ABSTRACT
Abstract
In the present disclosure, methods of operating a translation node in a communication network are discussed. The translation node receives Application service information for an application service, a distributed Artificial Intelligence AI model for the application service, and Model Deployment Map MDM information for the application service, translates the MDM information for the application service into network Quality of Service QoS parameters for the application service, and provides the distributed AI model with the network QoS parameters for distribution to at least one other node of the communication network. Related methods of operating SMF and NDWAF nodes are also discussed.
Description
TECHNICAL FIELD
The present disclosure relates generally to communications, and more particularly to communication methods and related devices and nodes supporting wireless communications.
BACKGROUND
Distributed Artificial Intelligence D-AI is discussed below.
In the evolution of 5th Generation (5G) and beyond mobile networks, Artificial Intelligence AI applications with heavy computation resource requirements are involved in all phases of the communications. Under the constrain of limited computation resources, and battery lifetime of UEs as well as the overhead of transferring all the data to the machine learning ML server, distributed AI (D-AI) is brought up as having parts of the algorithm deployed among the communication network and performing computations as the communication proceeds.
On the other hand, with privacy concerns, it may become an ethical and/or legal issue when data centers are allowed to retrieve the user experience related data from a user's private device for commercial use. Federated learning FL has been brought up to address this issue by transferring weight(s) of the trained model, other than data, to protect users from privacy leakage. Use of Federated Learning FL, however, may still require every worker to have the capability to perform the full-version of the Machine Learning ML model.
FIG. 1 (from FIG. 4 of Reference [1]) illustrates an Algorithmic View of an example of a Convolution-DDNN deployment. In FIG. 1 , the Fully Connect (FC) layers and Convolution (ConvP) layers (indicated with crosshatching) are the parts that deploy on end devices (indicated with dashed line boxes). Aggregation computation is assigned on edge and cloud. A latter layer with external output is deployed on the cloud which may have fewer computation constrains (e.g., battery, availability, etc.) and more computation capability. The structure of FIG. 1 is discussed in greater detail with respect to FIG. 4 of Reference [1].
Considering the inherent distributed nature and constrained performance in every individual node, Distributed Deep Neural Networks (e.g., see FIG. 1 of Reference [1]) is one of the most commercially implemented distributed AI forms. Recent studies have contributed to having DDNN deployed in a distributed computing hierarchy, e.g., distributed between end device(s) and the cloud. For a well-trained DNN, it is feasible to migrate the neural network to a distributed system under human data scientist knowledge. The motivation is both having the AI algorithm match accuracy, communication and latency requirements, and share the inherent merit of a distributed system including fault tolerance and privacy. FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F illustrates concepts of a DDNN.
FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F (from FIG. 2 of Reference [1]) illustrate an overview of DDNN architectures. The vertical lines represent the DNN pipeline, which connects the horizontal bars (Neural Network NN layers in FIG. 1 ). FIG. 2 A illustrates a standard DNN (processed entirely in the cloud), FIG. 2 B introduces end devices and a local exit point that may classify samples before the cloud, FIG. 2 C extends FIG. 2 B by adding multiple end devices which are aggregated together for classification, FIGS. 2 D and 2 E extend FIGS. 2 B and 2 C by adding edge layers between the cloud and end devices, and FIG. 2 F shows how the edge can also be distributed like the end devices. Structures of FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F are discussed in greater detail with respect to FIG. 2 of Reference [1].
Compared with Federated Learning (FL), DDNN can be differentiated in model O&M (operation and management). Federated learning may require a local workers' training in many local agents. Then, it may be required to transfer the training outcome to an aggregation point to combine the workers' training result to form a global weight. This may require all the nodes in the federation to have full knowledge of all of the model's hyper-parameters. Yet, DDNN is a distributed deployment approach for a single neural network model, aiming to reach an optimal trade-off between data traffic volume (for transfer of input data to a data center via a network) and end-to-end inference time. The former (input) layer can have little knowledge on how its output will be processed by latter layers of the neural network. DDNN is more widely adapted when there exist constrains on deploying a deep model on the edge/device. Yet for Federated Learning FL, the worker/training agent is expected to hold the full model and even train the full models with couples of batches of data.
FIG. 3 illustrates the basic SBA (Service Based Architecture) of the core network CN in 5G. Network Functions (NFs) expose their abilities as services that can be used by other NFs. In the current 3GPP specification for the 5G core network, the 5G System architecture is defined to support data connectivity and services enabling deployments to use techniques such as Network Function Virtualization and Software Defined Networking. The 5G System architecture may leverage service-based interactions between Control Plane (CP) Network Functions which are identified in Reference [2]. For example, AMF can provide a service that enables an NF to communicate with the user equipment UE and/or the AN (Access Network) through the AMF; and SMF exposes a service that allows the consumer NFs to handle PDU sessions of UEs.
Relevant NFs for the present disclosure include PCF, SMF, NEF as well as AF and NWDAF. As defined in the current 3GPP specifications (e.g., TS23.501, cited as Reference [2]), an AF may send requests to influence SMF routing decisions for traffic of specific PDU Sessions. The AF requests may influence User Plane Function UPF (re)selections and/or allow routing user traffic of a local access to a Data Network DN. A Network Data Analytic Function NWDAF provides analytics on several network Key Performance Indicators KPIs (e.g., network node load, slice load, Quality of Service QoS, Sustainability Analytics, etc.) to different Network Function NF consumers.
If the operator does not allow an AF to access the network directly, the AF shall use the NEF to interact with the 5th Generation Core 5GC.
The AF requests are sent to the Policy Control Function PCF or via the Network Exposure Function NEF. The AF requests that target existing or future Protocol Data Unit PDU Sessions of multiple UE(s) or of any UE are sent via the NEF and may target multiple PCF(s). The PCF(s) transform(s) the AF requests into policies that apply to PDU Sessions. The AF can also request to obtain Quality of Service QoS Sustainability Analytics for specific UEs from NWDAF, where the AF can also provide as input geographical areas and time windows used to tune the generation of QoS Sustainability Analytics.
Furthermore, UEs can have multiple Internet Protocol IP addresses, e.g. IPv6 multihoming or IP addresses with different PDU anchors.
Existing network architectures, however, may not adequately support distributed AI and/or DDNN deployment.
SUMMARY
According to some embodiments of inventive concepts, a method of operating a translation node in a communication network is provided. The translation node receives application service information for an application service, a distributed Artificial Intelligence AI model for the application service, and Model Deployment Map MDM information for the application service. The translation node translates the MDM information for the application service into network Quality of Service QoS parameters for the application service. The translation node provides the distributed AI model with the network QoS parameters for distribution to at least one other node of the communication network.
According to such embodiments, by providing translation of MDM information into network QoS parameters, DDNN deployment may be more efficiently implemented across a communication network, and/or user traffic management may be more efficiently configured. Moreover, impact on legacy operations/nodes/functions may be reduced.
According to some embodiments of inventive concepts, a method of operating a core network CN node in a communication network is provided. The CN node acquires a distributed artificial intelligence AI model for a communication device, wherein the distributed AI model includes a cloud model portion and a cloud model weight, an edge model portion and an edge model weight, and a local model portion and a local model weight. The CN node transmits the cloud model portion and the cloud model weight to a user plane function UPF node of the communication network. The CN node transmits the edge model portion and the edge model weight and the local model portion and the local model weight for distribution to a radio access network RAN node associated with the communication device.
According to such embodiments, an efficient/dynamic deployment may be provided for a distributed AI approach (e.g., DDNN) to set up the user plane for distributed AI traffic. Moreover, such deployments may be provided with reduced impact on legacy operations/nodes/functions.
