Communication method and apparatus for providing artificial intelligence service in real-time communication-based communication system

Abstract

The present disclosure relates to a communication method and apparatus for efficiently providing an AI service in an RTC-based communication system. According to an embodiment of the present disclosure, a method performed by a UE to provide a real-time AI service in an RTC-based communication system comprises the steps of: receiving, through an RTC AF, an ASPD for the real-time AI service, which is provided from an RTC ASP; registering information about the UE in an RTC AS in order to use the real-time AI service; performing negotiation for discovering an AI processing endpoint for transmitting and receiving media data of the real-time AI service to and from the RTC AS; and establishing a session for the real-time AI service with the discovered AI processing endpoint.

Description

Communication method and device for providing artificial intelligence services in a real-time communication-based communication system

[0001] The present disclosure relates to a communication method and device for providing AI (artificial intelligence) service in a RTC (real time communication) based communication system.

[0002] 5G mobile communication technology defines a wide frequency band to enable fast transmission speeds and new services, and can be implemented not only in the sub-6GHz frequency band such as 3.5 gigahertz (3.5GHz), but also in the ultra-high frequency band called millimeter wave (mmWave) such as 28GHz and 39GHz ('Above 6GHz'). In addition, for 6G mobile communication technology, which is called the system after 5G communication (Beyond 5G), implementation in the terahertz band (for example, the 3 terahertz (3THz) band at 95GHz) is being considered to achieve a transmission speed that is 50 times faster than 5G mobile communication technology and an ultra-low latency time that is reduced to one-tenth.

[0003] In the early stages of 5G mobile communication technology, the goal is to support services and satisfy performance requirements for enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC). These include beamforming and massive MIMO to mitigate path loss of radio waves in ultra-high frequency bands and increase the transmission distance of radio waves, support for various numerologies (such as operation of multiple subcarrier intervals) and dynamic operation of slot formats for efficient use of ultra-high frequency resources, initial access technology to support multi-beam transmission and wideband, definition and operation of BWP (Bidth Part), new channel coding methods such as LDPC (Low Density Parity Check) codes for large-capacity data transmission and Polar Code for reliable transmission of control information, and L2 pre-processing (L2). Standardization has been made for network slicing, which provides dedicated networks specialized for specific services, and pre-processing.

[0004] Currently, discussions are underway to improve and enhance the initial 5G mobile communication technology in consideration of the services that 5G mobile communication technology was intended to support, and physical layer standardization is in progress for technologies such as V2X (Vehicle-to-Everything) to help autonomous vehicles make driving decisions and increase user convenience based on their own location and status information transmitted by vehicles, NR-U (New Radio Unlicensed) for the purpose of system operation that complies with various regulatory requirements in unlicensed bands, NR terminal low power consumption technology (UE Power Saving), Non-Terrestrial Network (NTN), which is direct terminal-satellite communication to secure coverage in areas where communication with terrestrial networks is impossible, and Positioning.

[0005] In addition, standardization of wireless interface architecture / protocols is in progress for technologies such as intelligent factories (Industrial Internet of Things, IIoT) to support new services through linkage and convergence with other industries, Integrated Access and Backhaul (IAB) that provides nodes for expanding network service areas by integrating wireless backhaul links and access links, Mobility Enhancement technology including Conditional Handover and Dual Active Protocol Stack (DAPS) handover, and 2-step random access (2-step RACH for NR) that simplifies random access procedures. Standardization is also in progress for system architecture / services such as 5G baseline architecture (e.g., Service-based Architecture, Service-based Interface) for grafting Network Functions Virtualization (NFV) and Software-Defined Networking (SDN) technologies, and Mobile Edge Computing (MEC) that provides services based on the location of the terminal.

[0006] Once these 5G mobile communication systems are commercialized, an explosive increase in connected devices will be connected to the communication network, necessitating enhanced functionality and performance of 5G mobile communication systems and integrated operation of these connected devices. To this end, new research will be conducted on improving 5G performance and reducing complexity, supporting AI services, supporting metaverse services, and drone communications by utilizing eXtended Reality (XR), Artificial Intelligence (AI), and Machine Learning (ML) to efficiently support Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR).

[0007] In addition, the development of these 5G mobile communication systems includes new waveforms to ensure coverage in the terahertz band of 6G mobile communication technology, multi-antenna transmission technologies such as Full Dimensional MIMO (FD-MIMO), Array Antenna, and Large Scale Antenna, metamaterial-based lenses and antennas to improve the coverage of terahertz band signals, high-dimensional spatial multiplexing technology using Orbital Angular Momentum (OAM), Reconfigurable Intelligent Surface (RIS) technology, as well as full duplex technology to improve the frequency efficiency and system network of 6G mobile communication technology, satellite, AI (Artificial Intelligence) from the design stage and AI-based communication technology that realizes system optimization by internalizing end-to-end AI support functions, and ultra-high-performance communication and computing resources to provide services with complexity that exceeds the limits of terminal computing capabilities. It can serve as a basis for the development of next-generation distributed computing technologies that can be realized by utilizing them.

[0008] The present disclosure provides a communication method and device for efficiently providing AI services in an RTC-based communication system.

[0009] The present disclosure provides a communication method and device for efficiently providing AI services by using group identification information for input and output of an AI model in an RTC-based communication system.

[0010] The present disclosure provides a method and device for efficiently searching for an AI processing end point for transmitting and receiving media data of a real-time AI service in an RTC-based communication system.

[0011] The present disclosure provides a communication method and device for performing AI segmentation inference in a terminal and a network in an RTC-based communication system.

[0012] According to an embodiment of the present disclosure, a method performed in a terminal for providing a real-time AI service in a communication system based on RTC (real time communication) may include a process of receiving an ASPD (AI service provision description) for the real-time AI (artificial intelligence) service provided from an RTC ASP through an RTC AF (application function), a process of registering information about the terminal in an RTC AS (application server) for use of the real-time AI service, a process of performing negotiation for searching for an AI processing endpoint for transmitting and receiving media data of the real-time AI service with the RTC AS, and a process of establishing a session for the real-time AI service with the searched AI processing endpoint.

[0013] In one embodiment, the process of performing the negotiation in the method may further include the process of transmitting a first message including input group information of the AI ​​model to a SWAP (simple WebRTC application protocol) server within the RTC AS for searching for the AI ​​processing endpoint, and the process of receiving a second message including output group information of the AI ​​model from the searched AI processing endpoint through the SWAP server.

[0014] In one embodiment, the input group information and the output group information may each include at least one of language-specific text and voice information that can be selected as input or output in the AI ​​model.

[0015] In one embodiment, the first message may further include at least one of setting-related information including at least one of characteristic information of the AI ​​service selected by the terminal, end-to-end connection type information, and information about the AI ​​model, and matching criteria information used to search for the AI ​​processing endpoint, including at least one of performance specifications of the terminal and information about performance specifications provided by the target endpoint.

