WO2022139230A1 - 무선 통신 시스템에서 스플릿 포인트를 조정하는 방법 및 장치 - Google Patents
무선 통신 시스템에서 스플릿 포인트를 조정하는 방법 및 장치 Download PDFInfo
- Publication number
- WO2022139230A1 WO2022139230A1 PCT/KR2021/017979 KR2021017979W WO2022139230A1 WO 2022139230 A1 WO2022139230 A1 WO 2022139230A1 KR 2021017979 W KR2021017979 W KR 2021017979W WO 2022139230 A1 WO2022139230 A1 WO 2022139230A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- split
- split point
- intermediate data
- base station
- point
- Prior art date
Links
- 238000004891 communication Methods 0.000 title claims abstract description 94
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000013473 artificial intelligence Methods 0.000 claims description 111
- 230000015654 memory Effects 0.000 description 36
- 238000013528 artificial neural network Methods 0.000 description 23
- 238000012549 training Methods 0.000 description 20
- 230000008569 process Effects 0.000 description 16
- 238000010586 diagram Methods 0.000 description 14
- 238000005516 engineering process Methods 0.000 description 14
- 230000006870 function Effects 0.000 description 12
- 238000010801 machine learning Methods 0.000 description 11
- 230000005540 biological transmission Effects 0.000 description 10
- 238000012545 processing Methods 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 7
- 235000009499 Vanilla fragrans Nutrition 0.000 description 5
- 244000263375 Vanilla tahitensis Species 0.000 description 5
- 235000012036 Vanilla tahitensis Nutrition 0.000 description 5
- 230000001934 delay Effects 0.000 description 5
- 230000001413 cellular effect Effects 0.000 description 4
- 230000007423 decrease Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000003190 augmentative effect Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 239000004984 smart glass Substances 0.000 description 2
- 238000005406 washing Methods 0.000 description 2
- 241000760358 Enodes Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001151 other effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/098—Distributed learning, e.g. federated learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Definitions
- the following description relates to an apparatus and method for adjusting a split point based on AoI in a wireless communication system. More specifically, the present disclosure relates to an apparatus and method for determining a split point in a system that performs split inference through a U-shaped split AI/ML (artificial intelligence/machine learning) model.
- AI/ML artificial intelligence/machine learning
- a wireless access system is a multiple access system that can support communication with multiple users by sharing available system resources (bandwidth, transmission power, etc.).
- Examples of the multiple access system include a code division multiple access (CDMA) system, a frequency division multiple access (FDMA) system, a time division multiple access (TDMA) system, an orthogonal frequency division multiple access (OFDMA) system, and a single carrier frequency (SC-FDMA) system. division multiple access) systems.
- CDMA code division multiple access
- FDMA frequency division multiple access
- TDMA time division multiple access
- OFDMA orthogonal frequency division multiple access
- SC-FDMA single carrier frequency division multiple access
- an enhanced mobile broadband (eMBB) communication technology has been proposed compared to the existing radio access technology (RAT).
- eMBB enhanced mobile broadband
- RAT radio access technology
- UE reliability and latency sensitive services/user equipment
- mMTC massive machine type communications
- the present disclosure may provide an apparatus and method for split point coordination in a wireless communication system.
- the present disclosure may provide an apparatus and method for determining a split point in consideration of age of information (AoI) of data in a wireless communication system.
- AoI age of information
- the present disclosure may provide an apparatus and method for maintaining the freshness of information in a system performing split inference by adjusting a split point in consideration of AoI of data in a wireless communication system.
- the present disclosure may provide an apparatus and method for performing communication by reflecting a real-time environment in which split inference is performed by adjusting a split point in consideration of AoI of data in a wireless communication system.
- the present disclosure may provide a method of operating a terminal in a wireless communication system.
- first split inference split inference
- the first split point indicates a first point at which the terminal performs the first split inference based on a U-shaped segmented artificial intelligence learning model
- the two split points may indicate a second point at which the base station performs a second split inference based on the U-shaped split artificial intelligence learning model.
- the terminal transmits information on a time point when raw data of the terminal is transmitted to an input layer and information on the second split point together with the first intermediate data.
- the information on the first intermediate data reception time and information on the first split point transmitted to the base station and received by the base station may be received together with the second intermediate data.
- the step of the terminal adjusting the second split point may include measuring a Peak of AoI (PAoI) of the second intermediate data, and setting the measured PAoI value by the first base station. and adjusting the second split point based on a result of comparison with a threshold.
- PAoI Peak of AoI
- the second split point when the PAoI value is greater than the first threshold value, the second split point is moved or maintained in the direction of the input layer, and when the PAoI value is less than the first threshold value, the second split point is smaller than the first threshold value 2
- the split point may be moved or maintained in the direction of the output layer.
- the adjustment position of the second split point may be determined between after the first split point and before the output layer.
- the first split point may include: the base station measures the PAoI of the first intermediate data received from the terminal, and the measured PAoI value of the first intermediate data is a first value set by the base station When the value is less than the second threshold, it is moved or maintained in the direction of the output layer, and when the measured PAoI value of the first intermediate data is greater than the second threshold, it is moved or maintained in the direction of the input layer.
- the adjustment position of the first split point may be determined between after the input layer and before the second split point.
- a method of operating a base station in a wireless communication system performing initial settings for a first split point and a second split point, the first split point and the second split point to a terminal Transmitting information on at least one of the split points and receiving first intermediate data from the terminal, wherein the first intermediate data is a first split inference performed by the terminal based on the first split point (split inference), generating second intermediate data by performing a second split inference up to the second split point based on the first intermediate data, based on the first intermediate data It may include adjusting the first split point, and transmitting the generated second intermediate data and information on the adjusted first split point to the terminal.
- a transceiver in a terminal of a wireless communication system, a transceiver; and a processor coupled to the transceiver, wherein the processor receives information on at least one of a first split point and a second split point from a base station, and receives information on at least one of a first split point and a second split point based on the first split point perform split inference, generate first intermediate data, transmit the first intermediate data to the base station, and a second generated based on the second split point from the base station
- the intermediate data may be received, the remaining split inference may be performed from the second split point based on the second intermediate data, and the second split point may be adjusted based on the second intermediate data.
- a base station of a wireless communication system comprising: a transceiver; and a processor connected to the transceiver, wherein the processor performs initial settings for a first split point and a second split point, and provides a terminal with at least one of the first split point and the second split point.
- transmit information about one receive first intermediate data from the terminal, wherein the first intermediate data is based on a first split inference performed by the terminal based on a first split point generated, performing a second split inference up to the second split point based on the first intermediate data to generate second intermediate data, and adjusting the first split point based on the first intermediate data; , the generated second intermediate data and information on the adjusted first split point may be transmitted to the terminal.
- the base station performs initial setting for a first split point and a second split point, the base station provides the terminal with the first split point and transmitting information on at least one of the second split points, the terminal performing a first split inference based on the first split point, and first intermediate data generating, by the terminal, transmitting at least one of the first intermediate data and the adjusted second split point information to the base station; generating second intermediate data by performing a second split inference up to the second split point based on at least one of two split point information, the base station based on the first intermediate data received from the terminal adjusting the first split point; transmitting, by the base station, the generated second intermediate data and information on the adjusted first split point to the terminal;
- the method may include performing the remaining split inference from the second split point based on intermediate data, and the terminal adjusting the second split point based on the second intermediate data.
- a communication device at least one processor, at least one computer memory connected to the at least one processor, and storing instructions instructing operations according to execution by the at least one processor; comprising: receiving information about at least one of a first split point and a second split point from a base station, and performing a first split inference based on the first split point , generating first intermediate data, transmitting the first intermediate data to the base station, and adjusting the second intermediate data generated from the base station based on the second split point and the base station Receive information on the first split point, perform the remaining split inference from the second split point based on the second intermediate data, and determine the second split point based on the second intermediate data Can be adjusted.
- a non-transitory computer-readable medium storing at least one instruction
- the at least one command is configured to cause the device to receive information about at least one of a first split point and a second split point from a base station, and to perform a first step based on the first split point.
- Perform split inference generate first intermediate data, transmit the first intermediate data to the base station, and the second generated based on the second split point from the base station 2 intermediate data may be received, the remaining split inference may be performed from the second split point based on the second intermediate data, and the second split point may be adjusted based on the second intermediate data.
- a split point may be adjusted based on an age of information (AoI) of data in a wireless communication system.
- AoI age of information
- freshness of data can be maintained by adjusting a split point based on AoI of data in a wireless communication system.
- FIG. 1 is a diagram illustrating an example of a communication system applicable to the present disclosure.
- FIG. 2 is a diagram illustrating an example of a wireless device applicable to the present disclosure.
- FIG. 3 is a diagram illustrating another example of a wireless device applicable to the present disclosure.
- FIG. 4 is a diagram illustrating an example of a portable device applicable to the present disclosure.
- FIG. 5 is a diagram illustrating an example of a vehicle or autonomous driving vehicle applicable to the present disclosure.