According to some embodiments of inventive concepts, a method of operating a core network CN node in a communication network is provided. The CN node receives a distributed artificial intelligence AI model for an application service, wherein the AI model includes network QoS parameters for the application service. The CN node reports an alarm based on the network QoS parameters for the application service.
According to such embodiments, a more efficient collection of performance information from distributed AI components may be provided. Moreover, this performance information may be used to provide feedback to the application for potential redeployment.
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this application, illustrate certain non-limiting embodiments of inventive concepts. In the drawings:
FIG. 1 is a diagram illustrating an algorithmic view of a convolution-DDNN deployment;
FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F are diagrams illustrating an overview of DDNN architectures;
FIG. 3 is a block diagram illustrating a Service Based Architecture SBA of a 5G core network;
FIG. 4 is a diagram illustrating a DDNN architecture in a 3GPP network according to some embodiments of inventive concepts;
FIGS. 5 A and 5 B provide a message diagram illustrating network operations/messages during a bootstrapping phase according to some embodiments of inventive concepts;
FIGS. 6 A, 6 B, and 6 C provide a message diagram illustrating network operations/messages during an application runtime phase according to some embodiments of inventive concepts;
FIG. 7 is a message diagram illustrating operations/messages to set up a classifier and a UPF for DDNN application data traffic according to some embodiments of inventive concepts;
FIGS. 8 A and 8 B provide a message diagram illustrating operations/messages during handover with UPF relocation for DDNN according to some embodiments of inventive concepts;
FIG. 9 is a block diagram illustrating UAV assisted automated tower inspection in a DDNN deployment according to some embodiments of inventive concepts;
FIG. 10 is a block diagram illustrating a UAV of FIG. 9 according to some embodiments of inventive concepts;
FIG. 11 is a block diagram illustrating a wireless device UE according to some embodiments of inventive concepts;
FIG. 12 is a block diagram illustrating a radio access network RAN node (e.g., a base station eNB/gNB) according to some embodiments of inventive concepts;
FIG. 13 is a block diagram illustrating a core network CN node (e.g., an AMF node, an SMF node, etc.) according to some embodiments of inventive concepts;
FIG. 14 is a flow chart illustrating operations of a translation node according to some embodiments of inventive concepts;
FIG. 15 is a flow chart illustrating operations of an SMF node according to some embodiments of inventive concepts;
FIG. 16 is a flow chart illustrating operations of an NWDAF node according to some embodiments of inventive concepts;
FIG. 17 is a block diagram of a wireless network in accordance with some embodiments;
FIG. 18 is a block diagram of a user equipment in accordance with some embodiments
FIG. 19 is a block diagram of a virtualization environment in accordance with some embodiments;
</para
TECHNICAL FIELD
The present disclosure relates generally to communications, and more particularly to communication methods and related devices and nodes supporting wireless communications.
BACKGROUND
Distributed Artificial Intelligence D-AI is discussed below.
In the evolution of 5th Generation (5G) and beyond mobile networks, Artificial Intelligence AI applications with heavy computation resource requirements are involved in all phases of the communications. Under the constrain of limited computation resources, and battery lifetime of UEs as well as the overhead of transferring all the data to the machine learning ML server, distributed AI (D-AI) is brought up as having parts of the algorithm deployed among the communication network and performing computations as the communication proceeds.
On the other hand, with privacy concerns, it may become an ethical and/or legal issue when data centers are allowed to retrieve the user experience related data from a user's private device for commercial use. Federated learning FL has been brought up to address this issue by transferring weight(s) of the trained model, other than data, to protect users from privacy leakage. Use of Federated Learning FL, however, may still require every worker to have the capability to perform the full-version of the Machine Learning ML model.
FIG. 1 (from FIG. 4 of Reference [1]) illustrates an Algorithmic View of an example of a Convolution-DDNN deployment. In FIG. 1 , the Fully Connect (FC) layers and Convolution (ConvP) layers (indicated with crosshatching) are the parts that deploy on end devices (indicated with dashed line boxes). Aggregation computation is assigned on edge and cloud. A latter layer with external output is deployed on the cloud which may have fewer computation constrains (e.g., battery, availability, etc.) and more computation capability. The structure of FIG. 1 is discussed in greater detail with respect to FIG. 4 of Reference [1].
Considering the inherent distributed nature and constrained performance in every individual node, Distributed Deep Neural Networks (e.g., see FIG. 1 of Reference [1]) is one of the most commercially implemented distributed AI forms. Recent studies have contributed to having DDNN deployed in a distributed computing hierarchy, e.g., distributed between end device(s) and the cloud. For a well-trained DNN, it is feasible to migrate the neural network to a distributed system under human data scientist knowledge. The motivation is both having the AI algorithm match accuracy, communication and latency requirements, and share the inherent merit of a distributed system including fault tolerance and privacy. FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F illustrates concepts of a DDNN.
FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F (from FIG. 2 of Reference [1]) illustrate an overview of DDNN architectures. The vertical lines represent the DNN pipeline, which connects the horizontal bars (Neural Network NN layers in FIG. 1 ). FIG. 2 A illustrates a standard DNN (processed entirely in the cloud), FIG. 2 B introduces end devices and a local exit point that may classify samples before the cloud, FIG. 2 C extends FIG. 2 B by adding multiple end devices which are aggregated together for classification, FIGS. 2 D and 2 E extend FIGS. 2 B and 2 C by adding edge layers between the cloud and end devices, and FIG. 2 F shows how the edge can also be distributed like the end devices. Structures of FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F are discussed in greater detail with respect to FIG. 2 of Reference [1].
Compared with Federated Learning (FL), DDNN can be differentiated in model O&M (operation and management). Federated learning may require a local workers' training in many local agents. Then, it may be required to transfer the training outcome to an aggregation point to combine the workers' training result to form a global weight. This may require all the nodes in the federation to have full knowledge of all of the model's hyper-parameters. Yet, DDNN is a distributed deployment approach for a single neural network model, aiming to reach an optimal trade-off between data traffic volume (for transfer of input data to a data center via a network) and end-to-end inference time. The former (input) layer can have little knowledge on how its output will be processed by latter layers of the neural network. DDNN is more widely adapted when there exist constrains on deploying a deep model on the edge/device. Yet for Federated Learning FL, the worker/training agent is expected to hold the full model and even train the full models with couples of batches of data.
FIG. 3 illustrates the basic SBA (Service Based Architecture) of the core network CN in 5G. Network Functions (NFs) expose their abilities as services that can be used by other NFs. In the current 3GPP specification for the 5G core network, the 5G System architecture is defined to support data connectivity and services enabling deployments to use techniques such as Network Function Virtualization and Software Defined Networking. The 5G System architecture may leverage service-based interactions between Control Plane (CP) Network Functions which are identified in Reference [2]. For example, AMF can provide a service that enables an NF to communicate with the user equipment UE and/or the AN (Access Network) through the AMF; and SMF exposes a service that allows the consumer NFs to handle PDU sessions of UEs.
Relevant NFs for the present disclosure include PCF, SMF, NEF as well as AF and NWDAF. As defined in the current 3GPP specifications (e.g., TS23.501, cited as Reference [2]), an AF may send requests to influence SMF routing decisions for traffic of specific PDU Sessions. The AF requests may influence User Plane Function UPF (re)selections and/or allow routing user traffic of a local access to a Data Network DN. A Network Data Analytic Function NWDAF provides analytics on several network Key Performance Indicators KPIs (e.g., network node load, slice load, Quality of Service QoS, Sustainability Analytics, etc.) to different Network Function NF consumers.
If the operator does not allow an AF to access the network directly, the AF shall use the NEF to interact with the 5th Generation Core 5GC.