[0016] In one embodiment, the process of performing the negotiation in the method may further include a process of determining whether to execute a split inference of the AI ​​model between the terminal and the RTC AS based on the requirements of the AI ​​model and the performance of the terminal.

[0017] In addition, according to an embodiment of the present disclosure, a user equipment (UE) includes at least one transceiver, at least one processor communicatively coupled to the at least one transceiver, and at least one memory communicatively coupled to the at least one processor and storing instructions executable individually or in combination by the at least one processor, wherein the instructions cause the UE to receive, through the at least one transceiver via an RTC AF, an ASPD for a real-time AI service provided from an RTC ASP, register, through the at least one transceiver, information about the UE with an RTC AS for use of the real-time AI service, perform negotiation for searching for an AI processing endpoint for transmitting and receiving media data of the real-time AI service with the RTC AS, and establish, through the at least one transceiver, a session for the real-time AI service with the searched AI processing endpoint.

[0018] Also, according to an embodiment of the present disclosure, in an RTC-based communication system, an RTC AS includes at least one transceiver, at least one processor communicatively coupled to the at least one transceiver, and at least one memory communicatively coupled to the at least one processor and storing instructions executable individually or in combination by the at least one processor, wherein the instructions cause the RTC AS to receive, through the at least one transceiver, information about a terminal for using a real-time AI service from the terminal, to perform negotiation for searching for an AI processing end point for transmitting and receiving media data of the real-time AI service with the terminal, and to perform signaling for establishing a session for the real-time AI service between the terminal and the searched AI processing end point.

[0019] In one embodiment, the instructions executable individually or in combination by the at least one processor may cause the RTC AS to, through the at least one transceiver, cause a SWAP server within the RTC AS to receive a first message from the terminal, the first message including input group information of the AI ​​model for discovery of the AI ​​processing endpoint, and cause the discovered AI processing endpoint within the RTC AS to, through the at least one transceiver, transmit a second message including output group information of the AI ​​model via the SWAP server.

[0020] Figure 1 is a diagram showing an example configuration of a communication system for real-time communication (RTC).

[0021] FIG. 2 is a diagram showing an example configuration of a communication system for real-time communication (RTC) according to an embodiment of the present disclosure;

[0022] FIG. 3 is a drawing to explain the Input / Output / Split points and group identifiers of the AI ​​model according to an embodiment of the present disclosure.

[0023] FIG. 4 is a diagram illustrating a case where there are two options for each of the input and output points of an AI model according to an embodiment of the present disclosure;

[0024] FIG. 5 is a diagram showing an example of an AI model including a Split point according to an embodiment of the present disclosure;

[0025] FIG. 6 is a flowchart illustrating a procedure for a terminal to obtain / negotiate information about an entity for an AI service by communicating with an RTC AS according to an embodiment of the present disclosure; and

[0026] FIG. 7 is a diagram showing the configuration of a network entity in a wireless communication system according to one embodiment of the present disclosure.

[0027] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Furthermore, detailed descriptions of related known functions or configurations will be omitted if they are deemed to unnecessarily obscure the gist of the present disclosure. Furthermore, the terms described below are defined in light of their functions within the present disclosure and may vary depending on the intent or custom of the user or operator. Therefore, their definitions should be based on the overall content of this specification.

[0028] The advantages and features of the present disclosure, and methods for achieving them, will become clearer with reference to the embodiments described in detail below together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below and may be implemented in various different forms. These embodiments are provided solely to ensure that the disclosure of the present disclosure is complete and to fully inform those skilled in the art of the scope of the invention, and the present disclosure is defined only by the scope of the claims. Like reference numerals designate like elements throughout the specification.

[0029] At this time, it will be understood that each block of the processing flow diagrams and combinations of the flow diagrams can be performed by computer program instructions.

[0030] Additionally, each block may represent a module, segment, or portion of code that contains one or more executable instructions for performing a specific logical function(s). It should also be noted that in some alternative implementation examples, the functions described in the blocks may occur out of order. For example, two blocks depicted in succession may actually be executed substantially concurrently, or the blocks may sometimes be executed in reverse order, depending on their respective functions.

[0031] Here, the term '~ part' used in this embodiment means software or hardware components such as FPGA (Field Programmable Gate Array) or ASIC (Application Specific Integrated Circuit), and the '~ part' performs certain roles. However, the '~ part' is not limited to software or hardware. The '~ part' may be configured to be on an addressable storage medium or may be configured to play one or more processors. Therefore, as an example, the '~ part' includes components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and '~ parts' may be combined into a smaller number of components and '~ parts' or further separated into additional components and '~ parts'. Additionally, the components and '~parts' may be implemented to activate one or more CPUs within a device or secure multimedia card. In addition, in an embodiment, the '~parts' may include one or more processors.

[0032] In this disclosure, phrases such as "A and / or B", "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" can each include any one of the items listed together in that phrase, or all possible combinations thereof. Terms such as "first", "second", or "first" or "second" may be used merely to distinguish the corresponding component from other corresponding components and do not limit the corresponding components in any other respect (e.g., importance or order).

[0033] The terms used in the description of the present disclosure to identify connection nodes, terms referring to network entities, terms referring to messages, terms referring to interfaces between network entities, terms referring to various identification information, and the like are provided for convenience of explanation. Therefore, the present invention is not limited to the terms described below, and other terms referring to objects with equivalent technical meanings may be used.

[0034] For convenience of explanation, this disclosure uses terms and names defined in the 5GS and NR standards defined by the 3rd Generation Partnership Project (3GPP), among the existing communication standards. However, this disclosure is not limited to the above terms and names and can be equally applied to wireless communication systems that comply with other standards. This disclosure can be applied to 3GPP 5GS / NR (5th generation mobile communication standards).

[0035] In the present disclosure, a base station is an entity that performs resource allocation of a user equipment (UE), and may be at least one of an eNode B, a Node B, a BS (Base Station), a RAN (Radio Access Network), an AN (Access Network), a RAN node, a wireless access unit, a base station controller, or a node on a network. A UE may be at least one of a terminal, an MS (Mobile Station), a cellular phone, a smartphone, a computer, or a multimedia device capable of performing a communication function. In the present disclosure, a downlink (DL) refers to a wireless transmission path of a signal transmitted from a base station to a user equipment (UE), and an uplink (UL) refers to a wireless transmission path of a signal transmitted from a user equipment (UE) to a base station. In addition, although embodiments of the present disclosure are described below using an LTE (Long Term Evolution) or LTE-A system as an example, embodiments of the present disclosure may also be applied to other communication systems having a similar technical background or channel type. In addition, the embodiments of the present disclosure may be applied to other communication systems through some modifications within a scope that does not significantly deviate from the scope of the present disclosure, as judged by a person having skilled technical knowledge.

[0036] Meanwhile, the 3rd Generation Partnership Project (3GPP), which is responsible for cellular mobile communication standards, is standardizing a new core network structure called 5G Core (5GC) to facilitate the evolution of existing 4G LTE systems into 5G systems. Compared to the Evolved Packet Core (EPC), the network core for existing 4G systems, 5GC supports the following differentiated features.