- FIG. 6 is a diagram illustrating an example of AI (Artificial Intelligence) applicable to the present disclosure.
- AI Artificial Intelligence
- FIG. 7 is a diagram illustrating data related to data generation and collection according to an embodiment of the present invention.
- FIG 8 shows an example of a system for performing a split AI/ML model, according to an embodiment of the present disclosure.
- FIG. 9 illustrates an example of a system to which a U-shaped split AI/ML model is applied, according to an embodiment of the present disclosure.
- FIG. 10 illustrates an example of an operation for determining a split point of a split engine according to an embodiment of the present disclosure.
- FIG 11 illustrates an example of measuring age of information (AoI) according to an embodiment of the present disclosure.
- FIG. 12 illustrates an example of measuring AoI using a cost of update delay (CoUD) metric according to an embodiment of the present disclosure.
- CoUD cost of update delay
- FIG. 13 illustrates an example of a graph illustrating evaluation of communication performance of a device according to a candidate split point setting of a VGG-16 model according to an embodiment of the present disclosure.
- FIG. 14 illustrates an example of a system supporting a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- FIG. 15 illustrates an example of an operation of adjusting a split point in a U-shaped AI/ML split model according to an embodiment of the present disclosure.
- FIG. 16 illustrates an example of an operation in which a base station adjusts a first split point in a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- FIG 17 illustrates an example of an operation in which an apparatus adjusts a second split point in a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- FIG. 18 illustrates an operation in which a base station adjusts a first split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- FIG. 19 illustrates an operation in which an apparatus transmits information for split point adjustment to a base station in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- FIG. 20 illustrates an operation in which an apparatus adjusts a second split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- FIG. 21 illustrates an operation in which the device adjusts a split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- each component or feature may be considered optional unless explicitly stated otherwise.
- Each component or feature may be implemented in a form that is not combined with other components or features.
- some components and/or features may be combined to configure an embodiment of the present disclosure.
- the order of operations described in embodiments of the present disclosure may be changed. Some configurations or features of one embodiment may be included in other embodiments, or may be replaced with corresponding configurations or features of other embodiments.
- the base station has a meaning as a terminal node of a network that directly communicates with the mobile station.
- a specific operation described as being performed by the base station in this document may be performed by an upper node of the base station in some cases.
- the 'base station' is a term such as a fixed station, a Node B, an eNB (eNode B), a gNB (gNode B), an ng-eNB, an advanced base station (ABS) or an access point (access point).
- eNode B eNode B
- gNode B gNode B
- ng-eNB ng-eNB
- ABS advanced base station
- access point access point
- a terminal includes a user equipment (UE), a mobile station (MS), a subscriber station (SS), a mobile subscriber station (MSS), It may be replaced by terms such as a mobile terminal or an advanced mobile station (AMS).
- UE user equipment
- MS mobile station
- SS subscriber station
- MSS mobile subscriber station
- AMS advanced mobile station
- a transmitting end refers to a fixed and/or mobile node that provides a data service or a voice service
- a receiving end refers to a fixed and/or mobile node that receives a data service or a voice service.
- the mobile station may be a transmitting end, and the base station may be a receiving end.
- the mobile station may be the receiving end, and the base station may be the transmitting end.
- Embodiments of the present disclosure are wireless access systems IEEE 802.xx system, 3rd Generation Partnership Project (3GPP) system, 3GPP Long Term Evolution (LTE) system, 3GPP 5G (5th generation) NR (New Radio) system, and 3GPP2 system among It may be supported by standard documents disclosed in at least one, and in particular, embodiments of the present disclosure are supported by 3GPP TS (technical specification) 38.211, 3GPP TS 38.212, 3GPP TS 38.213, 3GPP TS 38.321 and 3GPP TS 38.331 documents. can be
- embodiments of the present disclosure may be applied to other wireless access systems, and are not limited to the above-described system. As an example, it may be applicable to a system applied after the 3GPP 5G NR system, and is not limited to a specific system.
- CDMA code division multiple access
- FDMA frequency division multiple access
- TDMA time division multiple access
- OFDMA orthogonal frequency division multiple access
- SC-FDMA single carrier frequency division multiple access
- LTE is 3GPP TS 36.xxx Release 8 or later
- LTE technology after 3GPP TS 36.xxx Release 10 may be referred to as LTE-A
- xxx Release 13 may be referred to as LTE-A pro.
- 3GPP NR may mean technology after TS 38.xxx Release 15.
- 3GPP 6G may mean technology after TS Release 17 and/or Release 18.
- "xxx" means standard document detail number LTE/NR/6G may be collectively referred to as a 3GPP system.
- FIG. 1 is a diagram illustrating an example of a communication system applied to the present disclosure.
- a communication system 100 applied to the present disclosure includes a wireless device, a base station, and a network.
- the wireless device means a device that performs communication using a wireless access technology (eg, 5G NR, LTE), and may be referred to as a communication/wireless/5G device.
- the wireless device may include a robot 100a, a vehicle 100b-1, 100b-2, an extended reality (XR) device 100c, a hand-held device 100d, and a home appliance. appliance) 100e, an Internet of Things (IoT) device 100f, and an artificial intelligence (AI) device/server 100g.
- a wireless access technology eg, 5G NR, LTE
- XR extended reality
- IoT Internet of Things
- AI artificial intelligence
- the vehicle may include a vehicle equipped with a wireless communication function, an autonomous driving vehicle, a vehicle capable of performing inter-vehicle communication, and the like.
- the vehicles 100b-1 and 100b-2 may include an unmanned aerial vehicle (UAV) (eg, a drone).
- UAV unmanned aerial vehicle
- the XR device 100c includes augmented reality (AR)/virtual reality (VR)/mixed reality (MR) devices, and includes a head-mounted device (HMD), a head-up display (HUD) provided in a vehicle, a television, It may be implemented in the form of a smartphone, a computer, a wearable device, a home appliance, a digital signage, a vehicle, a robot, and the like.
- the portable device 100d may include a smart phone, a smart pad, a wearable device (eg, smart watch, smart glasses), and a computer (eg, a laptop computer).
- the home appliance 100e may include a TV, a refrigerator, a washing machine, and the like.
- the IoT device 100f may include a sensor, a smart meter, and the like.
- the base station 120 and the network 130 may be implemented as a wireless device, and a specific wireless device 120a may operate as a base station/network node to other wireless devices.
- FIG. 2 is a diagram illustrating an example of a wireless device applicable to the present disclosure.
- a first wireless device 200a and a second wireless device 200b may transmit/receive wireless signals through various wireless access technologies (eg, LTE, NR).
- ⁇ first wireless device 200a, second wireless device 200b ⁇ is ⁇ wireless device 100x, base station 120 ⁇ of FIG. 1 and/or ⁇ wireless device 100x, wireless device 100x) ⁇ can be matched.
- the first wireless device 200a includes one or more processors 202a and one or more memories 204a, and may further include one or more transceivers 206a and/or one or more antennas 208a.
- the processor 202a controls the memory 204a and/or the transceiver 206a and may be configured to implement the descriptions, functions, procedures, suggestions, methods, and/or operational flowcharts disclosed herein.
- the processor 202a may process information in the memory 204a to generate first information/signal, and then transmit a wireless signal including the first information/signal through the transceiver 206a.
- the processor 202a may receive the radio signal including the second information/signal through the transceiver 206a, and then store the information obtained from the signal processing of the second information/signal in the memory 204a.
- the memory 204a may be connected to the processor 202a and may store various information related to the operation of the processor 202a.
- the memory 204a may provide instructions for performing some or all of the processes controlled by the processor 202a, or for performing the descriptions, functions, procedures, suggestions, methods, and/or operational flowcharts disclosed herein. may store software code including
- the processor 202a and the memory 204a may be part of a communication modem/circuit/chip designed to implement a wireless communication technology (eg, LTE, NR).
- a wireless communication technology eg, LTE, NR
- the transceiver 206a may be coupled to the processor 202a and may transmit and/or receive wireless signals via one or more antennas 208a.
- the transceiver 206a may include a transmitter and/or a receiver.
- the transceiver 206a may be used interchangeably with a radio frequency (RF) unit.
- RF radio frequency
- a wireless device may refer to a communication modem/circuit/chip.
- the second wireless device 200b includes one or more processors 202b, one or more memories 204b, and may further include one or more transceivers 206b and/or one or more antennas 208b.
- the processor 202b controls the memory 204b and/or the transceiver 206b and may be configured to implement the descriptions, functions, procedures, proposals, methods, and/or operational flowcharts disclosed herein.
- the processor 202b may process information in the memory 204b to generate third information/signal, and then transmit a wireless signal including the third information/signal through the transceiver 206b.
- the processor 202b may receive the radio signal including the fourth information/signal through the transceiver 206b, and then store information obtained from signal processing of the fourth information/signal in the memory 204b.
- the memory 204b may be connected to the processor 202b and may store various information related to the operation of the processor 202b.