The AF requests are sent to the Policy Control Function PCF or via the Network Exposure Function NEF. The AF requests that target existing or future Protocol Data Unit PDU Sessions of multiple UE(s) or of any UE are sent via the NEF and may target multiple PCF(s). The PCF(s) transform(s) the AF requests into policies that apply to PDU Sessions. The AF can also request to obtain Quality of Service QoS Sustainability Analytics for specific UEs from NWDAF, where the AF can also provide as input geographical areas and time windows used to tune the generation of QoS Sustainability Analytics.
Furthermore, UEs can have multiple Internet Protocol IP addresses, e.g. IPv6 multihoming or IP addresses with different PDU anchors.
Existing network architectures, however, may not adequately support distributed AI and/or DDNN deployment.
SUMMARY
According to some embodiments of inventive concepts, a method of operating a translation node in a communication network is provided. The translation node receives application service information for an application service, a distributed Artificial Intelligence AI model for the application service, and Model Deployment Map MDM information for the application service. The translation node translates the MDM information for the application service into network Quality of Service QoS parameters for the application service. The translation node provides the distributed AI model with the network QoS parameters for distribution to at least one other node of the communication network.
According to such embodiments, by providing translation of MDM information into network QoS parameters, DDNN deployment may be more efficiently implemented across a communication network, and/or user traffic management may be more efficiently configured. Moreover, impact on legacy operations/nodes/functions may be reduced.
According to some embodiments of inventive concepts, a method of operating a core network CN node in a communication network is provided. The CN node acquires a distributed artificial intelligence AI model for a communication device, wherein the distributed AI model includes a cloud model portion and a cloud model weight, an edge model portion and an edge model weight, and a local model portion and a local model weight. The CN node transmits the cloud model portion and the cloud model weight to a user plane function UPF node of the communication network. The CN node transmits the edge model portion and the edge model weight and the local model portion and the local model weight for distribution to a radio access network RAN node associated with the communication device.
According to such embodiments, an efficient/dynamic deployment may be provided for a distributed AI approach (e.g., DDNN) to set up the user plane for distributed AI traffic. Moreover, such deployments may be provided with reduced impact on legacy operations/nodes/functions.
According to some embodiments of inventive concepts, a method of operating a core network CN node in a communication network is provided. The CN node receives a distributed artificial intelligence AI model for an application service, wherein the AI model includes network QoS parameters for the application service. The CN node reports an alarm based on the network QoS parameters for the application service.
According to such embodiments, a more efficient collection of performance information from distributed AI components may be provided. Moreover, this performance information may be used to provide feedback to the application for potential redeployment.
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this application, illustrate certain non-limiting embodiments of inventive concepts. In the drawings:
FIG. 1 is a diagram illustrating an algorithmic view of a convolution-DDNN deployment;
FIGS. 2 A, 2 B, 2 C, 2 D, 2 E, and 2 F are diagrams illustrating an overview of DDNN architectures;
FIG. 3 is a block diagram illustrating a Service Based Architecture SBA of a 5G core network;
FIG. 4 is a diagram illustrating a DDNN architecture in a 3GPP network according to some embodiments of inventive concepts;
FIGS. 5 A and 5 B provide a message diagram illustrating network operations/messages during a bootstrapping phase according to some embodiments of inventive concepts;
FIGS. 6 A, 6 B, and 6 C provide a message diagram illustrating network operations/messages during an application runtime phase according to some embodiments of inventive concepts;
FIG. 7 is a message diagram illustrating operations/messages to set up a classifier and a UPF for DDNN application data traffic according to some embodiments of inventive concepts;
FIGS. 8 A and 8 B provide a message diagram illustrating operations/messages during handover with UPF relocation for DDNN according to some embodiments of inventive concepts;
FIG. 9 is a block diagram illustrating UAV assisted automated tower inspection in a DDNN deployment according to some embodiments of inventive concepts;
FIG. 10 is a block diagram illustrating a UAV of FIG. 9 according to some embodiments of inventive concepts;
FIG. 11 is a block diagram illustrating a wireless device UE according to some embodiments of inventive concepts;
FIG. 12 is a block diagram illustrating a radio access network RAN node (e.g., a base station eNB/gNB) according to some embodiments of inventive concepts;
FIG. 13 is a block diagram illustrating a core network CN node (e.g., an AMF node, an SMF node, etc.) according to some embodiments of inventive concepts;
FIG. 14 is a flow chart illustrating operations of a translation node according to some embodiments of inventive concepts;
FIG. 15 is a flow chart illustrating operations of an SMF node according to some embodiments of inventive concepts;
FIG. 16 is a flow chart illustrating operations of an NWDAF node according to some embodiments of inventive concepts;
FIG. 17 is a block diagram of a wireless network in accordance with some embodiments;
FIG. 18 is a block diagram of a user equipment in accordance with some embodiments
FIG. 19 is a block diagram of a virtualization environment in accordance with some embodiments;
FIG. 20 is a block diagram of a telecommunication network connected via an intermediate network to a host computer in accordance with some embodiments;
FIG. 21 is a block diagram of a host computer communicating via a base station with a user equipment over a partially wireless connection in accordance with some embodiments;
FIG. 22 is a block diagram of methods implemented in a communication system including a host computer, a base station and a user equipment in accordance with some embodiments;
FIG. 23 is a block diagram of methods implemented in a communication system including a host computer, a base station and a user equipment in accordance with some embodiments;
FIG. 24 is a block diagram of methods implemented in a communication system including a host computer, a base station and a user equipment in accordance with some embodiments; and
FIG. 25 is a block diagram of methods implemented in a communication system including a host computer, a base station and a user equipment in accordance with some embodiments.
DETAILED DESCRIPTION
Inventive concepts will now be described more fully hereinafter with reference to the accompanying drawings, in which examples of embodiments of inventive concepts are shown. Inventive concepts may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of present inventive concepts to those skilled in the art. It should also be noted that these embodiments are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present/used in another embodiment.
The following description presents various embodiments of the disclosed subject matter. These embodiments are presented as teaching examples and are not to be construed as limiting the scope of the disclosed subject matter. For example, certain details of the described embodiments may be modified, omitted, or expanded upon without departing from the scope of the described subject matter.
FIG. 11 is a block diagram illustrating elements of a communication device UE 1100 (also referred to as a mobile terminal, a mobile communication terminal, a wireless device, a wireless communication device, a wireless terminal, mobile device, a wireless communication terminal, user equipment, UE, a user equipment node/terminal/device, etc.) configured to provide wireless communication according to embodiments of inventive concepts. ( Communication device 1100 may be provided, for example, as discussed below with respect to wireless device 4110 of FIG. 17 .) As shown, communication device UE may include an antenna 1107 (e.g., corresponding to antenna 4111 of FIG. 17 ), and transceiver circuitry 1101 (also referred to as a transceiver, e.g., corresponding to interface 4114 of FIG. 17 ) including a transmitter and a receiver configured to provide uplink and downlink radio communications with a base station(s) (e.g., corresponding to network node 4160 of FIG. 17 , also referred to as a RAN node) of a radio access network. Communication device UE may also include processing circuitry 1103 (also referred to as a processor, e.g., corresponding to processing circuitry 4120 of FIG. 17 ) coupled to the transceiver circuitry, and memory circuitry 1105 (also referred to as memory, e.g., corresponding to device readable medium 4130 of FIG. 17 ) coupled to the processing circuitry. The memory circuitry 1105 may include computer readable program code that when executed by the processing circuitry 1103 causes the processing circuitry to perform operations according to embodiments disclosed herein. According to other embodiments, processing circuitry 1103 may be defined to include memory so that separate memory circuitry is not required. Communication device UE may also include an interface (such as a user interface) coupled with processing circuitry 1103 , and/or communication device UE may be incorporated in a vehicle.
