CN116346206A - AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS - Google Patents
AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS Download PDFInfo
- Publication number
- CN116346206A CN116346206A CN202310312924.4A CN202310312924A CN116346206A CN 116346206 A CN116346206 A CN 116346206A CN 202310312924 A CN202310312924 A CN 202310312924A CN 116346206 A CN116346206 A CN 116346206A
- Authority
- CN
- China
- Prior art keywords
- model
- transmission mode
- training data
- data
- satellite
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
- H04B7/18513—Transmission in a satellite or space-based system
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
- H04B7/18519—Operations control, administration or maintenance
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Astronomy & Astrophysics (AREA)
- Aviation & Aerospace Engineering (AREA)
- General Physics & Mathematics (AREA)
- Radio Relay Systems (AREA)
Abstract
The invention discloses an AI/ML model distributed transmission method, device and system based on a low orbit satellite and 5GS, wherein the AI/ML model is split into model data and training data, the model data is divided into delay sensitive components and set as a ground transmission mode, the training data is divided into delay tolerant components and set as a satellite transmission mode, and distributed transmission is carried out; when the AI/ML model is trained, the training data request is sent to the corresponding UE in a satellite transmission mode, and the training data fed back by the UE is transmitted back in the satellite transmission mode so as to be used for the AI/ML model training by the AI/ML system; and when the AI/ML model is used, feeding back the model data of the trained AI/ML model to the UE in a ground transmission mode. The invention can greatly save the ground network resources through the distributed transmission, so that the ground network resources can serve delay-sensitive business more, and the service experience of the terminal is improved.
Description
Technical Field
The invention relates to the field of 5G communication, in particular to data transmission in the field of 5G communication.
Background
With the development of communication networks, the kinds of services that the internet can provide to users are also rapidly increasing, for example: positioning, navigation, speech recognition, picture recognition, etc., while most of these services are based on advanced technologies such as artificial intelligence and machine learning (AI/ML), which are powerful but at the same time require more resource support. Such as greater memory space, more computing resources, shorter transmission delays, etc. In the face of various resource demands, the resources of the ground network are often insufficient to support, and the satellite access network is used as the supplement of the ground network, so that additional resource supplement can be provided for various intelligent services. However, since satellite access is used for transmission, the time delay is larger than that of terrestrial transmission, so that consideration needs to be given to how to reasonably utilize the terrestrial and satellite networks, and the terrestrial bandwidth of the 5G system is saved.
Therefore, a solution to the above problems is urgently needed.
Disclosure of Invention
The invention aims to provide an AI/ML model distributed transmission method device, a system and a computing device based on a low-orbit satellite and 5GS, which can greatly save ground network resources, enable the ground network resources to serve delay-sensitive business more and improve the service experience of a terminal.
In order to achieve the above purpose, the invention discloses an AI/ML model distributed transmission method based on a low orbit satellite and 5GS, which comprises the steps of dividing an AI/ML model into model data and training data, dividing the model data into delay sensitive components and setting the delay sensitive components as a ground transmission mode, and dividing the training data into delay tolerant components and setting the delay tolerant components as a satellite transmission mode; when an AI/ML model is trained, when a training data request sent by an AI/ML system is received, the training data request is sent to a corresponding UE in a satellite transmission mode, so that the UE collects corresponding training data, the training data is fed back in the satellite transmission mode, and the training data sent by the UE is sent to the AI/ML system for the AI/ML system to train the AI/ML model; when the AI/ML model is used, when a model subscription request sent by the UE is received, model data of the trained AI/ML model is fed back to the UE in a ground transmission mode.
Preferably, the NWDAF network element receives the training data request sent by the AI/ML system, and sends the training data request to the corresponding UE in a satellite transmission mode, and the NWDAF network element receives the training data fed back by the UE in a satellite transmission mode; and receiving and storing the model data of the AI/ML model trained by the AI/ML system through an edge server, receiving a model subscription request sent by the UE through an NWDAF network element, forwarding the model subscription request to the edge server, receiving the model data of the trained AI/ML model sent by the edge server through the NWDAF network element, and feeding back the model data of the trained AI/ML model to the UE in a ground transmission mode.