- the memory 204b may provide instructions for performing some or all of the processes controlled by the processor 202b, or for performing the descriptions, functions, procedures, suggestions, methods, and/or operational flowcharts disclosed herein. may store software code including
- the processor 202b and the memory 204b may be part of a communication modem/circuit/chip designed to implement a wireless communication technology (eg, LTE, NR).
- a wireless communication technology eg, LTE, NR
- the transceiver 206b may be coupled to the processor 202b and may transmit and/or receive wireless signals via one or more antennas 208b.
- Transceiver 206b may include a transmitter and/or receiver.
- Transceiver 206b may be used interchangeably with an RF unit.
- a wireless device may refer to a communication modem/circuit/chip.
- FIG. 3 is a diagram illustrating another example of a wireless device applied to the present disclosure.
- a wireless device 300 corresponds to the wireless devices 200a and 200b of FIG. 2 , and includes various elements, components, units/units, and/or modules. ) can be composed of
- the wireless device 300 may include a communication unit 310 , a control unit 320 , a memory unit 330 , and an additional element 340 .
- the communication unit may include communication circuitry 312 and transceiver(s) 314 .
- communication circuitry 312 may include one or more processors 202a, 202b and/or one or more memories 204a, 204b of FIG. 2 .
- the transceiver(s) 314 may include one or more transceivers 206a , 206b and/or one or more antennas 208a , 208b of FIG. 2 .
- the control unit 320 is electrically connected to the communication unit 310 , the memory unit 330 , and the additional element 340 and controls general operations of the wireless device.
- the controller 320 may control the electrical/mechanical operation of the wireless device based on the program/code/command/information stored in the memory unit 330 .
- control unit 320 transmits the information stored in the memory unit 330 to the outside (eg, another communication device) through the communication unit 310 through a wireless/wired interface, or externally (eg, through the communication unit 310) Information received through a wireless/wired interface from another communication device) may be stored in the memory unit 330 .
- FIG. 4 is a diagram illustrating an example of a mobile device applied to the present disclosure.
- the portable device may include a smart phone, a smart pad, a wearable device (eg, a smart watch, smart glasses), and a portable computer (eg, a laptop computer).
- the mobile device may be referred to as a mobile station (MS), a user terminal (UT), a mobile subscriber station (MSS), a subscriber station (SS), an advanced mobile station (AMS), or a wireless terminal (WT).
- MS mobile station
- UT user terminal
- MSS mobile subscriber station
- SS subscriber station
- AMS advanced mobile station
- WT wireless terminal
- the mobile device 400 includes an antenna unit 408 , a communication unit 410 , a control unit 420 , a memory unit 430 , a power supply unit 440a , an interface unit 440b , and an input/output unit 440c .
- the antenna unit 408 may be configured as a part of the communication unit 410 .
- Blocks 410 to 430/440a to 440c respectively correspond to blocks 310 to 330/340 of FIG. 3 .
- the communication unit 410 may transmit and receive signals (eg, data, control signals, etc.) with other wireless devices and base stations.
- the controller 420 may control components of the portable device 400 to perform various operations.
- the controller 420 may include an application processor (AP).
- the memory unit 430 may store data/parameters/programs/codes/commands necessary for driving the portable device 400 . Also, the memory unit 430 may store input/output data/information.
- the power supply unit 440a supplies power to the portable device 400 and may include a wired/wireless charging circuit, a battery, and the like.
- the interface unit 440b may support a connection between the portable device 400 and other external devices.
- the interface unit 440b may include various ports (eg, an audio input/output port and a video input/output port) for connection with an external device.
- the input/output unit 440c may receive or output image information/signal, audio information/signal, data, and/or information input from a user.
- the input/output unit 440c may include a camera, a microphone, a user input unit, a display unit 440d, a speaker, and/or a haptic module.
- the input/output unit 440c obtains information/signals (eg, touch, text, voice, image, video) input from the user, and the obtained information/signals are stored in the memory unit 430 . can be saved.
- the communication unit 410 may convert the information/signal stored in the memory into a wireless signal, and transmit the converted wireless signal directly to another wireless device or to a base station. Also, after receiving a radio signal from another radio device or base station, the communication unit 410 may restore the received radio signal to original information/signal.
- the restored information/signal may be stored in the memory unit 430 and output in various forms (eg, text, voice, image, video, haptic) through the input/output unit 440c.
- FIG. 5 is a diagram illustrating an example of a vehicle or autonomous driving vehicle applied to the present disclosure.
- the vehicle or autonomous driving vehicle may be implemented as a mobile robot, a vehicle, a train, an aerial vehicle (AV), a ship, and the like, but is not limited to the shape of the vehicle.
- AV aerial vehicle
- the vehicle or autonomous driving vehicle 500 includes an antenna unit 508 , a communication unit 510 , a control unit 520 , a driving unit 540a , a power supply unit 540b , a sensor unit 540c and autonomous driving.
- a unit 540d may be included.
- the antenna unit 550 may be configured as a part of the communication unit 510 .
- Blocks 510/530/540a to 540d respectively correspond to blocks 410/430/440 of FIG. 4 .
- the communication unit 510 may transmit/receive signals (eg, data, control signals, etc.) to and from external devices such as other vehicles, base stations (eg, base stations, roadside units, etc.), and servers.
- the controller 520 may control elements of the vehicle or the autonomous driving vehicle 500 to perform various operations.
- the controller 520 may include an electronic control unit (ECU).
- ECU electronice control unit
- AI devices include TVs, projectors, smartphones, PCs, laptops, digital broadcasting terminals, tablet PCs, wearable devices, set-top boxes (STBs), radios, washing machines, refrigerators, digital signage, robots, vehicles, etc. It may be implemented as a device or a mobile device.
- the AI device 600 includes a communication unit 610 , a control unit 620 , a memory unit 630 , input/output units 640a/640b , a learning processor unit 640c and a sensor unit 640d. may include.
- Blocks 910 to 930/940a to 940d may correspond to blocks 310 to 330/340 of FIG. 3 , respectively.
- the communication unit 610 uses wired and wireless communication technology to communicate with external devices such as other AI devices (eg, FIGS. 1, 100x, 120, 140) or an AI server ( FIGS. 1 and 140 ) and wired and wireless signals (eg, sensor information, user input, learning model, control signal, etc.). To this end, the communication unit 610 may transmit information in the memory unit 630 to an external device or transmit a signal received from the external device to the memory unit 630 .
- AI devices eg, FIGS. 1, 100x, 120, 140
- an AI server FIGS. 1 and 140
- wired and wireless signals eg, sensor information, user input, learning model, control signal, etc.
- the controller 620 may determine at least one executable operation of the AI device 600 based on information determined or generated using a data analysis algorithm or a machine learning algorithm. Then, the controller 620 may control the components of the AI device 600 to perform the determined operation. For example, the control unit 620 may request, search, receive, or utilize the data of the learning processor unit 640c or the memory unit 630, and is determined to be a predicted operation or desirable among at least one executable operation. Components of the AI device 600 may be controlled to execute the operation. In addition, the control unit 920 collects history information including user feedback on the operation contents or operation of the AI device 600 and stores it in the memory unit 630 or the learning processor unit 640c, or the AI server ( 1 and 140), and the like may be transmitted to an external device. The collected historical information may be used to update the learning model.
- the memory unit 630 may store data supporting various functions of the AI device 600 .
- the memory unit 630 may store data obtained from the input unit 640a , data obtained from the communication unit 610 , output data of the learning processor unit 640c , and data obtained from the sensing unit 640 .
- the memory unit 930 may store control information and/or software codes necessary for the operation/execution of the control unit 620 .
- the input unit 640a may acquire various types of data from the outside of the AI device 600 .
- the input unit 620 may obtain training data for model learning, input data to which the learning model is applied, and the like.
- the input unit 640a may include a camera, a microphone, and/or a user input unit.
- the output unit 640b may generate an output related to sight, hearing, or touch.
- the output unit 640b may include a display unit, a speaker, and/or a haptic module.
- the sensing unit 640 may obtain at least one of internal information of the AI device 600 , surrounding environment information of the AI device 600 , and user information by using various sensors.
- the sensing unit 640 may include a proximity sensor, an illuminance sensor, an acceleration sensor, a magnetic sensor, a gyro sensor, an inertial sensor, an RGB sensor, an IR sensor, a fingerprint recognition sensor, an ultrasonic sensor, an optical sensor, a microphone, and/or a radar. have.
- the learning processor unit 640c may train a model composed of an artificial neural network by using the training data.
- the learning processor unit 640c may perform AI processing together with the learning processor unit of the AI server ( FIGS. 1 and 140 ).
- the learning processor unit 640c may process information received from an external device through the communication unit 610 and/or information stored in the memory unit 630 . Also, the output value of the learning processor unit 940c may be transmitted to an external device through the communication unit 610 and/or stored in the memory unit 630 .
- AI The most important and newly introduced technology for 6G systems is AI.
- AI was not involved in the 4G system.
- 5G systems will support partial or very limited AI.
- the 6G system will be AI-enabled for full automation.
- Advances in machine learning will create more intelligent networks for real-time communication in 6G.
- Incorporating AI into communications can simplify and enhance real-time data transmission.