[0037] 5GC introduces the Network Slice feature. As a requirement of the 5G system, 5GC can support various types of terminal types and services (e.g., enhanced Mobile Broadband (eMBB), Ultra Reliable Low Latency Communications (URLLC), massive Machine Type Communications (mMTC)). Each of these terminals / services has different requirements for the core network. For example, eMBB service requires high data rate, while URLLC service requires high stability and low latency. The technology proposed to satisfy these various service requirements is Network Slice (or network slicing).

[0038] In the present disclosure, the network technology may refer to the standard specification for 5G Real-time Media Communication Architecture of TS 26.506 defined by the International Telecommunication Union (ITU) or 3GPP, for example, and the components included in the network structure of FIGS. 1 and 2 described below may mean a physical entity, or may mean software performing an individual function or hardware combined with software. Reference symbols shown as RTC-x, etc. in the drawings represent known interfaces between NFs in a 5G core network (CN). For the convenience of the description below, some of the terms and names defined in the 3GPP standards may be used. However, the present disclosure is not limited by the above terms and names, and may be equally applied to systems conforming to other standards.

[0039] Figure 1 is a diagram showing an example configuration of a communication system for real-time communication (RTC).

[0040] In the system of FIG. 1, the RTC structure illustrates an example of a structure that supports real-time communication between an endpoint (or one peer) and another endpoint (or another peer). Here, an endpoint is an entity that can transmit, receive, or forward communications, and examples of endpoints include a terminal (UE) or a server. In other words, real-time communication between terminals, between a terminal and a server, or between a terminal and a server and between terminals can be realized using the RTC structure. Real-time communication may not mean the transmission of media data that has been generated and stored in advance, but rather a case where media data such as video and / or audio (e.g., voice calls, video calls, etc.) acquired in real time by a transmitting endpoint is processed and consumed by a receiving endpoint. In the RTC, the media data may include delay-sensitive media data.

[0041] The RTC structure of FIG. 1 may include, for example, a terminal (110), an RTC AF (application function) (120), an RTC AS (application server) (130), an RTC ASP (application service provider) (140), a NEF (network exposure function) (150), and a PCF (policy control function) (160).

[0042] In FIG. 1, the RTC AF (120) receives information about the outline of the service that the RTC ASP (140) wants to provide from the service provider RTC ASP (140), and accordingly, prepares / sets the necessary mobile communication network components and provides the service to a terminal (UE) (110) using the mobile communication network.

[0043] In FIG. 1, the RTC AF (120) may include as sub-components at least one of a network support function (NS-AF) (121) for communicating with a 5G Core Network, a configuration function (122) for setting up a service, and a provision function (123) for providing a service.

[0044] In Fig. 1, RTC AS (130) can transmit and receive data with terminal (110) and / or RTC AF (120), perform computational processing on data, and transmit the processing result to terminal (110) and / or RTC AF (120).

[0045] In FIG. 1, the RTC AS (130) may include at least one of an application-supporting web function (131) for communicating with a terminal (110) through a web interface, an interworking function (135) for providing an interworking function that activates an MNO-supported RTC session including endpoints across various MNOs (Mobile Network Operators), a transport gateway function (132) and an ICE (Interactive Connectivity Establishment) function (134) for providing STUN (Session Traversal Utilities for NAT) and TURN (Traversal Using Relays around NAT) services that facilitate NAT (Network Address Translation) and firewall traversal, a WebRTC signaling function (133) for providing a standardized signaling protocol for session setup to RTC session participants, and a media function (136) for an MNO to provide media functions to support an RTC session as subcomponents. The above STUN is a network protocol that enables the end host to discover a public IP (internet protocol) address. The PCF (150) is a network function (NF) that manages operator policy information for service provision in wireless communication systems such as 5G systems. The NEF (160) is responsible for transmitting or receiving events occurring in wireless communication systems such as 5G systems and supported functions to the outside world.

[0046] In FIG. 1, the terminal (110) may include at least one of the following components: an RTC application (111), an RTC media session handler (MSH) (112), and an RTC media access function (MAF) (113).

[0047] In Fig. 1, the RTC application (111) can access services generated and provided by the RTC ASP (140) or provided by the RTC ASP (140). The RTC MSH (112) can communicate with the RTC AF (120) to receive and select details of services provided by the RTC ASP (140) via the RTC-5 interface. The RTC MAF (113) can communicate with the RTC AS (130) to use services provided by the RTC ASP (140) via the RTC-4 interface. The RTC ASP (140) can transmit information about an overview of services to be provided to the RTC AF (120) via the RTC-1 interface.

[0048] An embodiment of the present disclosure provides at least one of the methods for real-time AI services as described below 1) to 4).

[0049] 1) How RTC ASP provides real-time AI services through mobile communication network services;

[0050] 2) A method for a mobile communication network service to prepare / set up components for real-time AI services based on a request from an RTC ASP and provide terminals with access methods to these components.

[0051] 3) A method for deciding whether to execute on the terminal or to split execution between the terminal and the server based on the requirements of the AI ​​model to be used by the terminal and the performance of the terminal itself, and negotiating this with the mobile communication network service; and

[0052] 4) Method for a terminal to initiate and receive real-time AI services through a network

[0053] FIG. 2 is a diagram illustrating an example configuration of a communication system for real-time communication (RTC) according to an embodiment of the present disclosure.

[0054] The RTC structure of Fig. 2 may include, for example, a terminal (210), an RTC AF (220), an RTC AS (230), an RTC ASP (240), an NEF (250), and a PCF (260). The basic functions of each component in the RTC structure of Fig. 2 may refer to the description of the RTC structure of Fig. 1.

[0055] The terminal (210) may include at least one of the following components: an RTC application (211) for accessing an AI service provided / generated by an RTC ASP (240), an RTC media session handler (MSH) (212) for communicating with an RTC AF (220) for receiving and selecting information on details of an AI service provided / generated by the RTC ASP (240), and an RTC media access function (MAF) (213) for communicating with an RTC AS (130) for using an AI service provided by the RTC ASP (240).

[0056] The RTC application (211) may invoke an RTC media session handler (MSH) (212) and / or an RTC media access function (MAF) (213) to support transmission and reception of RTC media data. The RTC media session handler (MSH) (212) may communicate with an RTC AF (220) to establish, control, and / or support delivery of RTC media sessions. In addition, the RTC media access function (MAF) (213) may communicate with an RTC AS (130) to access and deliver RTC media content.

[0057] The above RTC AF (220) may include at least one of a network support function (NS-AF) (221) for communicating with a 5G core network, a setting function (222) for setting up an AI service, a provisioning function (223) for providing a service, and an AI service provisioning function (224), which is a provisioning entity for initiating a real-time AI service according to the present disclosure, as a subcomponent.