Specifically, the NWDAF network element sets a transmission mode of the training data request and the training data as a satellite transmission mode and stores the data, and the NWDAF network element sets a transmission mode of the model subscription request and the AI/ML model as a ground transmission mode and stores the data.
The invention also discloses a computing device comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is configured to store at least one executable instruction that causes the processor to perform operations corresponding to the low-orbit satellite and 5GS based AI/ML model distributed transmission method described above.
The invention also discloses a computer storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute the operation corresponding to the AI/ML model distributed transmission method based on the low-orbit satellite and 5 GS.
The invention also discloses an AI/ML model distributed transmission device based on the low orbit satellite and 5GS, which comprises: the conveying control module splits an AI/ML model into model data and training data, divides the model data into delay sensitive components and sets the delay sensitive components as a ground transmission mode, and divides the training data into delay tolerant components and sets the delay tolerant components as a satellite transmission mode; the first receiving and transmitting module is used for transmitting the training data request to corresponding UE in a satellite transmission mode when receiving the training data request transmitted by the AI/ML system during the AI/ML model training, so that the UE collects the corresponding training data and feeds back the training data in the satellite transmission mode, and the training data transmitted by the UE is transmitted to the AI/ML system for the AI/ML system to train the AI/ML model; and the second receiving and transmitting module feeds back the trained model data of the AI/ML model to the corresponding UE in a ground transmission mode when receiving a model subscription request sent by the UE when the AI/ML model is used.
Preferably, the transmission control module, the first transceiver module and the second transceiver module are deployed in an NWDAF network element, and the first transceiver module receives a training data request sent by the AI/ML system, sends the training data request to a corresponding UE in a satellite transmission mode, and receives training data fed back by the UE in a satellite transmission mode; and the second receiving and transmitting module receives the model data of the trained AI/ML model sent by the edge server and feeds back the model data of the trained AI/ML model to the UE in a ground transmission mode.
The invention also discloses an AI/ML model distributed transmission system based on the low orbit satellite and the 5GS, which comprises a satellite access point, a ground access point, a 5G core network, an NTN gateway and an AI/ML system, wherein the satellite access point is connected with the 5G core network through the NTN gateway, the ground access point is connected with the 5G core network, the AI/ML system is connected with the 5G core network, the 5G core network splits the AI/ML model into model data and training data, the model data is divided into delay sensitive components and is set as a ground transmission mode, and the training data is divided into delay tolerant components and is set as a satellite transmission mode; the AI/ML system sends a training data request to the 5G core network during AI/ML model training, the 5G core network transmits the training data request to corresponding UE in a satellite transmission mode through an NTN gateway and a satellite access point so that the UE collects corresponding training data and feeds back the training data in a satellite transmission mode through the satellite access point and the NTN gateway, the 5G core network transmits the training data fed back by the UE to the AI/ML system, and the AI/ML system carries out AI/ML model training according to the training data to obtain trained model data of an AI/ML model; and the 5G core network receives a model subscription request sent by the UE when the AI/ML model is used, and transmits model data of the trained AI/ML model to the corresponding UE through the ground access point according to the model subscription request in a ground transmission mode.
Preferably, the AI/ML distributed transmission system based on the low orbit satellite and 5GS further comprises an edge server, wherein the AI/ML system transmits model data of the trained AI/ML model to the edge server for storage; and the edge server receives a model subscription request sent by the UE when the AI/ML model is used, sends model data of the trained AI/ML model to the 5G core network according to the model subscription request, and the 5G core network transmits the model data of the trained AI/ML model to the corresponding UE in a ground transmission mode through the ground access point.
Preferably, the 5G core network includes an NWDAF network element, where the NWDAF network element sets a transmission mode of the training data request and the training data as a satellite transmission mode and stores the data, and the NWDAF network element sets a transmission mode of the model subscription request and the AI/ML model as a ground transmission mode and stores the data.
Compared with the prior art, the method has the advantages that the AI/ML model is split into two parts of model data and training data, the model data is divided into the delay sensitive parts according to the requirements of the terminal on the two parts, the delay sensitive parts are used for transmission through the ground access, the training data is divided into the delay tolerant parts, the satellite access is used for transmission, the distributed transmission is performed, finally, the ground network resources are effectively saved through the distributed transmission method, the ground network resources can serve delay sensitive business more, and the service experience of the terminal is improved.