As discussed herein, operations of communication device UE may be performed by processing circuitry 1103 and/or transceiver circuitry 1101 . For example, processing circuitry 1103 may control transceiver circuitry 1101 to transmit communications through transceiver circuitry 1101 over a radio interface to a radio access network node (also referred to as a base station) and/or to receive communications through transceiver circuitry 1101 from a RAN node over a radio interface. Moreover, modules may be stored in memory circuitry 1105 , and these modules may provide instructions so that when instructions of a module are executed by processing circuitry 1103 , processing circuitry 1103 performs respective operations (e.g., operations discussed below with respect to Example Embodiments relating to wireless communication devices). According to some embodiments, a communication device UE 1100 and/or an element(s)/function(s) thereof may be embodied as a virtual node/nodes and/or a virtual machine/machines.
FIG. 12 is a block diagram illustrating elements of a radio access network RAN node 1200 (also referred to as a network node, base station, eNodeB/eNB, gNodeB/gNB, etc.) of a Radio Access Network (RAN) configured to provide cellular communication according to embodiments of inventive concepts. ( RAN node 1200 may be provided, for example, as discussed below with respect to network node 4160 of FIG. 17 .) As shown, the RAN node may include transceiver circuitry 1201 (also referred to as a transceiver, e.g., corresponding to portions of interface 4190 of FIG. 17 ) including a transmitter and a receiver configured to provide uplink and downlink radio communications with mobile terminals. The RAN node may include network interface circuitry 1207 (also referred to as a network interface, e.g., corresponding to portions of interface 4190 of FIG. 17 ) configured to provide communications with other nodes (e.g., with other base stations) of the RAN and/or core network CN. The network node may also include processing circuitry 1203 (also referred to as a processor, e.g., corresponding to processing circuitry 4170 ) coupled to the transceiver circuitry, and memory circuitry 1205 (also referred to as memory, e.g., corresponding to device readable medium 4180 of FIG. 17 ) coupled to the processing circuitry. The memory circuitry 1205 may include computer readable program code that when executed by the processing circuitry 1203 causes the processing circuitry to perform operations according to embodiments disclosed herein. According to other embodiments, processing circuitry 1203 may be defined to include memory so that a separate memory circuitry is not required.
As discussed herein, operations of the RAN node may be performed by processing circuitry 1203 , network interface 1207 , and/or transceiver 1201 . For example, processing circuitry 1203 may control transceiver 1201 to transmit downlink communications through transceiver 1201 over a radio interface to one or more mobile terminals UEs and/or to receive uplink communications through transceiver 1201 from one or more mobile terminals UEs over a radio interface. Similarly, processing circuitry 1203 may control network interface 1207 to transmit communications through network interface 1207 to one or more other network nodes and/or to receive communications through network interface from one or more other network nodes. Moreover, modules may be stored in memory 1205 , and these modules may provide instructions so that when instructions of a module are executed by processing circuitry 1203 , processing circuitry 1203 performs respective operations (e.g., operations discussed below with respect to Example Embodiments relating to RAN nodes). According to some embodiments, RAN node 1200 and/or an element(s)/function(s) thereof may be embodied as a virtual node/nodes and/or a virtual machine/machines.
According to some other embodiments, a network node may be implemented as a core network CN node without a transceiver. In such embodiments, transmission to a wireless communication device UE may be initiated by the network node so that transmission to the wireless communication device UE is provided through a network node including a transceiver (e.g., through a base station or RAN node). According to embodiments where the network node is a RAN node including a transceiver, initiating transmission may include transmitting through the transceiver.
FIG. 13 is a block diagram illustrating elements of a core network CN node (e.g., an SMF node, an AMF node, etc.) of a communication network configured to provide cellular communication according to embodiments of inventive concepts. As shown, the CN node may include network interface circuitry 1307 (also referred to as a network interface) configured to provide communications with other nodes of the core network and/or the radio access network RAN. The CN node may also include a processing circuitry 1303 (also referred to as a processor) coupled to the network interface circuitry, and memory circuitry 1305 (also referred to as memory) coupled to the processing circuitry. The memory circuitry 1305 may include computer readable program code that when executed by the processing circuitry 1303 causes the processing circuitry to perform operations according to embodiments disclosed herein. According to other embodiments, processing circuitry 1303 may be defined to include memory so that a separate memory circuitry is not required.
As discussed herein, operations of the CN node may be performed by processing circuitry 1303 and/or network interface circuitry 1307 . For example, processing circuitry 1303 may control network interface circuitry 1307 to transmit communications through network interface circuitry 1307 to one or more other network nodes and/or to receive communications through network interface circuitry from one or more other network nodes. Moreover, modules may be stored in memory 1305 , and these modules may provide instructions so that when instructions of a module are executed by processing circuitry 1303 , processing circuitry 1303 performs respective operations (e.g., operations discussed below with respect to Example Embodiments relating to core network nodes). According to some embodiments, CN node 1300 and/or an element(s)/function(s) thereof may be embodied as a virtual node/nodes and/or a virtual machine/machines.
In current 3GPP communication systems, protocols within the network may not be adapted to the computation and/or communication co-proceeding for distributed-deployed AI inference components in an application bootstrapping phase of a mobile network (e.g., what signaling and method should be used to deploy different layers of a DDNN into a mobile network and/or what information should be provisioned for different components). For example, it may be undetermined how to decide which layers of a deep neural network should be deployed in which network components (including end user devices), for example, according to the device computation capabilities, latency requirements, data collection intensiveness, computation capabilities of network servers, etc.
One issue may be to address how to express DDNN deployment requirements, combined with network performance matrices, and thus how the corresponding network setup and configuration can be conducted.
Another issue may be to address how the DDNN deployment is handled by the network based on 5G network architecture, for example, including which NFs are impacted, what procedures are impacted, etc.
In the present disclosure, distributed deployment of a DNN in a 5G mobile network for inferencing is discussed. Deployment of a DDNN for learning is not a focus of the present disclosure.
Contribution S1-193606 (also referred to herein as Reference [4]) has been submitted to 3GPP SA1 with a study proposal on AI/ML Model Transfer in 5GS for upcoming Release 18, which has been approved, and related works should start during the second half of 2020. If studies are successfully completed in 3GPP SA1 during Release 18, this means that starting from Release 19 other groups in 3GPP could start working on technical solutions for AI/ML model transfer. This highlights that in 3GPP interest has started to grow in supporting AI/ML model transfer. In current discussions in 3GPP, however, the support of AI/ML model transfer is approached from the perspective of: (i) supporting AI/ML model transfer for mainly centralized AI deployments where âdistributionâ is seen from the perspective of providing an AI model to a UE upon request and not from a distributed AI approach such as DDNN; (ii) studying the use cases and potential service and performance requirements to identify traffic characteristics of AI/ML model transfer; and (iii) performing a gap analysis on performance requirements for AI/MEL model transfer (e.g., data rate, latency, reliability, coverage and capacity, etc.) for AI/ML model downloading/uploading. From this point of view, until the end of Release 19, 3GPP is not expected to cover network enhancements to support distributed AI architectures such as DDNN.
Support for distributed AI in mobile networks is believed to be a key network feature for upcoming network generations with a focus on an after Release 19 timeframe and can be considered as a native feature for 6G mobile networks.
According to some embodiments of inventive concepts, an approach to deploy DDNN in a service-based mobile network architecture is disclosed. New parameters are defined in the core network control plane signals to enable the application to influence the behavior of relevant network functions.