[0058] The RTC AS (230) may include at least one of an application-supporting web function (231) for communicating with a terminal (210) via a web interface, a transport gateway function (232) and an ICE function (234) for providing the STUN and TURN services that facilitate NAT and firewall traversal, a WebRTC signaling function (233) for providing a standardized signaling protocol for session setup to RTC session participants, an interworking function (235) for providing an interworking function that activates an MNO-supported RTC session, and AI service entities (236, 237, 238) for providing AI sessions, providing AI models, and processing AI according to the present disclosure. A specific description of the AI ​​service entities (236, 237, 238) will be described later.

[0059] PCF (260) is a network function (NF) that manages operator policy information for service provision in wireless communication systems such as 5G systems. NEF (250) is responsible for externally transmitting or receiving events occurring in wireless communication systems such as 5G systems and supported functions.

[0060] A mobile communication network service according to an embodiment of the present disclosure can provide a real-time AI service. A real-time AI service according to an embodiment of the present disclosure refers to a service in which a terminal can perform inference or learning in real time using AI technology on specific inputs of an AI model.

[0061] The RTC structure for a mobile communication network service according to the present disclosure may include new components within the RTC AF (220) and the RTC AS (230) and may define new message formats for exchanging and transmitting control and / or data between existing and new components.

[0062] The RTC ASP (240) and RTC AF (220) according to the present disclosure can exchange AI service provision descriptions to specify the real-time AI service that the RTC ASP (240) intends to provide. The RTC AF (220) can receive the AI ​​service provision description transmitted by the RTC ASP (240) and identify the features of the real-time AI service that the mobile communication network service can provide from it.

[0063] In addition, in FIG. 2, the RTC AF (220) may receive or be requested to create a list of entities that the RTC ASP (240) has created / configured in advance to provide AI services from the RTC configuration information transmitted / provided by the RTC ASP (240). As a method for the RTC ASP (240) or RTC AF (220) to create / configure entities to provide AI services, for example, a method may be used to create an Edge Application Server (EAS) instance in a 3GPP-based edge network and specify attribute information that can indicate that the purpose is to provide AI services.

[0064] In Fig. 2, the RTC AF (220) can instantiate an AI service entity requested by the RTC ASP (240) as a media function instance or an EAS instance. The RTC structure of Fig. 2 illustrates instantiated AI service entity(ies), and the AI ​​service entities (236, 237, 238) can include at least one of an AI session provisioning entity (236), an AI model repository entity (237), and an AI processing entity (238).

[0065] The AI ​​session providing entity (236) provides the terminal (210) with information about one or more features included in the AI ​​service to be used and information about each AI model that supports the AI ​​service (or the one or more features). The terminal (210) can request and receive information about a specific feature from the AI ​​session providing entity (236) among the features list in the SAI (service access information) received by the RTC MSH (212).

[0066] The above AI model repository entity (237) is an endpoint that stores and transmits AI models, and can communicate with other endpoints such as a terminal (210) to transmit a list of stored AI model(s) to the other endpoint or provide the AI ​​model(s). The terminal (210) can request and receive an AI model list (Model manifest), which is information about a specific AI model, from the AI ​​model repository entity (237).

[0067] The above AI processing entity (238) is an endpoint that performs at least one of the processes such as AI inference, segmentation inference, and learning, and can communicate with another endpoint such as a terminal (210) to receive input media or data to be applied / input to an AI model, and return output media or data obtained as a result of AI inference, segmentation inference, and / or learning to the other endpoint. The terminal (210) can delegate all or part of the execution / processing of the AI ​​model to the AI ​​processing entity (238) of the RTC AS (230) through performance negotiation, and can receive the execution / processing result from the AI ​​processing entity (238) of the RTC AS (230) and transmit it to the terminal itself or another endpoint. For example, when the terminal (210) runs out of memory or performance when executing an AI model, the terminal (210) can identify an executable split point within the terminal (210) from the AI ​​model list (Model manifest), delegate execution / processing after the split point to the AI ​​processing entity (238), and negotiate with the AI ​​processing entity (238) to transmit the execution / processing result to the terminal itself.

[0068] In Fig. 2, the RTC ASP (240) can transmit an AI Service Provision Description (ASPD) to enable the RTC AF (220) to announce a real-time AI service to a terminal (210) connected to a mobile communication network service, so that the terminal (210) can discover the real-time AI service.

[0069] In the present disclosure, the ASPD may be transmitted in the form of a sub-message or resource among the Maf_Provisioning service API, which is an Application Programming Interface (API) through which the RTC AF (220) can receive messages from the RTC ASP (240). For example, the ASPD may be included in a provisioning session message as dependent information of content-preparation-templates, content-hosting-configuration, content-publishing-configuration, and rtc-configuration, or as separate information called ai-service-provision.

[0070] The above RTC AF (220) can transmit information about the feature(s) of the real-time AI service that the terminal (210) can receive from the ASPD transmitted by the RTC ASP (240) and the AI ​​Service Entities (236, 237, 238) to the RTC MSH (212) of the terminal (210). The AI ​​Service Provisioning entity (222, 223) within the RTC AF (220) provides the terminal (210) with information related to the AI ​​service based on the ASPD received from the RTC ASP (240).

[0071] The above ASPD may be delivered in the form of a lower-level message or resource among the Maf_SessionHandling service APIs that can be received by the RTC MSH (212) and / or the RTC AS (230). The Maf_SessionHandling service API includes the SAI (Service Access Information) API and may include ASPD as a lower-level attribute of the Service Access Information resource. The ASPD may also be provided separated into lower-level components such as an AI service features list, an AI service type, and AI model information.

[0072] In the present disclosure, the ASPD (AI Service Provision Description) may be configured as in the example in [Table 1] below.

[0073] [Table 1]

[0074]

[0075] In the above [Table 1], the AI ​​service feature list (Service features list) may provide a list of features that the terminal can select in the format of [Table 2] below. Each of the above features may include detailed information such as the example in [Table 2] below.

[0076] [Table 2]

[0077]

[0078] The AI ​​service type list in [Table 1] above includes other endpoints that the terminal can select, particularly AI processing endpoints, and can present the AI ​​service type(s) to the terminal. The AI ​​service type may include detailed information, such as the example in [Table 3] below.

[0079] [Table 3]

[0080]

[0081] In the above [Table 3], the endpoint matching criteria can be provided when one or more endpoints are included in the list. When endpoint information (e.g., A001) within the endpoint list is searched using the conditions (e.g., identifiers) specified in the endpoint matching criteria, the endpoint with the A001 identifier can be searched and identified as an endpoint providing an AI Service.

[0082] In the above [Table 1], the AI ​​model information list is information about AI models provided by the RTC ASP (240), RTC AS (230), and AI model repository entity (237). The details of the AI ​​model information list are as shown in the example in [Table 4] below.