Drawings
Fig. 1 is a schematic diagram of the operation of the AI/ML model distributed transmission method based on low-orbit satellites and 5GS of the present invention.
Fig. 2 is a data transmission process diagram of the AI/ML model distributed transmission method based on the low-orbit satellite and 5GS according to the present invention.
Fig. 3 is a block diagram of the AI/ML model distributed transmission apparatus based on the low-orbit satellite and 5GS of the present invention.
Fig. 4 is a block diagram of the AI/ML model distributed transmission system based on low-orbit satellites and 5GS of the present invention.
Fig. 5 is a flowchart of the AI/ML model distributed transmission method, system based on low-orbit satellites and 5GS of the present invention.
Detailed Description
In order to describe the technical content, the constructional features, the achieved objects and effects of the present invention in detail, the following description is made in connection with the embodiments and the accompanying drawings.
Referring to fig. 1 and 5, the invention discloses an AI/ML model distributed transmission method based on a low-orbit satellite and 5GS, which comprises the steps of dividing an AI/ML model into model data and training data, dividing the model data into delay sensitive components and setting the delay sensitive components as a ground transmission mode, and dividing the training data into delay tolerant components and setting the delay tolerant components as a satellite transmission mode; when an AI/ML model is trained, when a training data request sent by an AI/ML system is received, the training data request is sent to a corresponding UE in a satellite transmission mode, so that the UE collects corresponding training data, the training data is fed back in the satellite transmission mode, and the training data sent by the UE is sent to the AI/ML system for the AI/ML system to train the AI/ML model; when the AI/ML model is used, when a model subscription request sent by the UE is received, model data of the trained AI/ML model is fed back to the UE in a ground transmission mode.
The UE performing the training data request may be a user end, or may be a device such as a remote monitoring end collecting multiple user data. The UE performing the training data request and the UE performing the model subscription request may be the same UE or different UEs.
Referring to fig. 2 and 5, in this embodiment, the NWDAF network element 20 receives a training data request S1 sent by the AI/ML system 10, and sends the training data request S1 to the corresponding UE40 in a satellite transmission manner, and the NWDAF network element 20 receives training data SM1 fed back by the UE40 in a satellite transmission manner; the method comprises the steps of receiving and storing model data SM2 of an AI/ML model trained by the AI/ML system 10 through an edge server 30 (EAS), receiving a model subscription request S2 sent by the UE40 through an NWDAF network element 20 and forwarding the model subscription request S2 to the edge server 30, receiving the model data SM2 of the trained AI/ML model sent by the edge server 30 through the NWDAF network element 20, and feeding back the model data SM2 of the trained AI/ML model to the UE in a ground transmission mode.
Specifically, the NWDAF network element 20 sets the transmission mode of the training data request S1 and the training data SM1 to a satellite transmission mode and stores them, and the NWDAF network element 20 sets the transmission mode of the model subscription request S2 and the AI/ML model SM2 to a terrestrial transmission mode and stores them.
The invention also discloses a computing device comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is configured to store at least one executable instruction that causes the processor to perform operations corresponding to the low-orbit satellite and 5GS based AI/ML model distributed transmission method described above.
The invention also discloses a computer storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute the operation corresponding to the AI/ML model distributed transmission method based on the low-orbit satellite and 5 GS.
Referring to fig. 3, the invention also discloses an AI/ML model distributed transmission device based on a low-orbit satellite and 5GS, which comprises a transmission control module 21, a first transceiver module 22 and a second transceiver module 23, wherein the transmission control module 21 splits the AI/ML model into model data SM2 and training data SM1, the model data SM2 is divided into a delay sensitive component and is set as a ground transmission mode, and the training data SM1 is divided into a delay tolerant component and is set as a satellite transmission mode. When receiving the training data S1 sent by the AI/ML system 10 during the AI/ML model training, the first transceiver module 22 sends the training data S1 to the corresponding UE40 in a satellite transmission manner, so that the UE40 collects the corresponding training data SM1, feeds back the training data SM1 in a satellite transmission manner, and sends the training data SM1 sent by the UE40 to the AI/ML system 10 for the AI/ML system 10 to perform the AI/ML model training. The second transceiver module 23 feeds back the model data SM2 of the trained AI/ML model to the corresponding UE40 in a ground transmission manner when receiving a model subscription request S2 sent by the UE40 when the AI/ML model is used.