According to some embodiments of inventive concepts, a 5G system is considered as an example, even though inventive concepts may also target a network feature(s) that could be considered as a baseline and/or native feature of an upcoming mobile network generation(s). Considering 5G systems, the SMF and PCF will setup the user plane for the distributed AI traffic. The present disclosure also considers the NWDAF to collect the performance information from the distributed AI components and to provide feedback to the application for the potential redeployments if needed.
According to some embodiments of inventive concepts, a Model Deployment Map MDM may be provided to express the DDNN which should be used by the network to handle the DDNN deployment. Such a Model Deployment Map may also be used to configure the user traffic management when the user device is involved in the DDNN.
According to some embodiments of inventive concepts, an approach may be provided to deploy a DDNN in a service-based mobile network architecture to reduce/avoid impact with respect to legacy network procedures. Such an approach may also consider UE mobility.
According to some embodiments of inventive concepts, an approach may be provided to update a DDNN deployment dynamically according to variations of network conditions.
Provisioning of parameters from an Application Function AF to the mobile network is discussed below.
Enhancements introduced according to some embodiments of the present disclosure may apply to a mobile network whose system architecture is designed based on a service-based approach with network functions (NFs) providing services to other NFs. In the remainder of the present disclosure, for the sake of illustration, embodiments of inventive concepts are explained as being applied to a 5G system, which has a service-based architecture. It is noted that embodiments of inventive concepts may provide enhancements for distributed AI support in a mobile network, and that this may also be a useful network feature for upcoming network generations (e.g., 6G mobile networks).
According to some embodiments of inventive concepts, for either an application request for distributed deployment of a model in the network, or for the network to suggest a distributed deployment topology for an application service in the network, a blueprint of the deployment methodology is introduced, where the blueprint of the deployment methodology is defined as Model Deployment Map (MDM) and is aimed to include information to describe the AI architecture and associated requirements. Paired with service feature information, MDM may be considered to include: (i) a static part of information which describes the deployment of the DDNN in Mobile network, including deploying device and deployment template; and (ii) a dynamic part that includes the runtime DDNN model performance metrics.
A static part of the MDM may be defined during a bootstrapping phase of the application, wherein the static part of the MDM may include 2 components: UE type information and a respective deployment template. UE type information describes the deploying device of the application, which indicates a computing capability. The UE type information may include 3 categories: 3GPP managed user equipment (UE, handheld device, etc.), 3GPP managed device-to-device D2D service (UAV, UGV, etc.), and 3GPP managed Internet of Things IoT related Features (Sensor). This information indicates a primary classification of the UE's capability to provide inferencing, which could lead to different deployment templates for different UE types.
According to the 3 categories discussed above (i.e., 3GPP managed user equipment, 3GPP managed D2D service, and 3GPP managed IoT related features), a deployment template may be prepared (if the application service is needed) to provide the deployment topology respectively. One example is DDNN where its MDM's deployment template includes: number of layers on each node (e.g., a node may be a UE, a radio access network RAN node or a core network CN function), filter size/number on each node, estimated communication cost (e.g., cost based on exchanging matrix size, latency, communication lifetime, etc.) between nodes. This information may be static unless a new model is updated from the application server. For DDNN, the deployment template is different from a federated learning FL weight aggregation process, since the information indicates a fragment of the model and weight (providing faster inference), instead of a full model's weight. It is also different from a distributed deployment with ensemble technologies, since the model is trained as a whole beforehand and the deployment template is specially for one application service (providing better inference accuracy). Yet, for ensemble tech, all portions may be trained separately and may be assembled in an ad-hoc way for different application purposes.
Further, MDM may contain the model performance metrics when it is deployed in a distributed system in its dynamic part. Referring to FIGS. 2 A, 2 B, 2 C, 2 D, 2 E , and 2 F, this information includes local accuracy and inference time, edge accuracy and inference time and individual accuracy and inference time. These Key Performance Indicators KPIs are noted as: Local accuracy & inference time (LAI); Edge accuracy & inference time (EAI); Cloud accuracy & inference time (CAI); and Individual accuracy & inference time (IAI).
Local accuracy and inference time (LAI) may be provided as the mean accuracy and inference time when exiting 100% samples at the local exit of DDNN, in the UE level.
Edge accuracy and inference time (EAI) may be provided as the mean accuracy and inference time when exiting 100% samples at the edge exit of DDNN, in the cell level.
Cloud accuracy and inference time (CAI) may be provided as the mean accuracy and inference time when exiting 100% samples at the cloud exit of DDNN, in the network level.
Individual accuracy and inference time (IAI) may be provided as the mean accuracy and inference time when deploying the AI model as the MDM deployment information, for a single UE in the network.
The above listed KPIs could also be expressed in other forms in addition to mean, e.g., Xth (e.g. 90th) percentile, minimum value (for accuracy), maximum value (for inference time). In addition, it could be complemented by additional information, e.g., mean KPI plus variance or confidence interval.
Some of the above listed KPIs could also be associated with additional information for triggering adaptations such as: KPI trigger threshold; and in-advance trigger time.
For a KPI trigger threshold, the KPI value has a threshold (e.g., 85%), and in this case an adaptation is triggered when the local accuracy value crosses the threshold (e.g., 85%), to reduce/avoid that adaptation being triggered only when the KPI goes below the minimum desired value.
In-advance trigger time (e.g., 20 seconds) may be used to trigger in-advance adaptation (i.e., before the KPI crosses the associated threshold). The in-advance trigger timer parameter may be used such that an adaptation is triggered when it is predicted that within the in-advance trigger time (e.g. within 20 seconds) the KPI would cross the associated threshold.
According to some embodiments, the KPI trigger threshold and/or in-advance trigger time parameters may be applied only to a subset of above KPIs (e.g., to inference time but not to accuracy). One example is when edge inference time has an associated trigger threshold (e.g., trigger an adaptation if the performance value of edge inference time crosses the threshold of 95% of the KPI value) and potentially an associated in-advance trigger time (e.g., trigger now an adaptation because in 20 seconds it is expected that edge inference time will cross the threshold of 95% of the KPI value).
FIG. 4 illustrates an example of a DDNN Architecture in a 3GPP Network (service-based system architecture) according to some embodiments of inventive concepts.
The MDM could be generated by an application server AS and provided to the network via an AF. According to some embodiments, the application server may interact with a NEF via an Application Function (AF). Multiple application servers could share the same AF. Furthermore, MDM generation could be done within the AF. The MDM could be considered to be used by the network as an input to build a Service Level Agreement SLA or a network slice template for distributed AI architectures which is then enforced to functions of the core network CN. The MDM could be also considered by the network as an âintentâ used by the application server to ask for a particular type of distributed AI deployment.
For a specific AI model, the above information could be dynamic and will be continuously updated for further deployment modification when the network environment changes. If any KPIs (or aggregated KPIs) are degraded below the thresholds, the network could trigger a re-organization of the deployment topology, traffic priorities, QoS management, computation capabilities, etc. This means that, in the present disclosure, the network (e.g., NWDAF) may be extended to provide additional analytics with respect to those standardized in 5G, i.e., the new analytics are Local accuracy and inference time (LAI), Edge accuracy and inference time (EAI), Cloud accuracy and inference time (CAI), Individual accuracy and inference time (IAI), potentially with associated KPI trigger threshold and in-advance trigger time. And request to update the static part of MDM. Detailed sequence/message diagrams are discussed below with respect to FIGS. 5 A and 5 B and FIGS. 6 A and 6 B (where FIGS. 5 A and 5 B , and FIGS. 6 A, 6 B, and 6 C show an application server interacting with a mobile core network CN, and where such interaction could happen via an AF to which the application server is associated).