[0083] [Table 4]

[0084]

[0085] In the above [Table 4], the AI ​​Model URN (Uniform Resource Name) includes the standard information that must be supported in order for a network entity to receive, understand, and execute an AI model. The standard identifier may be identified as a URN defined by each standard organization (e.g., 3GPP) or a URN defined by an AI model adoption service provider. The AI ​​Model Identifier is an identifier that can be received from the AI ​​Model Repository (237) or used when exchanging additional information about the AI ​​model. The AI ​​Model Repository URL list is a list of repositories that can store and provide the AI ​​model. The Supported Features list is a list of features that can be inferred or learned from the AI ​​model. The AI ​​Performance Index indicates the accuracy value achieved by the AI ​​model in the inference performance evaluation. The AI ​​Model manifest includes information about the input / output and split points of the AI ​​model. The details of the AI ​​Model manifest can be described as in the example of [Table 5] below. In addition, the Performance Requirements are performance specifications required to execute the entire AI model in real time. Performance specifications can include calculations per second and minimum video memory size. Each indicator has its own value.

[0086] [Table 5]

[0087]

[0088] In the above [Table 5], the Manifest URN is a specification identifier that defines the Manifest specification. The Input groups list is a list of groups that represent the number of cases that can be allowed at the input point of the AI ​​model. For example, when an AI model can receive text, video, or audio as input, or can receive video and audio simultaneously as input, the first group can be text, the second group can be video, the third group can be audio, and the fourth group can be video and audio. By dividing the input of the AI ​​model into groups, the user using the AI ​​model can determine which types of media, or combinations thereof, are allowable inputs, and also cannot use them arbitrarily in combination. The list of input groups is a list of group identifiers. The Split point groups list and the Output groups list also represent lists of groups that represent the number of cases that can be allowed at the split point and output point. By selecting and designating a group from the Output groups list, the terminal or server that finally executes the AI ​​model can output the results of AI inference with the media or data of the designated group. The above Group may include sub-information such as Point type, Group identifier, media, codec / profile, and performance requirements. The Point type indicates one of the input / split / output points. If the Point type is not provided, the Group identifier may be uniquely assigned to all input / split / output points. The Group identifier is an identifier for the group. The endpoints negotiating to execute the AI ​​model can use the Group identifier to determine how far each endpoint will execute and make a decision accordingly.A group has one or more Media, and each Media can be either Media or Data. For example, Video, Audio, or Data can be examples of Media. The Codec / Profile provides the standards and profile information that must be followed when generating or consuming each Media. For example, Video Media can be encoded using a codec such as HEVC (High Efficiency Video Coding) and can be understood using the same codec. Profile information can also be provided to explain which options of the codec can be used for decoding. The Performance requirements are a list of indicators that can be used to determine whether each Media has the performance to generate or consume it. For example, information such as the number of operations per second can be information that allows a terminal or server to make a judgment based on whether its own number of operations per second is high or low.

[0089] FIG. 3 is a drawing to explain the Input / Output / Split points and group identifiers of an AI model according to an embodiment of the present disclosure.

[0090] FIG. 3 illustrates an AI model (320) in which the input (310) of the AI ​​model (320) is one media (e.g., Korean speech) and the output (330) is one media (e.g., English speech). In this case, there are a total of two selectable group identifiers (340a, 340b) (Group 1, Group 2). If group 1 (340a) is presented as the input of the AI ​​processing endpoint, when Korean speech is transmitted from the terminal, the AI ​​processing endpoint may be considered to apply the Korean speech to the AI ​​model (320) to generate and output English speech.

[0091] FIG. 4 is a diagram illustrating a case where there are two choices for each of the input and output points of an AI model according to an embodiment of the present disclosure.

[0092] The example of FIG. 4 illustrates an AI model (320) in which the inputs (410a, 410b) of the AI ​​model (420) can select at least one from two media (e.g., Korean text and voice), and the outputs (430b, 430b) can select at least one from two media (e.g., English text and voice). In this case, the number of selectable group identifiers (440a, 440b, 440c, 440d) is four in total (Group 1 to Group 4). In the example of FIG. 4, if Group 1 (440a) is indicated / selected as a group identifier that identifies a group that is a unit that bundles one or more media that the terminal can transmit, the AI ​​processing endpoint can generate and output English text (430a) and / or English voice (430b) from, for example, Korean text (410a). The selection of the group identifier can be determined by the endpoint or the terminal.

[0093] FIG. 5 is a diagram showing an example of an AI model including a split point according to an embodiment of the present disclosure.

[0094] The example of FIG. 5 illustrates an AI model (520) in which the inputs (510a, 510b, 510c) of the AI ​​model (520) can be selected from three media (e.g., Korean text, voice, and text & voice), and the outputs (530b, 530b) can be selected from two media (e.g., English text, voice). In this case, there are six group identifiers (540a, 540b, 540c, 540d, 540e, 540f) in total (Group 1 to Group 6) that can be selected from among groups, which are units that bundle one or more media that the terminal can transmit. In the example of FIG. 5, if the terminal decides to perform AI inference up to the Split point (i.e., split inference), it can present / select Group identifier 4 (540d) to the AI ​​processing endpoint. The AI ​​processing endpoint can determine whether to generate / output group 5 or 6 (530a, 530b) from the intermediate data (520a) transmitted from the terminal and negotiate this with the terminal.

[0095] Negotiation for segmented inference of such AI models can be performed between the terminal (210) and the RTC AS (230) in the example of FIG. 2. The number of group identifiers of the AI ​​models in the example of FIG. 5 is an example and is not limited to the example of FIG. 5.

[0096] In an embodiment of the present disclosure, the terminal may receive an ASPD describing all real-time AI services provided by the RTC ASP from the RTC AF, or may receive the ASPD from the RTC AF in stages based on the selection and judgment of the terminal (or the RTC ASP or the RTC AF). For example, the terminal may first receive a list of Features of AI services provided from the RTC AF (e.g., the AI ​​Service features list in [Table 1] above), a step in which the terminal selects a Feature to be used from the AI ​​Service features list, a step in which the terminal receives information about an AI model providing the Feature from the RTC AF, a step in which the terminal determines whether to fully execute the AI ​​model on the terminal based on performance requirements or to partially execute the AI ​​model and delegate the remainder (i.e., split inference) to a server (i.e., RTC AS) for execution, and a step in which the terminal delegates operations related to the AI ​​model to the server (i.e., RTC AS).

[0097] In another embodiment, the terminal may receive access information of an AI Session Provisioning entity (ASPe) (corresponding to AI Service Provisioning Function (224) in FIG. 2), which is a provisioning entity for initiating a real-time AI service from the RTC AF, and may first receive a list of Features of the AI ​​service provided from the ASPe, a step in which the terminal selects a Feature of the AI ​​service it wishes to use from the AI ​​Service features list, a step in which the terminal receives information about an AI model that provides a Feature of the AI ​​service from the RTC AF, a step in which the terminal determines whether to fully execute the AI ​​model on the terminal based on performance requirements or to partially execute the AI ​​model and delegate the remainder (i.e., split inference) to a server (i.e., RTC AS) for execution, and a step in which the terminal delegates operations related to the AI ​​model to the server (i.e., RTC AS).