Referring to fig. 3, the transmission control module 21, the first transceiver module 22 and the second transceiver module 23 are disposed in the NWDAF network element 20, and the first transceiver module 22 receives the training data S1 sent by the AI/ML system 10, sends the training data S1 to the corresponding UE40 in a satellite transmission manner, and receives the training data SM1 fed back by the UE40 in a satellite transmission manner; the second transceiver module 23 receives the trained AI/ML model data SM2 sent by the edge server 30 (EAS), and feeds back the trained AI/ML model data SM2 to the UE40 in a terrestrial transmission manner.
Referring to fig. 4 and 5, the present invention also discloses an AI/ML model distributed transmission system based on low-orbit satellites and 5GS, which comprises a satellite access point 51, a ground access point 52, a 5G core network 200, an NTN gateway 53 and an AI/ML system 10. The satellite access point 51 is connected to the 5G core network 200 through the NTN gateway 53, the terrestrial access point 52 is connected to the 5G core network 200, and the AI/ML system 10 is connected to the 5G core network 200.
Specifically, the 5G core network 200 splits the AI/ML model into model data SM2 and training data SM1, divides the model data SM2 into delay sensitive components and sets the delay sensitive components as a terrestrial transmission mode, and divides the training data SM1 into delay tolerant components and sets the delay tolerant components as a satellite transmission mode.
During AI/ML model training, the AI/ML system 10 sends training data to the 5G core network 200 to obtain S1, the 5G core network 200 transmits the training data to the corresponding UE40 in a satellite transmission mode through the NTN gateway 53 and the satellite access point 51, so that the UE40 collects the corresponding training data SM1, and feeds back the training data SM1 in a satellite transmission mode through the satellite access point 51 and the NTN gateway 53, the 5G core network 200 transmits the training data SM1 fed back by the UE40 to the AI/ML system 10, and the AI/ML system 10 performs AI/ML model training according to the training data SM1 to obtain model data SM2 of a trained AI/ML model.
The 5G core network 200 receives a model subscription request S2 sent by the UE40 when the AI/ML model is used, and transmits the model data SM2 of the trained AI/ML model to the corresponding UE40 through the ground access point 52 according to the model subscription request S2 in a ground transmission manner.
With continued reference to fig. 4 and 5, the low-orbit satellite and 5GS based AI/ML model distributed transmission system further includes an edge server 30 (EAS), the AI/ML system 10 delivering model data SM2 of the trained AI/ML model to the edge server 30 for storage; the edge server 30 receives a model subscription request S2 sent by the UE40 when the AI/ML model is used, and sends model data SM2 of the trained AI/ML model to the 5G core network 200 according to the model subscription request S2. The 5G core network 200 transmits the model data SM2 of the trained AI/ML model to the corresponding UE40 through the ground access point 52 by using a ground transmission mode.
Specifically, the 5G core network 200 includes an NWDAF network element 20, where the NWDAF network element 20 sets a transmission mode of the training data S1 and the training data SM1 to a satellite transmission mode and stores the transmission mode, and the NWDAF network element 20 sets a transmission mode of the model subscription request S2 and the AI/ML model to a ground transmission mode and stores the transmission mode.
The working process of the present invention is described with reference to fig. 5, which includes steps 1 to 7 when the AI/ML model is trained, and steps 8 to 11 when the AI/ML model is used.
1. During AI/ML model training, the AI/ML system sends a training data request S1 to the 5G core network.
2. The 5G core network forwards the training data request S1 to the UE in a satellite transmission mode.
3. The UE collects and feeds back the collected training data SM1 to the 5G core network in a satellite transmission mode.
4. The 5G core network forwards the training data SM1 to the AI/ML system.
5. The AI/ML system performs data training on the constructed AI/ML model according to training data SM1 to obtain a trained AI/ML model. During training, the AI/ML model needs to be trained to a certain precision (the quality of the model can be evaluated by using MAE\MSE, etc.).