FIGS. 5 A and 5 B provide a message/sequence Diagram as applied, for example, to a 5G network for an Application Server Bootstrapping Phase. In FIGS. 5 A and 5 B , each of the NWDAF, PCF, UPF, and NEF may be respectively provided as a core network node including a network interface 1307 , a processor 1303 , and memory 1305 as discussed above with respect to core network node 1300 of FIG. 13 , such that communications between two different core network nodes may be provided through respective network interfaces.
In the bootstrapping phase of FIGS. 5 A and 5 B , the network should expose the network features (e.g., RAN deployment situations in some geographic area, edge server deployments, their computing abilities, etc.) to the application server AS of machine learning ML at operations
501 and 502 , through the network exposure function NEF. The application server can then use that information to choose good deployment options and derive initial MDM information. Once the MDM has been prepared, the application server could provide it to the NEF at operation 503 , and the NEF can store the MDM information and Application Service Information at operation 504 . At operation 505 , the NEF can then translate the MDM parameters into 3gpp QoS parameters and provide the translated version of MDM to the PCF via NEF translation at operation 506 . In such embodiments, the NEF may act as a translation node in embodiments where the network operator does not allow access to the PCF directly by an Application Server. In some other embodiments where
CLAIMS
Claims ( 22 )
1 . A method of operating a translation node in a communication network, the method comprising:
receiving application service information for an application service, a distributed Artificial Intelligence, AI, model for the application service, and Model Deployment Map, MDM, information for the application service; translating the MDM information for the application service into network Quality of Service, QoS, parameters for the application service; and providing the distributed AI model with the network QoS parameters for distribution to at least one other node of the communication network.
2 . The method of claim 1 , wherein the network QoS parameters for the application service include at least one of an individual accuracy and inference time, IAI, for the AI model, a local accuracy and inference time, LAI, for the AI model, an edge accuracy and inference time, EAI, for the AI model, and/or a cloud accuracy and inference time, CAI, for the AI model.
3 . The method of claim 2 , wherein the network QoS parameters further include a trigger threshold associated with at least one of the IAI, LAI, EAI, and/or CAI, wherein the trigger threshold defines a value of at least one of the IAI, LAI, EAI, and/or CAI that is used to trigger an adaptation of the AI model.
4 . The method of claim 3 , wherein the network QoS parameters further include an in-advance trigger time, wherein the in-advance trigger time is used to trigger an adaptation of the AI model in advance of the at least one of the IAI, LAI, EAI, and/or CAI satisfying the trigger threshold.
5 . The method of claim 1 , wherein providing the distributed AI model with the network QoS parameters comprises transmitting the distributed AI model with the network QoS parameters to a policy control function, PCF, node of the communication network.
6 . The method of claim 1 , wherein the application service information includes at least one of an input data size for the application service and/or a potential bandwidth requirement for the application service.
7 . The method of claim 1 , wherein the application service is a first application service, wherein the application service information is first application service information, wherein the AI model is a first AI model, and wherein the MDM information is first MDM information, the method further comprising:
storing the first MDM information for the first application service in association with the first application service information for the first application service; receiving second application service information for a second application service, a second distributed AI model for the second application service, and second MDM information for the second application service; responsive to a similarity between the first application service information and the second application service information, aligning the network QoS parameters for the first application service with the second application service; and providing the second distributed AI model with the network QoS parameters for distribution to the at least one other node of the communication network.
8 . The method of claim 7 , wherein providing the first distributed AI model with the network QoS parameters comprises transmitting the first distributed AI model with the network QoS parameters to a policy control function, PCF, node of the communication network, and wherein providing the second distributed AI model with the network QoS parameters comprises transmitting the second distributed AI model with the network QoS parameters to the policy control function, PCF, node of the communication network.
9 . The method of claim 7 , wherein the first application service information indicates a first input data size for the first application service and/or a first potential bandwidth requirement for the first application service, wherein the second application service information indicates a second input data size for the second application service and/or a second potential bandwidth requirement for the second application service, and wherein the network QoS parameters for the first application are aligned with the second application service responsive to a similarity between the first and second input data sizes and/or responsive to a similarity between the first and second potential bandwidth requirements.
10 . The method of claim 1 , wherein the communication network includes a core network, and wherein the translation node is integrated in a network exposure function, NEF, node of the core network and/or in a policy control function, PCF, node of the core network.
11 . A method of operating a core network, CN, node in a communication network, the method comprising:
acquiring a distributed artificial intelligence, AI, model for a communication device, wherein the distributed AI model includes a cloud model portion and a cloud model weight, an edge model portion and an edge model weight, and a local model portion and a local model weight; transmitting the cloud model portion and the cloud model weight to a user plane function, UPF, node of the communication network; and transmitting the edge model portion and the edge model weight and the local model portion and the local model weight for distribution to a radio access network, RAN, node associated with the communication device.
12 . The method of claim 11 further comprising:
receiving a session request for a session for an AI service associated with the distributed AI model from the communication device;
wherein the distributed AI model is acquired responsive to receiving the session request from the communication device.
13 . The method of claim 12 , wherein acquiring the distributed AI model comprises transmitting a request to a policy control function, PCF, node of the communication network responsive to receiving the session request for the distributed AI model, and receiving the distributed AI model from the PCF node.
14 . The method of claim 12 , wherein the session request comprises a request to establish and/or update the session for the distributed AI model.
15 . The method of claim 12 , wherein the session for the distributed AI model comprises a protocol data unit, PDU, session for the distributed AI model.
16 . The method of claim 12 , wherein transmitting the edge model portion and the edge model weight and the local model portion and the local model weight comprises transmitting a session response for the session for the distributed AI model, wherein the session response is transmitted in response to the session request, and wherein the session response includes the edge model portion and the edge model weight and the local model portion and the local model weight.
17 . The method of claim 16 , wherein the session response is transmitted through an access and mobility function, AMF, node of the communication network to the RAN node associated with the communication node.
18 . The method of claim 17 , wherein the session request is received from the communication device through the RAN node and the AMF node.
19 . The method of claim 16 , wherein the session response includes an Internet Protocol, IP, address to be allocated to the communication device for traffic of the distributed AI model.
20 . (canceled)
21 . A method of operating a core network, CN, node in a communication network, the method comprising:
receiving a distributed artificial intelligence, AI, model for an application service, wherein the AI model includes network QoS parameters for the application service; and reporting an alarm based on the network QoS parameters for the application service.
22 .- 37 . (canceled)
US18/035,634
2020-11-26
2020-11-26
Providing distributed ai models in communication networks and related nodes/devices
Pending
US20230412513A1
( en )
Applications Claiming Priority (1)
Application Number
Priority Date
Filing Date
Title
PCT/SE2020/051130
WO2022115011A1
( en )
2020-11-26
2020-11-26
Providing distributed ai models in communication networks and related nodes/devices
Publications (1)
Publication Number
Publication Date
US20230412513A1
true
US20230412513A1 ( en )
2023-12-21
Family
ID=73699381
Family Applications (1)
Application Number
Title
Priority Date
Filing Date
US18/035,634
Pending
US20230412513A1
( en )
2020-11-26
2020-11-26
Providing distributed ai models in communication networks and related nodes/devices
Country Status (2)
Country
Link
US
( 1 )
US20230412513A1
( en )
WO
( 1 )
WO2022115011A1
( en )
Cited By (19)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20220286894A1
( en )
*
2021-03-08
2022-09-08
Zscaler, Inc.
Intelligent steering in 5G
CN115134778A
( en )
*
2022-06-28
2022-09-30
ä¸å大å¦æ·±å³ç 究é¢
A computing offloading method for Internet of Vehicles based on multi-user game and federated learning
US20230336966A1
( en )
*
2020-12-31
2023-10-19
Samsung Electronics Co., Ltd.