- AI can use numerous analytics to determine how complex target tasks are performed. In other words, AI can increase efficiency and reduce processing delays.
- AI can also play an important role in M2M, machine-to-human and human-to-machine communication.
- AI can be a rapid communication in the BCI (brain computer interface).
- BCI brain computer interface
- AI-based communication systems can be supported by metamaterials, intelligent structures, intelligent networks, intelligent devices, intelligent cognitive radios, self-sustaining wireless networks, and machine learning.
- AI-based physical layer transmission means applying a signal processing and communication mechanism based on an AI driver rather than a traditional communication framework in a fundamental signal processing and communication mechanism.
- a signal processing and communication mechanism based on an AI driver rather than a traditional communication framework in a fundamental signal processing and communication mechanism.
- deep learning-based channel coding and decoding, deep learning-based signal estimation and detection, deep learning-based multiple input multiple output (MIMO) mechanism It may include AI-based resource scheduling and allocation.
- Machine learning may be used for channel estimation and channel tracking, and may be used for power allocation, interference cancellation, and the like in a physical layer of a downlink (DL). In addition, machine learning may be used for antenna selection, power control, symbol detection, and the like in a MIMO system.
- DL downlink
- machine learning may be used for antenna selection, power control, symbol detection, and the like in a MIMO system.
- Deep learning-based AI algorithms require large amounts of training data to optimize training parameters.
- a lot of training data is used offline. This is because static training on training data in a specific channel environment may cause a contradiction between dynamic characteristics and diversity of a wireless channel.
- signals of the physical layer of wireless communication are complex signals.
- further research on a neural network for detecting a complex domain signal is needed.
- Machine learning refers to a set of operations that trains a machine to create a machine that can perform tasks that humans can or cannot do.
- Machine learning requires data and a learning model.
- data learning methods can be roughly divided into three types: supervised learning, unsupervised learning, and reinforcement learning.
- Neural network learning is to minimize output errors. Neural network learning repeatedly inputs learning data into the neural network, calculates the output and target errors of the neural network for the training data, and backpropagates the neural network error from the output layer of the neural network to the input layer in the direction to reduce the error. ) to update the weight of each node in the neural network.
- Supervised learning uses training data in which the correct answer is labeled in the training data, and in unsupervised learning, the correct answer may not be labeled in the training data. That is, for example, learning data in the case of supervised learning related to data classification may be data in which categories are labeled for each of the training data.
- the labeled training data is input to the neural network, and an error can be calculated by comparing the output (category) of the neural network with the label of the training data.
- the calculated error is back propagated in the reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back propagation.
- the change amount of the connection weight of each node to be updated may be determined according to a learning rate.
- the computation of the neural network on the input data and the backpropagation of errors can constitute a learning cycle (epoch).
- the learning rate may be applied differently depending on the number of repetitions of the learning cycle of the neural network. For example, in the early stage of learning a neural network, a high learning rate can be used to increase the efficiency by allowing the neural network to quickly obtain a certain level of performance, and in the late learning period, a low learning rate can be used to increase the accuracy.
- the learning method may vary depending on the characteristics of the data. For example, when the purpose of accurately predicting data transmitted from a transmitter in a communication system is at a receiver, it is preferable to perform learning using supervised learning rather than unsupervised learning or reinforcement learning.
- the learning model corresponds to the human brain, and the most basic linear model can be considered. ) is called
- the neural network cord used as a learning method is largely divided into deep neural networks (DNN), convolutional deep neural networks (CNN), and recurrent boltzmann machine (RNN) methods. and such a learning model can be applied.
- DNN deep neural networks
- CNN convolutional deep neural networks
- RNN recurrent boltzmann machine
- AI/ML artificial intelligence/machine learning
- a large amount of data (high-dimensional data collected through a biosensor, a large-capacity image, and a large-capacity video) may be generated, collected, and processed through the AI/ML model.
- the device for data generation and collection may be a cellular Internet of things (IoT) device.
- IoT Internet of things
- FIG. 7 is a diagram illustrating data related to data generation and collection according to an embodiment of the present invention.
- FIG. 7A it can be seen that the amount of cellular IoT connection for data generation and collection is steadily increasing.
- Today due to the increase in the number of cellular IoT devices, the amount of cellular IoT connections has increased. This also made it possible to obtain large amounts of data from the device.
- the device can classify and predict large amounts of high-dimensional data through AI/ML models. The performance of AI/ML models to process large amounts of data is also improving.
- the graph is a large neural network (LNN), a medium neural network (MNN), a small neural network (SNN), and a traditional learning algorithm (TLA).
- LNN large neural network
- MNN medium neural network
- SNN small neural network
- TLA traditional learning algorithm
- AI/ML models can build deep and wide models by continuously learning using a large amount of data to improve performance. Accordingly, the size of the model gradually increases, and the model may include up to several million parameters.
- inference may be a process of passing data through a network and outputting a result according to an input.
- inference may be a process of performing calculation on at least one layer.
- training may be a process of comparing the inference result with the correct answer, and adjusting variables in the network so that the inference result approximates or matches the correct answer when performing the inference later.
- split AI/ML model is used to split the AI/ML model established between the device and the network. It may be necessary to build
- FIG. 8 shows an example of a system 800 for performing a split AI/ML model, according to an embodiment of the present disclosure.
- the system 800 performs split inference in units of layers divided according to the requirements (computation capability, delay speed, privacy protection) of services (object recognition, augmented reality). ) can be assigned to devices and servers (base stations) to perform split AI/ML models.
- the split model 810 assigned to the device performs data processing by forward propagation up to the split point 850 , and an intermediate result of performing up to the split point 850 .
- Intermediate data 820 may be transmitted to the split model 840 allocated to the server. That is, the split point 850 is a point at which inference performance of the AI/ML model is split.
- the device may perform split inference on the layers up to the split point, and the server may perform split inference on the layers after the split point.
- Split inference can be performed.
- the split model 840 assigned to the server may perform split inference using the intermediate data 820 received from the split model assigned to the device.
- the split model 840 assigned to the server may transmit a label 830 that is a result of completing the split inference to the device.
- the device 810 may receive the label 830 from the server and perform the inference.
- a back propagation procedure may be performed by receiving a gradient of a loss function for the intermediate data 820 from the server.
- Split inference may be referred to as split inference, partial inference, or other terms having an equivalent technical meaning, and may not be limited to a specific name. However, for convenience of description, split inference will be used in the following description.
- a split point is required for a device and a server to perform split inference through a split AI/ML model.
- the split point may be a point at which the model performing the inference is divided into a device and a server.
- the initial split point may be set at a time point when training of the split model is completed.
- the split model format shown in FIG. 8 may be referred to as vanilla split modeling.
- the device does not transmit the raw data of the device to the base station, but it still has to transmit the label, so a privacy problem may occur for a label containing important information such as control and monitoring contents of devices. have.
- a U-shaped split AI/ML model may be used as shown in FIG. 9 below.
- FIG. 9 illustrates an example of a system 900 to which a U-shaped split AI/ML model is applied, according to an embodiment of the present disclosure.
- the system 900 may allocate a split AI/ML model for performing split inference in units of divided layers to the device 910 and the base station 920 .
- the system 900 may implement the U-shaped split AI/ML model by setting two split points 940 and 942 so that the device 910, not the base station 920, acquires the label 932 .
- the split AI/ML model assigned to the device 910 performs inference on the layers from the input layer to the first split point 940 using input data 930 .
- the device 910 may generate first intermediate data as a result of the inference, and may transmit the first intermediate data to the base station 920 .
- the split AI/ML model assigned to the base station 920 is inferred to the layers from the first split point 940 to the second split point 942 using the first intermediate data received from the device 910 . can be performed.
- the base station 920 may generate second intermediate data as a result of the inference, and may transmit the second intermediate data to the device 910 .
- the split AI/ML model assigned to the device 910 may perform inference from the second split point 942 to the output layer using the second intermediate data received from the base station 920 .
- the device 910 can generate the label 932 as a result of the completion of the inference. In this case, since the label 932 is obtained from the device 910 rather than the base station 920 , the privacy problem for the label 932 can be solved.
- a split point which is a location for splitting the AI/ML model, may be determined through a split engine.
- the split engine receives the size of intermediate data transmitted from the device uplink to the server and the amount of computation required to calculate the layers from the device to the split point, determines when the split point needs to be changed, and can adjust the split point.
- the split engine notifies the determination result and the adjustment result to the apparatus and the server, and the apparatus and the server perform the inference according to the adjusted split point.
- FIG. 10 illustrates an example of an operation for determining a split point of a split engine according to an embodiment of the present disclosure.
- the device 1030 may generate the intermediate data 1040 by performing the split inference 1032 using the split AI/ML model assigned to the device.
- the device 1030 may transmit the generated intermediate data 1040 to the server 1010 .
- the server 1010 may perform the split inference 1012 using the split AI/ML model allocated to the network using the received intermediate data 1040 .