6. The AI/ML system sends model data SM2 of the trained AI/ML model to an edge memory (EAS).
7. An edge memory (EAS) stores model data SM2.
8. When the AI/ML model is used, the UE sends a model subscription request S2 to the 5G core network. The transmission mode of the model subscription request S2 is determined by the communication protocol between the UE and the 5G core network, and is not limited to the terrestrial transmission mode, but may be transmitted by other transmission modes, including a satellite transmission mode. In this embodiment, the UE transmits the model subscription requests S2 to 5G core network in a terrestrial transmission manner.
9. The 5G core network forwards the model subscription request S2 to an edge memory (EAS) in the vicinity of the UE.
10. An edge memory (EAS) feeds back model data SM2 to 5G core networks of the trained AI/ML model.
11. The 5G core network forwards the model data SM2 of the trained AI/ML model to the corresponding UE in a ground transmission mode.
The steps 1-7 are repeated in a preset period, and the purpose is to continuously update the model so that the model can adapt to the service change of the UE and the precision of the model is kept at a higher level.
Wherein, before step 1, there is also a step of establishing a PDU session to enable the UE to maintain communication with each network element (each network element of the 5G core network).
Compared with the prior art, the method has the advantages that the AI/ML model is split into two parts of the model data SM2 and the training data SM1, the model data SM2 is divided into the delay sensitive parts according to the requirements of the terminal on the two parts, the delay tolerant parts are transmitted through the ground access, the training data SM1 is divided into the delay tolerant parts, the satellite access is used for transmission, and the distributed transmission is carried out, so that the ground network resources are effectively saved, the ground network resources can serve delay sensitive services more, and the service experience of the terminal is improved.
The present invention can be implemented by using only hardware or by using software and a necessary general hardware platform through the description of the foregoing embodiments. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product. The software product may be stored in a non-volatile or non-transitory storage medium, which may be a compact disk read-only memory (CD-ROM), a USB flash disk, or a removable hard disk. The software product comprises instructions that enable a computer device (personal computer, server, or network device) to perform the method provided in the embodiments of the present invention. For example, such execution may correspond to simulation of a logical operation as described herein. The software product may additionally or alternatively include a plurality of instructions that enable a computer apparatus to perform operations for configuring or programming digital logic devices in accordance with embodiments of the present invention.
The foregoing description of the preferred embodiments of the present invention is not intended to limit the scope of the claims, which follow, as defined in the claims.
Claims (10)
1. An AI/ML model distributed transmission method based on a low orbit satellite and 5GS is characterized in that:
dividing an AI/ML model into model data and training data, dividing the model data into delay sensitive components and setting the delay sensitive components as a ground transmission mode, and dividing the training data into delay tolerant components and setting the delay tolerant components as a satellite transmission mode;
when an AI/ML model is trained, when a training data request sent by an AI/ML system is received, the training data request is sent to a corresponding UE in a satellite transmission mode, so that the UE collects the corresponding training data, the training data is fed back in the satellite transmission mode, and the training data sent by the UE is sent to the AI/ML system for the AI/ML system to train the AI/ML model;
when the AI/ML model is used, when a model subscription request sent by the UE is received, model data of the trained AI/ML model is fed back to the corresponding UE in a ground transmission mode.
2. The AI/ML model distributed transmission method based on low-orbit satellites and 5GS according to claim 1, wherein: receiving a training data request sent by the AI/ML system through an NWDAF network element, sending the training data request to corresponding UE in a satellite transmission mode, and receiving training data fed back by the UE in the satellite transmission mode through the NWDAF network element; and receiving and storing the model data of the AI/ML model trained by the AI/ML system through an edge server, receiving a model subscription request sent by the UE through an NWDAF network element, forwarding the model subscription request to the edge server, receiving the model data of the trained AI/ML model sent by the edge server through the NWDAF network element, and feeding back the model data of the trained AI/ML model to the corresponding UE in a ground transmission mode.
3. The AI/ML model distributed transmission method based on low-orbit satellites and 5GS according to claim 2, wherein: the NWDAF network element sets the transmission mode of the training data request and the training data as a satellite transmission mode and stores the data, and the NWDAF network element sets the transmission mode of the model subscription request and the AI/ML model as a ground transmission mode and stores the data.