Method and apparatus for controlling communication of plurality of devices belonging to communication group
US20230370892A1
( en )
*
2022-05-12
2023-11-16
Verizon Patent And Licensing Inc.
Systems and methods for identifying and correcting issues associated with a wireless network
US20240114359A1
( en )
*
2022-09-30
2024-04-04
Mediatek Inc.
Ai-ml model storage in ott server and transfer through up traffic
US20250022114A1
( en )
*
2023-07-14
2025-01-16
International Business Machines Corporation
Cell site equipment inspection using edge-based image analysis
US20250126444A1
( en )
*
2022-03-31
2025-04-17
Qualcomm Incorporated
Centralized machine learning model configurations
US20250168081A1
( en )
*
2022-03-31
2025-05-22
Qualcomm Incorporated
Distributed machine learning model configurations
US20250184248A1
( en )
*
2023-12-05
2025-06-05
Nokia Solutions And Networks Oy
Methods and apparatus for quantifying probability of expected network performance
WO2025167632A1
( en )
*
2024-02-08
2025-08-14
å为ææ¯æéå ¬å¸
Quality of service management method and related device
US20250310214A1
( en )
*
2021-11-12
2025-10-02
Interdigital Patent Holdings, Inc.
5g support for ai/ml communications
WO2025200831A1
( en )
*
2024-03-28
2025-10-02
å为ææ¯æéå ¬å¸
Configuration method for distributed ai task, management node, and computing node
WO2025208480A1
( en )
*
2024-04-03
2025-10-09
å京å°ç±³ç§»å¨è½¯ä»¶æéå ¬å¸
Communication method, user equipment, communication device, ott server and communication system
WO2025208478A1
( en )
*
2024-04-03
2025-10-09
å京å°ç±³ç§»å¨è½¯ä»¶æéå ¬å¸
Communication method, user equipment, communication device, ott server, and communication system
WO2025228151A1
( en )
*
2024-04-30
2025-11-06
å为ææ¯æéå ¬å¸
Communication method, apparatus, and system
WO2025247479A1
( en )
*
2024-05-28
2025-12-04
Huawei Technologies Co., Ltd.
Devices and methods for split-ai in-network computing in a mobile network
US12526669B2
( en )
*
2022-02-11
2026-01-13
Nokia Solutions And Networks Oy
Network data analytics function accuracy enhancement
WO2026031730A1
( en )
*
2024-08-09
2026-02-12
å为ææ¯æéå ¬å¸
Communication methods, communication apparatuses, and storage medium
US20260058882A1
( en )
*
2022-08-11
2026-02-26
Beijing Xiaomi Mobile Software Co., Ltd.
Communication method and apparatus
Families Citing this family (8)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
WO2024020752A1
( en )
*
2022-07-25
2024-02-01
å京å°ç±³ç§»å¨è½¯ä»¶æéå ¬å¸
Artificial intelligence (ai)-based method for providing service, apparatus, device and storage medium
WO2024030333A1
( en )
*
2022-08-01
2024-02-08
Apple Inc.
Method and apparatus for ai model definition and ai model transfer
WO2024063710A1
( en )
*
2022-09-20
2024-03-28
Telefonaktiebolaget Lm Ericsson (Publ)
Mapping of artificial intelligence-related messages
US20260067174A1
( en )
*
2022-09-27
2026-03-05
Innopeak Technology, Inc.
Wireless communication method for data transfer, device and chip
EP4615032A4
( en )
*
2022-11-06
2025-12-17
Beijing Xiaomi Mobile Software Co Ltd
COMMUNICATION METHOD AND DEVICE AS WELL AS STORAGE MEDIUM
CN118158655A
( en )
*
2022-12-07
2024-06-07
ç»´æ²ç§»å¨éä¿¡æéå ¬å¸
Transmission method, device, terminal and network side equipment
CN118283590A
( en )
*
2022-12-30
2024-07-02
å为ææ¯æéå ¬å¸
Communication method and device
CN119450413A
( en )
*
2023-08-02
2025-02-14
ç»´æ²è½¯ä»¶ææ¯æéå ¬å¸
Business processing method, device, communication equipment and readable storage medium
Citations (8)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20190318268A1
( en )
*
2018-04-13
2019-10-17
International Business Machines Corporation
Distributed machine learning at edge nodes
US20200196155A1
( en )
*
2018-12-12
2020-06-18
Verizon Patent And Licensing Inc.
Utilizing machine learning to provide closed-loop network management of a fifth generation (5g) network
US11429853B2
( en )
*
2018-11-13
2022-08-30
Gyrfalcon Technology Inc.
Systems and methods for determining an artificial intelligence model in a communication system
US11481571B2
( en )
*
2018-01-12
2022-10-25
Microsoft Technology Licensing, Llc
Automated localized machine learning training
US20230072080A1
( en )
*
2015-09-14
2023-03-09
Telefonaktiebolaget Lm Ericsson (Publ)
Radio access nodes and terminal devices in a communication network
US11734611B2
( en )
*
2019-02-26
2023-08-22
Global Elmeast Inc.
Artificial intelligence delivery edge network
US11812294B2
( en )
*
2017-07-07
2023-11-07
Telefonaktiebolaget Lm Ericsson (Publ)
Transfer of measurement configuration information in wireless communication networks
US20240172283A1
( en )
*
2021-04-08
2024-05-23
Telefonaktiebolaget Lm Ericsson (Publ)
Methods and nodes in a communications network
Family Cites Families (1)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US11108655B2
( en )
*
2018-07-06
2021-08-31
International Business Machines Corporation
Automated application deployment in a managed services domain
2020
2020-11-26
WO
PCT/SE2020/051130
patent/WO2022115011A1/en
not_active
Ceased
2020-11-26
US
US18/035,634
patent/US20230412513A1/en
active
Pending
Patent Citations (10)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20230072080A1
( en )
*
2015-09-14
2023-03-09
Telefonaktiebolaget Lm Ericsson (Publ)
Radio access nodes and terminal devices in a communication network
US11812294B2
( en )
*
2017-07-07
2023-11-07
Telefonaktiebolaget Lm Ericsson (Publ)
Transfer of measurement configuration information in wireless communication networks
US11481571B2
( en )
*
2018-01-12
2022-10-25
Microsoft Technology Licensing, Llc
Automated localized machine learning training
US20190318268A1
( en )
*
2018-04-13
2019-10-17
International Business Machines Corporation
Distributed machine learning at edge nodes
US11429853B2
( en )
*
2018-11-13
2022-08-30
Gyrfalcon Technology Inc.
Systems and methods for determining an artificial intelligence model in a communication system
US20200196155A1
( en )
*
2018-12-12
2020-06-18
Verizon Patent And Licensing Inc.
Utilizing machine learning to provide closed-loop network management of a fifth generation (5g) network
US10750371B2
( en )
*
2018-12-12
2020-08-18
Verizon Patent And Licensing, Inc.
Utilizing machine learning to provide closed-loop network management of a fifth generation (5G) network
US11734611B2
( en )
*
2019-02-26
2023-08-22
Global Elmeast Inc.
Artificial intelligence delivery edge network
US20230334369A1
( en )
*
2019-02-26
2023-10-19
Global Elmeast Inc.
Artificial intelligence delivery edge network
US20240172283A1
( en )
*
2021-04-08
2024-05-23
Telefonaktiebolaget Lm Ericsson (Publ)
Methods and nodes in a communications network
Cited By (22)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20230336966A1
( en )
*
2020-12-31
2023-10-19
Samsung Electronics Co., Ltd.
Method and apparatus for controlling communication of plurality of devices belonging to communication group
US12167273B2
( en )
*
2021-03-08
2024-12-10
Zscaler, Inc.