[0098] In order for the above RTC AF to provide ASPe access information to the terminal, at least one of the items in [Table 6] may be considered / included in the ServiceAccessInformation resource information in the MafSessionHandling service provided to the RTC MSH of the terminal. In one implementation, the three endpoints in [Table 6] may be different or the same endpoint.

[0099] [Table 6]

[0100]

[0101] The terminal may receive from the RTC AF one or more endpoint information accessible from the aiSessionProvisioningEndpoints list as in the example of [Table 6] above, and may request and receive from the RTC AF an AI Service Features list as exemplified in [Table 1] and [Table 2] above that may be provided to the terminal using a message as in the example of [Table 7] below.

[0102] [Table 7]

[0103]

[0104] From the AI ​​Service Features list above, the terminal can select one or more AI service features and identify one or more AI models that support them. To determine whether the AI ​​model can be executed on the terminal, the terminal can use a message like [Table 8] below to receive detailed information related to the AI ​​model's performance from the AI ​​Model Repository within the RTC AS.

[0105] [Table 8]

[0106]

[0107] After the terminal determines whether to execute the AI ​​service on the terminal itself or to execute it separately from another endpoint such as a server (i.e., RTC AS), the terminal specifies this as the AI ​​Model identifier of [Table 4] and the group identifier of [Table 5] and transmits it as a message to the AI ​​processing endpoint (e.g., the AI ​​processing entity). The endpoint (AI processing entity) that receives the message receives the specified AI model from the AI ​​repository endpoint (e.g., the AI ​​model repository entity) and determines whether the endpoint (AI processing entity) can execute the AI ​​model based on the specified group identifier information and AI model manifest. In [Table 9] below, if the AI ​​model is not supported by the AI ​​processing entity or is not available within the service, the AI ​​Model identifier can reply to the terminal with not supported or not available, respectively. In [Table 9] below, if the performance specifications of the endpoint (AI processing entity) are sufficient when performing segmentation inference or learning after the point specified in the group identifier, True can be returned to the terminal, otherwise False can be returned.

[0108] [Table 9]

[0109]

[0110] The RTC ASP can use mobile network services to provide real-time AI inference services to the RTC Application on the terminal. The application layer of the mobile network can include the RTC AF and the RTC AS. Conventionally, the RTC ASP can use the RTC Configuration message to instruct the RTC AF about the services it wishes to provide.

[0111] The RTC ASP and RTC AF according to the present disclosure can extend the RTC Configuration message as shown in [Table 10] below for additional configuration required for AI inference services. [Table 10] below exemplifies information included in the RTC Configuration message, and at least one of the information exemplified below can be included in the RTC Configuration message.

[0112] [Table 10]

[0113]

[0114] If aiSesionProvisioningFunctionEndpoints, aiModelRepositoryEndpoints and / or aiProcessingEndpoints are specified in the Configuration message received by the RTC AF from the RTC ASP, a list of aiSessionProvisioningFunctions, aiModelRepositoryEndpoints and aiProcessingEndpoints created by the RTC ASP through pre-negotiation may be delivered to the terminal. If the above aiSessionProvisioningFunction, aiModelRepositoryEndpoints or aiProcessingEndpoints are enabled but the list is not specified, the RTC AF instructs the RTC AS to create the aiSessionProvisioningFunction, aiModelRepositoryEndpoints and aiProcessingEndpoints, and the RTC AS can create the aiSessionProvisioningFunction, aiModelRepositoryEndpoints and aiProcessingEndpoints as an instance of the MF and then reply to the RTC AF.

[0115] In another implementation, the aiSessionProvisioningFunction, aiModelRepositoryEndpoints, and aiProcessingEndpoints described above can be configured as Edge Resources in edge computing proposed in the 3GPP standard. The RTC ASP can present an Edge Resource Configuration resource to the RTC AF for the purpose of instantiating an Edge Application Server (EAS) instance in the edge network using the Edge computing resources of the RTC AS.

[0116] The format of requirements considered for conventional EAS instantiation is as follows [Table 11].

[0117] [Table 11]

[0118]

[0119] In order to support real-time AI services in this disclosure, the following extensions may be considered for easType or easFeatures in [Table 11].

[0120] When aiSessionProvisioning is specified for the above easType or easFeatures, an EAS instance that serves as AI service provisioning is instantiated. RTC AF can provide ASPD information to the created instance.

[0121] When aiModelRepository is specified for the above easType or easFeatures, an EAS instance that acts as an AI model repository is instantiated. The RTC AF can instruct the created instance to receive the AI ​​Model.

[0122] When aiProcessing is specified for the above easType or easFeatures, an EAS instance that acts as an AI processing engine is instantiated.

[0123] The terminal can search for other endpoints by listing only aiProcessingEndpoints in the eas or processing items, and send the message in [Table 9] above to the searched endpoints to request the initiation of real-time AI services based on the results.

[0124] This disclosure describes a negotiation method based on the Simple WebRTC Application Protocol (SWAP) message. The SWAP protocol operates through a reliable, full-duplex WebSocket connection between two endpoints or between an endpoint and a SWAP server.

[0125] In one implementation, the terminal may receive SWAP server endpoint information upon receiving SAI information, and may receive an AI processing endpoint by specifying conditions for communicating with the SWAP server to search for an entity for an AI service.

[0126] FIG. 6 is a flowchart illustrating a procedure for a terminal to obtain / negotiate information about an entity for an AI service by communicating with an RTC AS according to an embodiment of the present disclosure. A basic description of the network entities in the embodiment of FIG. 6 may refer to the description of the corresponding network entities in the RTC structure of FIG. 2 . The SWAP server function in FIG. 6 may be performed by the WebRTC signaling function (233) in FIG. 2 .

[0127] Referring to FIG. 6, Service Provision for AI service is initiated at step 601.

[0128] In step 602, the RTC ASP may transmit service access information (SAI), including ASPD, which is information about the real-time AI service to be provided, to the RTC AF.

[0129] At steps 603 and 604, the RTC AF may create or instantiate the SWAP server and AI processing endpoints described in the SAI.

[0130] In step 605, the RTC AF may provide an SAI to the terminal upon request from the terminal. The SAI may include the address of the SWAP server.

[0131] The registration phase is initiated at step 606.

[0132] In step 607, the terminal registers itself with the SWAP server. All endpoints, including the terminal, can register themselves with the SWAP server upon receiving information from the SWAP server for RTC communication. The SWAP server's register message provides a matching_criteria parameter, and at least one of the keywords in [Table 12] can be used as the matching criteria:

[0133] [Table 12]

[0134]

[0135] Terminals and servers that wish to use or provide RTC-based real-time AI services to other endpoints can specify rtc_ai_service under service. The app is described by the RTC ASP as information that uniquely identifies its RTC application.