- the split engine 1020 may adjust the split point based on the size of the intermediate data 1040 and the amount of computation required for the device 1030 to calculate the layers up to the split point. For example, if the split engine 1020 determines that the device 1030 is suitable for performing split inference on more layers than before, the split engine 1020 may adjust the split point in the direction of the output layer. In addition, if the split engine 1020 determines that the server 1010 is suitable for performing split inference on more layers than before, the split engine 1020 may adjust the split point in the direction of the input layer.
- a method of determining a split point using the split engine shown in FIG. 10 is designed based on the vanilla split AI/ML model of FIG. 8 . Therefore, when determining the split point using the split engine in the vanilla split AI/ML model, only intermediate data transmitted by the device to the base station through the uplink can be considered.
- the U-shaped split AI/ML model differs from the vanilla split AI/ML model in determining and adjusting the split point, not only the intermediate data that the device transmits to the base station via the uplink, but also the intermediate data that the device transmits to the base station via the downlink. Intermediate data passing to the device also needs to be considered.
- the split engine may semi-statically adjust the split point based on the size of the intermediate data and the amount of computation required for the device to calculate the layers from the input layer to the split point.
- communication performance and freshness of information may depend on a transmission state of intermediate data.
- the method of adjusting the split point using the split engine may be difficult to reflect the real-time environment of the intermediate data transmitted periodically and continuously.
- AoI Age of information
- the present disclosure provides AoI of intermediate data transferred between the device and the base station in the uplink and downlink so that the latest information can be maintained when the device and the base station perform inference through the U-shaped split AI/ML model.
- AoI which is an index for split adjustment in the U-shaped split AI/ML model.
- age or another term having an equivalent technical meaning, and may not be limited to a specific name.
- AoI is the difference between the creation time of data and the usage time of the data.
- AoI( ) may be as in Equation 1 below.
- AoI The time elapsed from to the current time t.
- AoI is a measure of the freshness/freshness of information from the point of view of the receiver. The smaller the AoI is, the more recent the packet has arrived at the receiver. That is, the smaller the AoI, the more up-to-date information can be maintained.
- FIG 11 illustrates an example of measuring AoI according to an embodiment of the present disclosure.
- AoI ( ) 1110 may be packet delays 1120 and 1122 and an inter-delivery time 1130 .
- the packet delays 1120 and 1122 are time elapsed from generation to delivery of the packet 1150
- the inter-delivery time 1130 is a new packet 1150 after the packet 1150 is delivered to the receiver 1140 . It may be the time taken until it is transmitted to the receiving side 1140 .
- the packet delays 1120 and 1122 may include at least one of a processing delay, a queuing delay, a transmission delay, and a propagation delay.
- the system AoI ( ) 1110 may be linearly increased until the packet 1150 is delivered to the receiving side 1140 . For example, when the packet 1150 arrives at the receiving side 1140, the system AoI ( ) 1110 can be reduced by as much as packet delays 1120 and 1122 .
- Calculating and minimizing the AoI may be mathematically very complex depending on the application of the service provided. Therefore, the device and the base station may utilize the Peak Age of Information (PAoI), which measures the maximum AoI value immediately before information is received at the receiving side, for split point adjustment. The device and the base station may adjust the split point by comparing the PAoI and an arbitrarily set threshold value.
- PAoI Peak Age of Information
- Cost of Update Delay (C(t)) may be a metric modeling a “dissatisfaction” level due to staleness of information on the receiving side or a “need” level for new information.
- the CoUD metric may be different for each service to be provided or for each application applied to the service. CoUD is expressed by Equation 2 below.
- CoUD is a nonlinear function that does not decrease in Equation 1 for AoI.
- FIG. 12 illustrates an example of measuring AoI using a CoUD metric according to an embodiment of the present disclosure.
- Equation 3 an example of a graph to which a non-linear CoUD metric is applied can be confirmed.
- AoI( ) can be expressed by Equation 3 as follows.
- AoI( ) as a service-dependent nonlinear function
- CoUD Co-dependent nonlinear function
- the U-shaped split model In a system to which the U-shaped split model is applied for data security, multiple devices can perform inference with the base station, and the wireless environment of each device communicating with the base station through uplink and downlink may be different. have. Therefore, in the system to which the U-shaped split model is applied, when the device and the base station adjust the split point, the AoI ( ), but the CoUD metric for the application link of the service is applied. is available.
- FIG. 13 illustrates an example of a graph illustrating evaluation of communication performance of a device according to a candidate split point setting of a VGG-16 model according to an embodiment of the present disclosure.
- Table 2 illustrates the required uplink data rate and split inference latency in the device according to the configured candidate split point when the VGG-16 model operates at 30 FPS. Table 2 shows the results based on the 227x227 input image.
- split point Approximate output data size (MByte) Required UL data rate (Mbps) @30FPS Device-side inference latency (ms) @33FPS Candidate split point 0 (Cloud-based inference) 0.6 145 N/A Candidate split point 1(after pool1 layer) 3 720 55 Candidate split point 2(after pool2 layer) 1.5 360 115 Candidate split point 3(after pool3 layer) 0.8 192 240 Candidate split point 4(after pool4 layer) 0.5 120 390 Candidate split point 5(after pool5 layer) 0.1 24 470 Candidate split point 6 (Device-based inference) N/A N/A 730
- the size of the intermediate data generated by the device performing split inference according to the set split point, the uplink data rate on the radio for transmitting the generated intermediate data to the base station, and the split in the device It can be seen that the inference latencies are different. This means that when the device and the base station perform split inference through the split AI/ML model, all three factors mentioned above must be considered. In addition, in general, since a wireless environment is a lossy environment, distortion and error generated by wirelessly transmitting intermediate data must be considered.
- the split point may be adjusted in the split AI/ML model in consideration of all the factors mentioned above, an excessive burden may occur compared to the computing power of the device.
- the base station can be connected to multiple devices in an N:1 manner, if the split point for transmitting and receiving intermediate data is adjusted for each device in consideration of the aforementioned factors, an excessive burden may occur compared to the computing power of the base station. have.
- the split point since the device and the base station perform inference using intermediate data transmitted through the uplink and downlink, the AoI of the intermediate data has an important influence on the inference performance. Therefore, in order to prevent excessive burden of devices and base stations and loss of data, the split point may be adjusted using AoI, which is an index indicating the freshness/freshness of intermediate data.
- the present disclosure provides an uplink/down link between a device and a base station to improve the freshness/freshness of the inference result in a situation where split inference is performed through a U-shaped split AI/ML model between multiple devices and a base station
- FIG. 14 illustrates an example of a system 1400 supporting a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- a system 1400 may include a plurality of devices 1410 , 1412 , and 1414 and a base station 1420 .
- each of the plurality of devices 1410, 1412, 1414 and the base station 1420 is equipped with an AI/ML model supporting the split AI/ML model, and each of the devices 1410, 1412, 1414 and the base station ( 1420) can hold the full AI/ML model.
- the base station 1420 uses the U-shaped split AI/ML model to provide services (control and status monitoring of devices), two initially set split points and an intermediate transmitted through uplink/downlink A non-decrementing non-linear function of the CoUD metric used to compute the non-linearly increasing AoI of the data.
- each of the devices 1410 , 1412 , and 1414 may be transmitted to each of the devices 1410 , 1412 , and 1414 .
- the devices 1410 , 1412 , 1414 and the base station 1420 perform learning and then apply a CoUD metric from the time point of performing the inference based on the AoI of intermediate data measured to adjust the split point.
- FIG. 15 illustrates an example of an operation of adjusting a split point in a U-shaped AI/ML split model 1500 according to an embodiment of the present disclosure.
- the device 1510 and the base station 1520 may perform a U-shaped split inference formed by setting a first split point 1550 and a second split point 1552 .
- the first split point 1550 may be a first split point established between the device 1510 and the base station 1520 .
- the second split point 1552 may be another split point established between the base station 1520 and the device 1510 in the direction of the output layer from the first split point 1550 .
- the device 1510 and the base station 1520 may perform the U-shaped split inference through the assigned split AI/ML model, respectively.
- the device 1510 may perform split inference on the layers from the input layer to the first split point 1550 .
- the device 1510 may generate the first intermediate data 1560 as a result of performing the split inference up to the first split point 1550 .
- the device 1510 may transmit the generated first intermediate data 1560 to the base station 1520 .
- the base station 1520 performs split inference on layers from the first split point 1550 to the second split point 1552 using the first intermediate data 1560 received from the device 1510 . can do.
- the base station 1520 may generate the second intermediate data 1562 as a result of performing the split inference up to the second split point 1540 .
- the base station 1520 may transmit the generated second intermediate data 1562 to the device 1510 .
- the apparatus 1510 may perform split inference on the layers from the second split point 1552 to the output layer using the second intermediate data 1562 received from the base station 1520 .
- an AI/ML split model for performing U-shaped inference may be allocated to the device 1510 and the base station 1520 after initial setting.
- the device 1510 and the base station 1520 may each perform inference through the assigned AI/ML split model.
- the device 1510 and the base station 1520 may adjust the split point to maintain the latest information by minimizing the AoI.
