include: S1. The decision control function of the service-oriented RAN initiates a request to the computing plane control function to launch the AI ​​analysis service for edge video stream data. S2. The computing surface control function feeds back the computing resource scheduling results to the decision control function; S3. The decision control function initiates a request to the service scheduling function to schedule a service-based RAN protocol stack dedicated to 4K video streaming and AI analysis services for edge video stream data. S4. The service dispatch function launches a service-oriented RAN protocol stack function instance dedicated to 4K video streaming; S5. The service scheduling function launches the AI ​​analysis service for edge video stream data.

Images