[0136] In step 608, the AI ​​processing endpoint also registers itself with the SWAP server. For instances created within the RTC AS, sub-items of the Matching criteria, such as the eas item (see [Table 12]) or the processing item, may include the aiSessionProvisioningFunction, aiModelRepositoryEndpoints, and aiProcessingEndpoints described in [Table 10].

[0137] At step 609, negotiation is performed between the terminal and the AI ​​processing endpoint.

[0138] In step 610, a target endpoint for exchanging real-time AI services is searched. The terminal searches for a target server or target terminal to initiate a service type such as terminal-terminal or terminal-server-terminal. To search for a target endpoint that provides AI processing services, the terminal creates an App-specific message (type 1) and transmits it to the SWAP server. At this time, the App-specific message may include at least one of configuration information, matching criteria, and a list of group identifiers (i.e., information about an AI model's input point group, split point group, or output point group).

[0139] The above configuration information (i.e., configuration-related information) may include at least one of the AI ​​service feature selected by the terminal, end-to-end connection type information, and AI model information. Depending on the connection type, for example, the terminal may search for another target terminal, search for a target server, or request communication with another target terminal using a server as an intermediary.

[0140] The above matching criteria may include performance specifications of the terminal and performance specifications that the target endpoint must provide.

[0141] The above Group Identifiers list is a list of Group identifiers that the terminal wishes to transmit. Based on the AI ​​model manifest information and the Group Identifiers, the terminal can specify part or all of the scope of the AI ​​model it wishes to implement.

[0142] In steps 611, 612, and 613, the SWAP server searches for AI processing endpoints that have registered themselves with a register message in response to an App-specific message (type 1) request from a terminal, or for AI processing endpoints that match the address and identifier in the matching criteria, match the performance requirements, or provide higher performance among other terminal endpoints, and forwards the App-specific message of the terminal to the searched AI processing endpoints and notifies the terminal of this. For example, the SWAP server only handles mediation between endpoints, the target of mediation is the registered endpoint(s), and the condition of mediation is the matching criteria. For example, assuming that terminals 1, 2, and 3 and servers 1, 2, and 3 have registered themselves on the SWAP server, and that only server 3 has registered that it has an AI processing function, and if terminal 1 requests the SWAP server to find an endpoint with the AI ​​processing function and group identifier 4 as a condition, the request from terminal 1 is transmitted to server 3, and if server 3 can receive the information described in group identifier 4 from the terminal, complete the remaining processing, and generate an output, the above 615 message can be transmitted.

[0143] The AI ​​processing endpoint that received the App-specific message in step 614 identifies output Group information that can be output based on the AI ​​model and Group identifier information.

[0144] At step 615, the AI ​​processing endpoint replies to the SWAP server with an App-specific message (type 2) containing the output Group identifier that it can provide.

[0145] In steps 616 and 617, the SWAP server transmits the App-specific message of the AI ​​processing endpoint to the terminal and notifies the AI ​​processing endpoint of this.

[0146] In step 618, the terminal determines whether it can receive an appropriate output from the received group identifier. If the group identifier is not a group identifier belonging to the output point, the terminal may repeat the operations of steps 609 to 618 to request a search for another second AI processing endpoint that uses the group identifier of the first AI processing endpoint as input.

[0147] For example, as in the embodiment of FIG. 5, when the terminal transmits an App-specific message (type 1) including one of group identifiers 1, 2, and 3 (540a, 540b, 540c), if the AI ​​processing endpoint responds with group identifiers 5 and 6 (540e, 540f), the terminal can expect full execution at the AI ​​processing endpoint; however, if the AI ​​processing endpoint responds with group identifier 4 (540d), the terminal can repeat the operations of steps 609 to 618 to request the SWAP server for another second AI processing endpoint that can generate group identifiers 5 and 6 (540e, 540f) from group identifier 4 (540d). The second AI processing endpoint can be an RTC MAF in the terminal or another AI processing endpoint in the network.

[0148] At step 619, a session is established between the terminal and the AI ​​processing endpoint.

[0149] In steps 620 and 621, the terminal generates an SDP (Session Description Protocol) offer for transmitting media and data belonging to the group identifier specified in the App-specific message (type 1) and transmits it to the SWAP server, and the SWAP server transmits it to the AI ​​processing endpoint and notifies the terminal that it has been transmitted.

[0150] In step 622, the AI ​​processing endpoint generates an SDP answer for transmitting media and data belonging to the group identifier specified in the App-specific message (type 2) and transmits it to the SWAP server. The SWAP server then forwards it to the terminal and notifies the AI ​​processing endpoint that the SDP answer has been delivered. The App-specific message (type 2) may include a list of group identifiers.

[0151] FIG. 7 is a diagram showing the configuration of a network entity in a communication system of an RTC structure according to one embodiment of the present disclosure.

[0152] The network entity of FIG. 7 may be one of the network entities described in the embodiments of FIGS. 2 to 6.

[0153] As illustrated in FIG. 7, the network entity may include a processor (701), a transceiver (703), and a memory (705). The processor (701), the transceiver (703), and the memory (705) of the network entity may operate according to the communication method of the network entity described above in the embodiments of FIGS. 2 to 6 . However, the components of the network entity are not limited to the examples described above. For example, the network entity may include more or fewer components than the components described above. In addition, the processor (701), the transceiver (703), and the memory (705) may be implemented in the form of a single chip.

[0154] The transceiver (703) is a general term for a receiver of a network entity and a transmitter of the network entity, and can transmit and receive signals with a terminal or another network entity. At this time, the transmitted and received signal may include at least one of control information and data. To this end, the transceiver (703) may include a wired / wireless transceiver and may include various configurations for transmitting and receiving signals. In addition, the transceiver (703) may receive a signal through a predetermined communication interface, output it to the processor (701), and transmit the signal output from the processor (701). In addition, when the network entity of FIG. 7 is a terminal, the transceiver (703) may include an RF transmitter that up-converts and amplifies the frequency of a transmitted signal, and an RF receiver that low-noise amplifies and frequency-converts a received signal. In addition, the transceiver (703) can receive a communication signal and output it to the processor (701), and transmit the signal output from the processor (701) to a terminal or another network entity through a network. The memory (705) can store programs and data necessary for the operation of the network entity according to at least one of the embodiments of FIGS. 2 to 6. In addition, the memory (705) can store control information or data included in a signal obtained from the network entity. The memory (705) can be configured as a storage medium or a combination of storage media, such as a ROM, a RAM, a hard disk, a CD-ROM, and a DVD.

[0155] The processor (701) may control a series of processes so that a network entity can operate according to at least one of the embodiments of FIGS. 2 to 6. The processor (701) may include at least one processor. The methods according to the embodiments described in the claims or specification of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software. In the case of software implementation, a computer-readable storage medium storing one or more programs (software modules) may be provided. The one or more programs stored in the computer-readable storage medium are configured for execution by one or more processors in an electronic device. The one or more programs include instructions that cause the electronic device to execute the methods according to the embodiments described in the claims or specification of the present disclosure.