Intelligent steering in 5G
US20220286894A1
( en )
*
2021-03-08
2022-09-08
Zscaler, Inc.
Intelligent steering in 5G
US20250310214A1
( en )
*
2021-11-12
2025-10-02
Interdigital Patent Holdings, Inc.
5g support for ai/ml communications
US12580826B2
( en )
*
2021-11-12
2026-03-17
Interdigital Patent Holdings, Inc.
5G support for AI/ML communications
US12526669B2
( en )
*
2022-02-11
2026-01-13
Nokia Solutions And Networks Oy
Network data analytics function accuracy enhancement
US20250126444A1
( en )
*
2022-03-31
2025-04-17
Qualcomm Incorporated
Centralized machine learning model configurations
US20250168081A1
( en )
*
2022-03-31
2025-05-22
Qualcomm Incorporated
Distributed machine learning model configurations
US20230370892A1
( en )
*
2022-05-12
2023-11-16
Verizon Patent And Licensing Inc.
Systems and methods for identifying and correcting issues associated with a wireless network
US12200536B2
( en )
*
2022-05-12
2025-01-14
Verizon Patent And Licensing Inc.
Systems and methods for identifying and correcting issues associated with a wireless network
CN115134778A
( en )
*
2022-06-28
2022-09-30
ä¸å大å¦æ·±å³ç 究é¢
A computing offloading method for Internet of Vehicles based on multi-user game and federated learning
US20260058882A1
( en )
*
2022-08-11
2026-02-26
Beijing Xiaomi Mobile Software Co., Ltd.
Communication method and apparatus
US20240114359A1
( en )
*
2022-09-30
2024-04-04
Mediatek Inc.
Ai-ml model storage in ott server and transfer through up traffic
US20250022114A1
( en )
*
2023-07-14
2025-01-16
International Business Machines Corporation
Cell site equipment inspection using edge-based image analysis
US20250184248A1
( en )
*
2023-12-05
2025-06-05
Nokia Solutions And Networks Oy
Methods and apparatus for quantifying probability of expected network performance
WO2025167632A1
( en )
*
2024-02-08
2025-08-14
å为ææ¯æéå ¬å¸
Quality of service management method and related device
WO2025200831A1
( en )
*
2024-03-28
2025-10-02
å为ææ¯æéå ¬å¸
Configuration method for distributed ai task, management node, and computing node
WO2025208478A1
( en )
*
2024-04-03
2025-10-09
å京å°ç±³ç§»å¨è½¯ä»¶æéå ¬å¸
Communication method, user equipment, communication device, ott server, and communication system
WO2025208480A1
( en )
*
2024-04-03
2025-10-09
å京å°ç±³ç§»å¨è½¯ä»¶æéå ¬å¸
Communication method, user equipment, communication device, ott server and communication system
WO2025228151A1
( en )
*
2024-04-30
2025-11-06
å为ææ¯æéå ¬å¸
Communication method, apparatus, and system
WO2025247479A1
( en )
*
2024-05-28
2025-12-04
Huawei Technologies Co., Ltd.
Devices and methods for split-ai in-network computing in a mobile network
WO2026031730A1
( en )
*
2024-08-09
2026-02-12
å为ææ¯æéå ¬å¸
Communication methods, communication apparatuses, and storage medium
Also Published As
Publication number
Publication date
WO2022115011A1
( en )
2022-06-02
Similar Documents
Publication
Publication Date
Title
WO2022115011A1
( en )
2022-06-02
Providing distributed ai models in communication networks and related nodes/devices
US12267737B2
( en )
2025-04-01
Report NSA/SA NR indicator
US11265801B2
( en )
2022-03-01
Information exchange for initial user equipment access
US20240406772A1
( en )
2024-12-05
Measurement Reporting Timer
US20210075496A1
( en )
2021-03-11
Method for enabling new radio (nr) integrated access and backhaul (iab) nodes to operate in non-standalone (nsa) cells
US11930383B2
( en )
2024-03-12
Methods and apparatus for categorising wireless devices
EP4329342B1
( en )
2025-05-28
Tele-operated driving event prediction, adaption and trigger
KR102339059B1
( en )
2021-12-13
Efficient PLMN encoding for 5G
EP3662599B1
( en )
2020-09-30
Avoiding multiple retransmissions of signalling transported by 5g nas transport
US12408087B2
( en )
2025-09-02
Methods supporting a capability to modify session traffic in response to a handover and related network nodes
WO2021037709A1
( en )
2021-03-04
V2x application enabler for tele-operated driving
US12095623B2
( en )
2024-09-17
Performance-aware system and method for adaptable service mesh data plane
US20220158973A1
( en )
2022-05-19
Data network name (dnn) manipulation
WO2020167187A1
( en )
2020-08-20
Enhancements to mdt
US12348977B2
( en )
2025-07-01
Methods providing generation and deployment of network configurations and related management nodes, computer programs, and computer program products
US20210044999A1
( en )
2021-02-11
Minimization of Driving Test Measurements
US11792693B2
( en )
2023-10-17
Methods and apparatuses for redirecting users of multimedia priority services
US20210359905A1
( en )
2021-11-18
Network function upgrade method, system and apparatus
WO2021165977A1
( en )
2021-08-26
Task planning-driven adaptive slices
Legal Events
Date
Code
Title
Description
2023-05-05
AS
Assignment
Owner name : TELEFONAKTIEBOLAGET ERICSSON LM (PUBL), SWEDEN
Free format text : ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:JIN, YIFEI;FU, ZHANG;CONDOLUCI, MASSIMO;SIGNING DATES FROM 20201127 TO 20201218;REEL/FRAME:063552/0628
2023-08-21
AS
Assignment
Owner name : TELEFONAKTIEBOLAGET LM ERICSSON (PUBL), SWEDEN
Free format text : CORRECTIVE ASSIGNMENT TO CORRECT THE NAME OF THE RECEIVING PARTY IS TELEFONAKTIEBOLAGET LM ERICSSON (PUBL) PREVIOUSLY RECORDED ON REEL 063552 FRAME 0628. ASSIGNOR(S) HEREBY CONFIRMS THE ASSIGNMENT;ASSIGNORS:JIN, YIFEI;FU, ZHANG;CONDOLUCI, MASSIMO;SIGNING DATES FROM 20201127 TO 20201218;REEL/FRAME:064660/0417
2023-09-20
STPP
Information on status: patent application and granting procedure in general
Free format text : DOCKETED NEW CASE - READY FOR EXAMINATION
2025-06-09
STPP
Information on status: patent application and granting procedure in general
Free format text : NON FINAL ACTION MAILED
2025-08-17
STPP
Information on status: patent application and granting procedure in general
Free format text : RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER
2025-09-09
STPP
Information on status: patent application and granting procedure in general
Free format text : NON FINAL ACTION COUNTED, NOT YET MAILED
2025-09-11
STPP
Information on status: patent application and granting procedure in general
Free format text : NON FINAL ACTION MAILED
2025-12-10
STPP
Information on status: patent application and granting procedure in general
Free format text : RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER
Free format text : FINAL REJECTION COUNTED, NOT YET MAILED
2025-12-17
STPP
Information on status: patent application and granting procedure in general
Free format text : RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER
2026-02-21
STPP
Information on status: patent application and granting procedure in general
Free format text : FINAL REJECTION COUNTED, NOT YET MAILED
2026-02-24
STPP
Information on status: patent application and granting procedure in general
Free format text : FINAL REJECTION MAILED
2026-04-25
STPP
Information on status: patent application and granting procedure in general
Free format text : ADVISORY ACTION COUNTED, NOT YET MAILED
2026-04-29
STPP
Information on status: patent application and granting procedure in general
Free format text : ADVISORY ACTION MAILED