- the device 1510 and the base station 1520 may measure AoI, which is a reference for split point adjustment.
- the device 1510 and the base station 1520 may measure PAoI, which is the maximum value of AoI at a time point when intermediate data is received, to simplify the AoI calculation.
- the device 1510 transmits the first intermediate data 1560, which is the result of inference for the layers from the input layer 1530 to the first split point 1550, to the base station 1520 in the uplink. It is possible to increase the AoI at this point.
- the device 1510 is a time point when the base station 1520 receives the second intermediate data 1562, which is a result of inference for the layers from the first split point 1550 to the second split point 1552, through the downlink. CoUD metric results in can be checked. calculated when the device 1510 receives the second intermediate data 1562 is the same as in Equation 4 below.
- Equation 4 may be the time at which the base station 1520 most recently received the first intermediate data 1560 from the device 1510 on the uplink at time t. At this time may be the freshest time.
- the base station 1520 may increase the AoI at the time when the U-shaped split inference operation starts.
- the base station 1520 is the result of the CoUD metric at the time when the device 1510 receives the first intermediate data 1560 transmitted through the uplink. can be checked. Calculated when the base station 1520 receives the first intermediate data 1560 is the same as in Equation 5 below.
- Equation 5 may be the time at which raw data of the device 1510 most recently collected/generated from a sensor or the like at time t is transmitted to the input layer. At this time may be the freshest time.
- the device 1510 and the base station 1520 may adjust the split point according to the measured AoI.
- the split point may be adjusted while maintaining the initially set number of split points.
- the first intermediate data 1560 and the second intermediate data 1562 transmitted through the uplink/downlink maintain the minimum AoI. It is assumed that the latest data is transmitted.
- the split point is adjusted in the direction of the input layer 1530 based on the entire model, the split point is up to the hidden layer that the device can pass through an activation function at least once in order to preserve the privacy of the raw data. can be adjusted.
- the split point may be adjusted only before the output layer 1540 in order to maintain the structure of the U-shaped split AI/ML model.
- the procedure for the base station to adjust the split point may be as follows.
- the base station has a threshold value based on the CoUD metric initially set to perform the U-shaped split inference. can be set. may be set in consideration of the inference latency of the device up to the first split point initially set between the device and the base station and a loss rate of the first intermediate data transmitted to the base station through the uplink.
- the base station may receive information necessary for split-point coordination transmitted by the device through the uplink.
- the information required for split point adjustment includes information on when data collected/generated by sensors, monitors, etc. is delivered to the input layer, and the device performs an inference on the layers from the input layer to the first split point. Adjusted based on the AoI measurement value of the first intermediate data that is the generated result, and the second intermediate data that is the result that the base station performs inference on the layers from the first split point to the second split point It may include at least one of location information of the second split point. In this case, when the base station performs the split point adjustment procedure for the first time, since the initially set second split point is not adjusted, the base station may not receive location information of the second split point.
- the base station may check the PAoI through the CoUD metric at the time when information necessary for split inference is received from the device.
- the base station may adjust the first split point according to a result of comparing the checked PAoI value and the threshold value. Also, the base station may perform split inference based on the information received from the device and the first intermediate data.
- the base station Since the base station has the entire split AI/ML model, the size of the first intermediate data resulting from the device performing split inference up to the first split point can be known. Therefore, when the base station transmits information on the reception time of the first intermediate data to the device through the downlink and the second intermediate data that is a result of the base station performing the inference, the adjusted location information of the first split point and the device are Resources that can be transmitted on the next uplink may be delivered together.
- FIG. 16 illustrates an example of an operation in which a base station adjusts a first split point in a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- the base station The first split point may be adjusted according to a result of comparing the 1610 and the PAoI 1620 .
- the base station According to ( 1610 ), the first split point may be adjusted or maintained in the direction of the output layer or the direction of the input layer.
- the base station has a PAoI (1620) value If it belongs to a section smaller than the value (1610), the first split point may be moved or maintained in the direction of the output layer.
- the base station has a PAoI (1620) value in a state where the first split point is located after the first hidden layer. If it belongs to a section smaller than the value (1610), the first split point may be maintained.
- the base station has a PAoI (1620) value If it belongs to a section greater than the value, the first split point may be moved or maintained in the direction of the input layer.
- the position of the first split point may be determined to correspond to Equation 6 below. That is, the position of the first split point may be determined between after the input layer and before the second split point.
- the AoI measured when the base station receives the first intermediate data that is the result of performing the split inference of the device decreases by the packet delay of the first intermediate data, and then gradually increases until another first intermediate data is received can do.
- the procedure for the device to adjust the split point may be as follows.
- the device performs the U-shaped split inference based on the initially set CoUD metric received from the base station, the threshold value can be set. may be set in consideration of the split inference latency of the initially set base station and the loss rate of the second intermediate data, which is a result of the base station performing inference delivered to the device through the downlink.
- the device may transmit information necessary for split point adjustment to the base station through the uplink.
- the information required for split point adjustment includes information on when data collected/generated by devices such as sensors and monitors is delivered to the input layer, and the device has an inference to the layers from the input layer to the first split point. Based on the AoI value of the first intermediate data, which is a result generated by performing It may include at least one of the adjusted location information of the second split point.
- the base station performs the split point adjustment procedure for the first time, the second split point is not adjusted and the initial setting is maintained, so the adjusted location information of the second split point may not be received by the base station.
- the device may receive information on the reception time of the first intermediate data from the base station, the second intermediate data that is a result of the base station performing the inference, and location information of the adjusted first split point.
- the device may check the PAoI through the CoUD metric at the time of receiving the second intermediate data from the base station through the downlink.
- the device may adjust the second split point according to a result of comparing the checked PAoI value and the threshold value.
- the base station receives the first intermediate data and the second intermediate data, which is a result of performing split inference on some layers to which the base station is allocated, from the base station, the second intermediate data is used to Split inference may be performed on layers from the second split point to the output layer.
- the apparatus may perform split inference on new raw data in consideration of the position of the adjusted first split point received from the base station and generate new first intermediate data. Thereafter, the device may transmit information on when new raw data is transmitted to the input layer to the base station, new first intermediate data, and position data of the adjusted second split point to the base station.
- FIG 17 illustrates an example of an operation in which an apparatus adjusts a second split point in a U-shaped split AI/ML model according to an embodiment of the present disclosure.
- the device is The second split point may be adjusted according to a result of comparing the 1710 and the PAoI 1720 .
- the second split point may be adjusted in the output layer direction or the input layer direction, or may be maintained.
- the device has a PAoI (1720) value If it belongs to a section greater than the (1710) value, the second split point may be moved or maintained in the direction of the input layer. That is, the second split point may move to the left.
- the device has a PAoI (1720) value If it belongs to a section smaller than the value, the second split point may be moved or maintained in the direction of the output layer. That is, the second split point may move to the right.
- the PAoI 1720 value is in a state where the second split point is located immediately before the output layer.
- the second split point may be maintained without moving in the direction of the output layer.
- the position of the second split point may be determined to correspond to Equation 7 below. That is, the location of the second split point may be determined between after the second split point and before the output layer.
- the device receives the second intermediate data, which is the result of the split inference of the base station, the AoI measured decreases by the packet delay of the second intermediate data, and then gradually increases until another first intermediate data is received.
- FIG. 18 illustrates an operation in which a base station adjusts a first split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- the following steps may be omitted depending on circumstances and/or settings, and are not limited to specific steps.
- the base station may receive information for adjusting the first split point from the device.
- the first split point may include at least one of a first split point initially set between the device and the base station, or a first split point adjusted in the initially set first split point.
- the information for adjusting the first split point received by the base station includes information on when raw data is input to the input layer, location data of the second split point, and layers from the input layer to the first split point of the device. It may include first intermediate data that is a result of performing the inference on .
- the base station may perform initial setup for performing split inference.
- the initial setting may include at least one of setting a CoUD metric for calculating the AoI of intermediate data, which is a result of performing split inference of the device and the base station, and starting the AoI increase.
- the base station may measure the PAoI of the uplink data at the time of receiving the information for the first split point adjustment.
- the uplink data may include first intermediate data that is a result of inference of the device.
- the measured PAoI value may be a standard for split point adjustment.
- the base station may perform split inference on the layers from the first split point to the second split point using the first intermediate data received from the device.
- the base station may generate the second intermediate data as a result of the split inference.
- the base station may transmit the generated second intermediate data to the device for performing the remaining split inference.
- the base station may adjust the first split point.
- the base station may adjust the first split point by comparing the measured PAoI value of the uplink data and an initially set threshold value. For example, when the PAoI value of the uplink data is less than or equal to the threshold value, the first split point may be moved in the direction of the output layer or may be maintained. In this case, the first split point may be moved in the output layer direction only before the second split point. For example, when the PAoI value of the uplink data is greater than the threshold value, the first split point may be moved or maintained in the input layer direction. In this case, the first split point may be moved in the input layer direction only after the first hidden layer in order to preserve the privacy of the input data.