[0156] These programs (software modules, software) may be stored in random access memory, non-volatile memory including flash memory, read only memory (ROM), electrically erasable programmable read only memory (EEPROM), magnetic disc storage devices, compact disc-ROMs (CD-ROMs), digital versatile discs (DVDs) or other forms of optical storage devices, magnetic cassettes, or may be stored in memories formed by a combination of some or all of these. In addition, each configuration memory may include multiple copies. The above program may be stored on an attachable storage device that is accessible via a communication network such as the Internet, an intranet, a local area network (LAN), a wide area network (WAN), a storage area network (SAN), or a combination thereof. This storage device may be connected to a device performing an embodiment of the present disclosure via an external port. Additionally, a separate storage device on the communication network may be connected to a device performing an embodiment of the present disclosure.

[0157] In the specific embodiments of the present disclosure described above, components included in the disclosure are expressed in the singular or plural form, depending on the specific embodiment presented. However, the singular or plural expressions are selected to suit the presented situation for convenience of explanation, and the present disclosure is not limited to singular or plural components. Components expressed in the plural form may be composed of singular elements, or components expressed in the singular form may be composed of plural elements.

[0158] While the detailed description of this disclosure has described specific embodiments, it should be understood that various modifications are possible without departing from the scope of this disclosure. Therefore, the scope of this disclosure should not be limited to the described embodiments, but should be defined not only by the scope of the claims described below, but also by equivalents thereof.

Claims

1. In a method performed by a terminal in a communication system based on RTC (real time communication), A process of receiving ASPD (AI service provision description) for real-time AI (artificial intelligence) service provided from RTC ASP (application service provider) through RTC AF (application function); A process of registering information about the terminal in the RTC AS (application server) to use the above real-time AI service; A process of performing negotiation to search for an AI processing end point for transmitting and receiving media data of the RTC AS and the real-time AI service; and A method comprising establishing a session for the above-described AI processing endpoint and the above-described real-time AI service.

2. In paragraph 1, The process of performing the above negotiation includes: a process of transmitting a first message including input group information of the AI ​​model to the SWAP (simple WebRTC application protocol) server within the RTC AS for searching the AI ​​processing endpoint; and A method further comprising the step of receiving a second message including output group information of the AI ​​model through the SWAP server from the searched AI processing endpoint.

3. In paragraph 2, A method wherein the input group information and the output group information each include at least one of language-specific text and voice information that can be selected as input or output in the AI ​​model.

4. In paragraph 2, The first message above is, A method further comprising at least one of setting-related information including at least one of characteristic information of the AI ​​service selected by the terminal, end-to-end connection type information, and information about the AI ​​model, and matching criteria information used to search for the AI ​​processing endpoint, including at least one of performance specifications of the terminal and information about performance specifications provided by the target endpoint.

5. In paragraph 1, A method in which the process of performing the above negotiation further includes a process in which the terminal determines whether to execute split inference of the AI ​​model between the terminal and the RTC AS based on the requirements of the AI ​​model and the performance of the terminal.

6. In the terminal (user equipment: UE), At least one transmitter / receiver; At least one processor communicatively coupled to said at least one transceiver; and At least one memory communicatively coupled to said at least one processor and storing instructions executable individually or in combination by said at least one processor, said instructions causing said terminal to: Receives an ASPD (AI service provision description) for a real-time AI (artificial intelligence) service provided from an RTC (real time communication) ASP (application service provider) through at least one transceiver via an RTC AF (application function), Through at least one of the above transmitters and receivers, register information about the terminal to the RTC AS (application server) for use of the real-time AI service, Through the at least one transceiver, negotiation is performed to search for an AI processing end point for transmitting and receiving media data of the RTC AS and the real-time AI service, and A terminal that causes a session to be established for the real-time AI service with the searched AI processing endpoint through at least one transceiver.

7. In paragraph 6, The instructions executable individually or in combination by the at least one processor cause the terminal to: transmit, through the at least one transceiver, a first message including input group information of the AI ​​model for searching the AI ​​processing endpoint to a SWAP (simple WebRTC application protocol) server within the RTC AS; A terminal further causing the terminal to receive a second message including output group information of the AI ​​model through the SWAP server from the searched AI processing endpoint via the at least one transceiver.

8. In paragraph 7, A terminal including at least one of the language-specific text and voice information that can be selected as input or output in the AI ​​model, respectively, and the input group information and the output group information.

9. In paragraph 7, The first message above is, Setting-related information including at least one of characteristic information of the AI ​​service selected by the terminal, end-to-end connection type information, and information about the AI ​​model; A terminal further comprising at least one of the matching criteria information used to search for the AI ​​processing endpoint, the matching criteria information including at least one of the performance specifications of the terminal and the performance specifications that the target endpoint must provide.

10. In paragraph 6, The instructions executable individually or in combination by the at least one processor are configured to cause the terminal to: A terminal that further causes the terminal to decide whether to execute split inference of the AI ​​model between the terminal and the RTC AS based on the requirements of the AI ​​model and the performance of the terminal while performing the above negotiation.

11. In a communication system based on RTC (real time communication), in the RTC AS (application server), At least one transmitter / receiver; At least one processor communicatively coupled to said at least one transceiver; and At least one memory communicatively coupled to said at least one processor and storing instructions executable individually or in combination by said at least one processor, said instructions causing said RTC AS to: Through at least one of the above transmitters and receivers, information about the terminal is received from the terminal for using the real-time AI service, Through at least one of the above transceivers, negotiation is performed to search for an AI processing end point for transmitting and receiving media data of the terminal and the real-time AI service, and An RTC AS that causes signaling to be performed for session establishment for the real-time AI service between the terminal and the searched AI processing endpoint.

12. In paragraph 11, The instructions executable individually or in combination by the at least one processor are: Through the at least one transceiver, the SWAP (simple WebRTC application protocol) server within the RTC AS receives a first message including input group information of the AI ​​model for searching the AI ​​processing endpoint from the terminal, and An RTC AS that causes the discovered AI processing endpoint within the RTC AS to transmit a second message containing output group information of the AI ​​model through the SWAP server via the at least one transceiver.

13. In paragraph 12, The above input group information and the above output group information are RTC ASs that each include at least one of language-specific text and voice information that can be selected as input or output in the AI ​​model.

14. In paragraph 12, The first message above is, An RTC AS further comprising at least one of setting-related information including at least one of characteristic information of the AI ​​service selected by the terminal, end-to-end connection type information, and information about the AI ​​model, and matching criteria information used to search for the AI ​​processing endpoint, including at least one of performance specifications of the terminal and information about performance specifications provided by the target endpoint.

15. In paragraph 11, Whether to execute the split inference of the AI ​​model between the RTC AS and the terminal is determined by the RTC AS based on the requirements of the AI ​​model and the performance of the terminal.

Images