4. A computing device, characterized by: comprising the following steps: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction that causes the processor to perform operations corresponding to the low-orbit satellite and 5GS based AI/ML model distributed transmission method as recited in any one of claims 1-3.
5. A computer storage medium, characterized by: at least one executable instruction stored in the storage medium causes the processor to perform operations corresponding to the low-orbit satellite and 5GS based AI/ML model distributed transmission method as recited in any one of claims 1-3.
6. An AI/ML model distributed transmission device based on low orbit satellite and 5GS, which is characterized in that: comprising the following steps:
the conveying control module splits an AI/ML model into model data and training data, divides the model data into delay sensitive components and sets the delay sensitive components as a ground transmission mode, and divides the training data into delay tolerant components and sets the delay tolerant components as a satellite transmission mode;
the first receiving and transmitting module is used for transmitting the training data request to corresponding UE in a satellite transmission mode when receiving the training data request transmitted by the AI/ML system during AI/ML model training, so that the UE collects the corresponding training data and feeds back the training data in the satellite transmission mode, and the training data transmitted by the UE is transmitted to the AI/ML system for AI/ML model training by the AI/ML system;
and the second receiving and transmitting module feeds back the trained model data of the AI/ML model to the corresponding UE in a ground transmission mode when receiving a model subscription request sent by the UE when the AI/ML model is used.
7. The AI/ML model distributed transmission apparatus based on low-orbit satellites and 5GS according to claim 6, wherein: the conveying control module, the first receiving and transmitting module and the second receiving and transmitting module are deployed in an NWDAF network element, the first receiving and transmitting module receives a training data request sent by the AI/ML system, the training data request is sent to corresponding UE in a satellite transmission mode, and training data fed back by the UE is received in a satellite transmission mode; and the second receiving and transmitting module receives the model data of the trained AI/ML model sent by the edge server, and feeds back the model data of the trained AI/ML model to the corresponding UE in a ground transmission mode.
8. An AI/ML model distributed transmission system based on low orbit satellite and 5GS, characterized in that: the system comprises a satellite access point, a ground access point, a 5G core network, an NTN gateway and an AI/ML system, wherein the satellite access point is connected with the 5G core network through the NTN gateway, the ground access point is connected with the 5G core network, the AI/ML system is connected with the 5G core network, the 5G core network splits an AI/ML model into model data and training data, the model data is divided into delay sensitive components and is set as a ground transmission mode, and the training data is divided into delay tolerant components and is set as a satellite transmission mode;
the AI/ML system sends a training data request to the 5G core network during AI/ML model training, the 5G core network transmits the training data request to corresponding UE in a satellite transmission mode through an NTN gateway and a satellite access point so that the UE collects corresponding training data and feeds back the training data in a satellite transmission mode through the satellite access point and the NTN gateway, the 5G core network transmits the training data fed back by the UE to the AI/ML system, and the AI/ML system carries out AI/ML model training according to the training data to obtain trained model data of an AI/ML model;
and the 5G core network receives a model subscription request sent by the UE when the AI/ML model is used, and transmits model data of the trained AI/ML model to the corresponding UE through a ground transmission mode according to the model subscription request through the ground access point.
9. The low-orbit satellite and 5GS based AI/ML model distributed transmission system according to claim 8, wherein: the AI/ML system transmits model data of the trained AI/ML model to the edge server for storage; and the edge server receives a model subscription request sent by the UE when the AI/ML model is used, sends the model data of the trained AI/ML model to the 5G core network according to the model subscription request, and the 5G core network transmits the model data of the trained AI/ML model to the corresponding UE in a ground transmission mode through the ground access point.