- the base station may adjust the AoI of the uplink data.
- the base station may compare a time difference between a time point (Di) at which uplink data is received from the device and a time point (Si) at which raw data is input to the input layer.
- the time point at which raw data is input may be a time point at which split inference starts in the device.
- Information on when raw data is input may be transmitted from the device to the base station.
- the base station in the PAoI value of the measured uplink data AoI can be dropped by subtracting .
- the base station may increase the dropped AoI by PAoI again.
- the base station may transmit information for adjusting the second split point to the device.
- the second split point may include at least one of a first split point between the device and the base station initially set after the first split point, or a second split point adjusted in the initial setup.
- the information for adjusting the second split point includes information on the first intermediate data reception time of the base station, location data of the first split point, and the base station interfering with the layer from the first split point to the second split point. It may include at least one of the second intermediate data that is a result of performing the operation. In this case, since the base station has the entire split AI/ML model, the size of the first intermediate data resulting from the device performing split inference up to the first split point can be known.
- the base station transmits information on the reception time of the first intermediate data to the device through the downlink and the second intermediate data that is a result of the base station performing the inference, the adjusted location information of the first split point and the device are Resources that can be transmitted on the next uplink may be delivered together.
- FIG. 19 illustrates an operation in which an apparatus transmits information for split point adjustment to a base station in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- the following steps may be omitted depending on circumstances and/or settings, and are not limited to specific steps.
- the device may check initial information for transmitting information for split point adjustment to the base station.
- the device may determine whether raw data is input to the input layer. If raw data is input to the input layer, the device may record the input time point.
- the device may check whether there is information on the first split point received from the base station. If the first split point information is not received from the base station, the device may use the initially set first split point for which no adjustment is made. Upon receiving the first split point information from the base station, the device may use the first split point reset by reflecting the received first split point information.
- step S1903 after checking the initial information, the device may perform split inference on the layers from the input layer to the first split point. The device may generate first intermediate data as a result of the split inference.
- the device may check the second split point information. As an example, the device may determine whether there is a calculated second split point.
- the calculated second split point may be a second split point adjusted according to an AoI value of downlink data calculated based on a time point when the base station receives information for adjusting the first split point. For example, if the second split point is calculated, the device may use the calculated second split point, and if not calculated, the device may use an initially set second split point.
- the device may transmit information for split point adjustment in the U-shaped split AI/ML model to the base station.
- the information for split inference includes information on when raw data is input to the input layer, location data of the second split point, and the first split inference result obtained by the device performing split inference up to the first split point. It may include at least one of intermediate data.
- the base station may check whether information for split point adjustment has been received from the device, and the subsequent procedure may be as shown in FIG. 18 .
- FIG. 20 illustrates an operation in which an apparatus adjusts a second split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- the following steps may be omitted depending on circumstances and/or settings, and are not limited to specific steps.
- the device may receive information for adjusting the second split point from the base station.
- the second split point may include at least one of a second split point initially set between the device and the base station after the first split point, or a second split point adjusted in the initially set second split point.
- the information for adjusting the second split point includes information on when the base station receives the first intermediate data, the first split point location data, and the base station for layers from the first split point to the second split point. Second intermediate data that is a result of performing the inference may be included.
- step S2003 if the device receives the information for the second split point adjustment from the base station, it may measure the PAoI of the downlink data at the time of receiving the information for the second split point adjustment.
- the downlink data may include second intermediate data that is an inference result of the base station.
- the measured PAoI value may be a standard for split point adjustment.
- step S2005 the device may perform split inference on the layers from the second split point to the output layer using the second intermediate data received from the base station. That is, the apparatus may perform split inference on remaining layers after the base station performs split inference on the layers from the first split point to the second split point. The device may generate the label as a result of performing split inference to the output layer.
- the device may adjust the second split point.
- the device may adjust the second split point by comparing the measured PAoI value of the downlink data with an initially set threshold value.
- the second split point may be moved or maintained in the direction of the output layer. In this case, the second split point may be moved in the direction of the output layer only before the output layer.
- the second split point may be moved or maintained in the input layer direction. In this case, the second split point may be moved in the input layer direction only after the first split point.
- the device may adjust the AoI of the downlink data.
- the apparatus may compare a time difference between a time point (Di) at which the downlink data is received from the base station and a time point (Si) at which the base station receives the first intermediate data from the terminal. Thereafter, the device uses the PAoI value of the measured downlink data.
- AoI can be dropped by subtracting . may be different for each device or for each applied application. Thereafter, the base station may increase the dropped AoI by PAoI again. Thereafter, the base station may increase the dropped AoI by PAoI again.
- the split point adjustment operation of the apparatus and the base station as described in FIGS. 18 and 20 may be repeatedly performed until the base station notifies the end of the service.
- FIG. 21 illustrates an operation in which the device adjusts a split point in a system to which a U-shaped split AI/ML model is applied according to an embodiment of the present disclosure.
- the device may receive at least one of information on a first split point and a second split point from the base station.
- Each of the split points may indicate a point for performing inference through the split AI/ML model.
- the first split point and the second split point may be split points initially set by the base station, respectively.
- at least one of the first split point and the second split point may be a split point adjusted from an initially set split point.
- the device may perform a first split inference based on the first split point and generate first intermediate data.
- the device may perform split inference on the layers from the input layer to the first split point based on the split point information received from the base station.
- the device may generate the first intermediate data as a result of the split inference.
- the device may transmit the first intermediate data to the base station.
- the device may transmit information on a time point when raw data of the device is transmitted to the input layer along with the first intermediate data and information on the second split point to the base station.
- the device may receive the second data generated based on the second split point from the base station.
- the base station may perform split inference using the first intermediate data received from the device.
- the base station may perform split inference on the layers from the first split point to the second split point, and may generate second intermediate data as a result of the split inference.
- the apparatus may receive information about a first intermediate data reception time and information on a first split point received by the base station together with the second intermediate data from the base station.
- the device may perform the remaining split inference from the second split point on the basis of the second intermediate data received from the base station.
- the device may perform split inference on the layers from the second split point to the output layer.
- the device may generate labels as a result of performing the split inference. Privacy of input data and labels can be preserved by allowing the device and the base station to perform split inference on the three-partitioned layers.
- the device may adjust the second split point based on the second intermediate data received from the base station. For example, the device may adjust the second split point by comparing the measured PAoI value of the second intermediate data with a threshold value initially set by the base station. For example, when the PAoI value of the second intermediate data is less than or equal to the threshold value, the second split point may be moved or maintained in the direction of the output layer. In this case, the second split point may be moved in the direction of the output layer only before the output layer. For example, when the PAoI value of the second intermediate data is greater than the threshold value, the second split point may be moved or maintained in the input layer direction. In this case, the second split point may be moved in the input layer direction only after the first split point.
- examples of the above-described proposed method may also be included as one of the implementation methods of the present disclosure, it is clear that they may be regarded as a kind of proposed method.
- the above-described proposed methods may be implemented independently, or may be implemented in the form of a combination (or merge) of some of the proposed methods.
- a rule may be defined so that the base station informs the terminal of whether the proposed methods are applied or not (or information about the rules of the proposed methods) through a predefined signal (eg, a physical layer signal or a higher layer signal). .
- Embodiments of the present disclosure may be applied to various wireless access systems.
- various radio access systems there is a 3rd Generation Partnership Project (3GPP) or a 3GPP2 system.
- 3GPP 3rd Generation Partnership Project
- 3GPP2 3rd Generation Partnership Project2
- Embodiments of the present disclosure may be applied not only to the various wireless access systems, but also to all technical fields to which the various wireless access systems are applied. Furthermore, the proposed method can be applied to mmWave and THz communication systems using very high frequency bands.
- embodiments of the present disclosure may be applied to various applications such as free-running vehicles and drones.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
Description
(pkt/sec) |
[delay] (sec) |
[Inter-delivery] (sec) |
Average AoI (Sec) |
|
① | 0.01 | 1.01 | 100.00 | 101.00 |
② | 0.53 | 2.13 | 1.89 | 3.48 |
③ | 0.99 | 100.00 | 1.01 | 100.02 |
Split point | Approximate output data size (MByte) | Required UL data rate (Mbps) @30FPS |
Device-side inference latency (ms) @33FPS |
Candidate split point 0 (Cloud-based inference) |
0.6 | 145 | N/A |
Candidate split point 1(after pool1 layer) | 3 | 720 | 55 |
Candidate split point 2(after pool2 layer) | 1.5 | 360 | 115 |
Candidate split point 3(after pool3 layer) | 0.8 | 192 | 240 |
Candidate split point 4(after pool4 layer) | 0.5 | 120 | 390 |
Candidate split point 5(after pool5 layer) | 0.1 | 24 | 470 |
Candidate split point 6 (Device-based inference) |
N/A | N/A | 730 |
Claims (14)
- 무선 통신 시스템에서 단말의 동작 방법에 있어서,기지국으로부터 제1 스플릿 포인트(split point) 및 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 수신하는 단계;상기 제1 스플릿 포인트에 기초하여 제1 스플릿 인퍼런스(split inference)를 수행하고, 제1 중간 데이터(intermediate data)를 생성하는 단계;상기 기지국으로 상기 제1 중간 데이터를 송신하는 단계;상기 기지국으로부터 상기 제2 스플릿 포인트에 기초하여 생성된 제2 중간 데이터를 수신하는 단계;상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트부터 나머지 스플릿 인퍼런스를 수행하는 단계; 및상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트를 조정하는 단계를 포함하는, 단말 동작 방법.