10. The low-orbit satellite and 5GS based AI/ML model distributed transmission system according to claim 8, wherein: the 5G core network comprises an NWDAF network element, the NWDAF network element sets a transmission mode of training data request and training data as a satellite transmission mode and stores the transmission mode, and the NWDAF network element sets a transmission mode of the model subscription request and an AI/ML model as a ground transmission mode and stores the transmission mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310312924.4A CN116346206A (en) | 2023-03-27 | 2023-03-27 | AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310312924.4A CN116346206A (en) | 2023-03-27 | 2023-03-27 | AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116346206A true CN116346206A (en) | 2023-06-27 |
Family
ID=86883557
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310312924.4A Pending CN116346206A (en) | 2023-03-27 | 2023-03-27 | AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116346206A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1747353B (en) * | 2004-09-08 | 2013-04-10 | 阿尔卡特公司 | Rural broadband hybrid satellite/terrestrial network architecture |
| CN113115461A (en) * | 2021-04-23 | 2021-07-13 | 北京科技大学 | QMIX-based wireless resource allocation optimization method and device |
| CN113873538A (en) * | 2020-06-30 | 2021-12-31 | 华为技术有限公司 | A model data transmission method and communication device |
| CN114339821A (en) * | 2020-09-30 | 2022-04-12 | 英特尔公司 | Method and apparatus for machine learning model sharing between distributed NWDAFs |
| CN115514769A (en) * | 2022-09-14 | 2022-12-23 | 中山大学 | Satellite Elastic Internet Resource Scheduling Method, System, Computer Equipment and Medium |
-
2023
- 2023-03-27 CN CN202310312924.4A patent/CN116346206A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1747353B (en) * | 2004-09-08 | 2013-04-10 | 阿尔卡特公司 | Rural broadband hybrid satellite/terrestrial network architecture |
| CN113873538A (en) * | 2020-06-30 | 2021-12-31 | 华为技术有限公司 | A model data transmission method and communication device |
| CN114339821A (en) * | 2020-09-30 | 2022-04-12 | 英特尔公司 | Method and apparatus for machine learning model sharing between distributed NWDAFs |
| CN113115461A (en) * | 2021-04-23 | 2021-07-13 | 北京科技大学 | QMIX-based wireless resource allocation optimization method and device |
| CN115514769A (en) * | 2022-09-14 | 2022-12-23 | 中山大学 | Satellite Elastic Internet Resource Scheduling Method, System, Computer Equipment and Medium |
Non-Patent Citations (3)
| Title |
|---|
| ""draft-29520-g20-rm_mcc"", 3GPP TSG_CT\\WG3_INTERWORKING_EX-CN3 * |
| 欧阳晔等: ""通信人工智能的下一个十年"", 《电信科学》, no. 3, pages 1 - 32 * |
| 牛煜霞等: ""NWDAF网络数据分析功能的标准演进"", 《移动通信》, vol. 47, pages 29 - 33 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11762703B2 (en) | Multi-region request-driven code execution system | |
| US10372499B1 (en) | Efficient region selection system for executing request-driven code | |
| US20190028548A1 (en) | Transport of control data in proxy-based network communications | |
| CN112417255B (en) | Online state receiving method and device, storage medium and electronic device | |
| US20250184224A1 (en) | Virtual resource processing method and apparatus | |
| CN111431730B (en) | Service processing method, system, computer equipment and readable medium | |
| JP2022546108A (en) | Information processing method, device, equipment and computer-readable storage medium | |
| JP2022084661A (en) | Synchronization method and equipment | |
| CN117082477A (en) | Task processing method, device, system, storage medium and electronic equipment | |
| US9106596B2 (en) | Method and apparatus of configuring a data broadcast service | |
| CN102148839A (en) | Method, device and system for managing FTP (file transfer protocol) server | |
| CN116346206A (en) | AI/ML model distributed transmission method, device and system based on low orbit satellite and 5GS | |
| CN112910958B (en) | Method, device, communication system and storage medium for unifying feature vectors | |
| CN103034632A (en) | Information transmitting method and a system | |
| CN115622992A (en) | File downloading method, device, equipment and medium based on Wi-Fi direct connection | |
| CN114416275A (en) | Method and system for synchronizing virtual machine states by multiple management clients | |
| CN114465997A (en) | Data synchronization method and device, electronic equipment and storage medium | |
| CN113259874A (en) | Message processing method, electronic device and storage medium | |
| CN113301104A (en) | Data processing system and method | |
| US12193113B2 (en) | Bulk data processing system | |
| CN111988859A (en) | Downlink control information transmission method | |
| CN113300971A (en) | Data processing system and method | |
| CN114367114A (en) | Data transmission method, device, device, medium and program product of virtual scene | |
| WO2017032212A1 (en) | Data stream processing method and apparatus | |
| CN111355768B (en) | Video data transmission method and device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230627 |
|
| RJ01 | Rejection of invention patent application after publication |