- 제 1항에 있어서,상기 제1 스플릿 포인트는, U자형 분할 인공지능 학습 모델에 기초하여 상기 단말이 상기 제1 스플릿 인퍼런스(split inference)를 수행하는 제1 지점을 지시하고,상기 제2 스플릿 포인트는, 상기 U자형 분할 인공지능 학습 모델에 기초하여 상기 기지국이 제2 스플릿 인퍼런스를 수행하는 제2 지점을 지시하는, 단말 동작 방법.
- 제 2항에 있어서,상기 단말은 상기 단말의 로우 데이터(raw data)가 인풋 레이어(input layer)로 전달된 시점에 대한 정보, 상기 제2 스플릿 포인트에 대한 정보를 상기 제1 중간 데이터와 함께 상기 기지국으로 전송하고,상기 기지국이 수신한 상기 제1 중간 데이터 수신 시점에 대한 정보, 상기 제1 스플릿 포인트에 대한 정보를 상기 제2 중간 데이터와 함께 수신하는, 단말 동작 방법.
- 제 3항에 있어서,상기 단말이 상기 제2 스플릿 포인트를 조정하는 단계는,상기 제2 중간 데이터의 PAoI(Peak of AoI)를 측정하고, 측정된 상기 PAoI 값을 상기 기지국이 설정한 제1 임계값과 비교한 결과에 기초하여 상기 제2 스플릿 포인트를 조정하는 단계를 포함하는, 단말 동작 방법.
- 제 4항에 있어서,상기 PAoI 값이 상기 제1 임계값보다 큰 경우 상기 제2 스플릿 포인트를 상기 인풋 레이어 방향으로 이동시키거나 유지하고, 상기 PAoI 값이 상기 제1 임계값보다 작은 경우 상기 제2 스플릿 포인트를 상기 아웃풋 레이어 방향으로 이동시키거나 유지하는, 단말 동작 방법.
- 제 5항에 있어서,상기 제2 스플릿 포인트의 조정 위치는, 상기 제1 스플릿 포인트 이후와 상기 아웃풋 레이어 이전 사이에서 결정되는, 단말 동작 방법.
- 제 2항에 있어서,상기 제1 스플릿 포인트는, 상기 기지국이 상기 단말로부터 수신한 상기 제1 중간 데이터의 PAoI를 측정하고, 측정된 상기 제1 중간 데이터의 PAoI 값이 상기 기지국이 설정한 제2 임계값보다 작은 경우 상기 아웃풋 레이어 방향으로 이동되거나 유지되고, 측정된 상기 제1 중간 데이터의 PAoI 값이 상기 제2 임계값보다 큰 경우 상기 인풋 레이어 방향으로 이동되거나 유지되는, 단말 동작 방법.
- 제 7항에 있어서,상기 제1 스플릿 포인트의 조정 위치는, 인풋 레이어 이후와 상기 제2 스플릿 포인트 이전 사이에서 결정되는, 단말 동작 방법.
- 무선 통신 시스템에서 기지국 동작 방법에 있어서,제1 스플릿 포인트(split point) 및 제2 스플릿 포인트에 대한 초기 설정을 수행하는 단계;단말로 상기 제1 스플릿 포인트 및 상기 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 송신하는 단계;상기 단말로부터 제1 중간 데이터를 수신하는 단계로써, 상기 제1 중간 데이터는 제1 스플릿 포인트에 기초하여 상기 단말이 수행한 제1 스플릿 인퍼런스(split inference)에 기초하여 생성되고;상기 제1 중간 데이터에 기초하여 상기 제2 스플릿 포인트까지 제2 스플릿 인퍼런스를 수행하여 제2 중간 데이터를 생성하는 단계,상기 제1 중간 데이터에 기초하여 상기 제1 스플릿 포인트를 조정하는 단계; 및상기 생성된 제2 중간 데이터 및 조정된 상기 제1 스플릿 포인트에 대한 정보를 상기 단말로 전송하는 단계를 포함하는, 기지국 동작 방법.
- 무선 통신 시스템의 단말에 있어서,송수신기; 및상기 송수신기와 연결된 프로세서를 포함하고,상기 프로세서는,기지국으로부터 제1 스플릿 포인트(split point) 및 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 수신하고,상기 제1 스플릿 포인트에 기초하여 제1 스플릿 인퍼런스(split inference)를 수행하고, 제1 중간 데이터(intermediate data)를 생성하고,상기 기지국으로 상기 제1 중간 데이터를 송신하고,상기 기지국으로부터 상기 제2 스플릿 포인트에 기초하여 생성된 제2 중간 데이터를 수신하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트부터 나머지 스플릿 인퍼런스를 수행하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트를 조정하는, 단말.
- 무선 통신 시스템의 기지국에 있어서,송수신기; 및상기 송수신기와 연결된 프로세서를 포함하며,상기 프로세서는,제1 스플릿 포인트(split point) 및 제2 스플릿 포인트에 대한 초기 설정을 수행하고,단말로 상기 제1 스플릿 포인트 및 상기 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 송신하고,상기 단말로부터 제1 중간 데이터를 수신하고, 상기 제1 중간 데이터는 제1 스플릿 포인트에 기초하여 상기 단말이 수행한 제1 스플릿 인퍼런스(split inference)에 기초하여 생성되고,상기 제1 중간 데이터에 기초하여 상기 제2 스플릿 포인트까지 제2 스플릿 인퍼런스를 수행하여 제2 중간 데이터를 생성하고,상기 제1 중간 데이터에 기초하여 상기 제1 스플릿 포인트를 조정하고,상기 생성된 제2 중간 데이터 및 조정된 상기 제1 스플릿 포인트에 대한 정보를 상기 단말로 전송하는, 기지국.
- 무선 통신 시스템에서 동작 방법에 있어서,기지국이 제1 스플릿 포인트(split point) 및 제2 스플릿 포인트에 대한 초기 설정을 수행하는 단계;상기 기지국이 단말로 상기 제1 스플릿 포인트 및 상기 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 송신하는 단계;상기 단말이 상기 제1 스플릿 포인트에 기초하여 제1 스플릿 인퍼런스(split inference)를 수행하고, 제1 중간 데이터(intermediate data)를 생성하는 단계;상기 단말이 상기 기지국으로 상기 제1 중간 데이터 및 조정된 제 2 스플릿 정보 중 적어도 어느 하나를 송신하는 단계;상기 기지국이 상기 단말로부터 수신한 상기 제1 중간 데이터 및 상기 조정된 제 2 스플릿 정보 중 적어도 어느 하나에 기초하여 상기 제2 스플릿 포인트까지 제2 스플릿 인퍼런스를 수행하여 제2 중간 데이터를 생성하는 단계;상기 기지국이 상기 단말로부터 수신한 상기 제1 중간 데이터에 기초하여 상기 제1 스플릿 포인트를 조정하는 단계;상기 기지국이 상기 생성된 제2 중간 데이터 및 조정한 상기 제1 스플릿 포인트에 대한 정보를 상기 단말로 전송하는 단계;상기 단말이 상기 기지국으로부터 수신한 상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트부터 나머지 스플릿 인퍼런스를 수행하는 단계; 및상기 단말이 상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트를 조정하는 단계를 포함하는, 동작 방법.
- 통신 장치에 있어서,적어도 하나의 프로세서;상기 적어도 하나의 프로세서와 연결되며, 상기 적어도 하나의 프로세서에 의해 실행됨에 따라 동작들을 지시하는 명령어를 저장하는 적어도 하나의 컴퓨터 메모리를 포함하며,상기 동작들은,기지국으로부터 제1 스플릿 포인트(split point) 및 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 수신하고,상기 제1 스플릿 포인트에 기초하여 제1 스플릿 인퍼런스(split inference)를 수행하고, 제1 중간 데이터(intermediate data)를 생성하고,상기 기지국으로 상기 제1 중간 데이터를 송신하고,상기 기지국으로부터 상기 제2 스플릿 포인트에 기초하여 생성된 제2 중간 데이터 및 상기 기지국이 조정한 상기 제1 스플릿 포인트에 대한 정보를 수신하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트부터 나머지 스플릿 인퍼런스를 수행하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트를 조정하는, 통신 장치.
- 적어도 하나의 명령어(instructions)을 저장하는 비-일시적인(non-transitory) 컴퓨터 판독 가능 매체(computer-readable medium)에 있어서,프로세서에 의해 실행 가능한(executable) 상기 적어도 하나의 명령어를 포함하며,상기 적어도 하나의 명령어는, 장치가,기지국으로부터 제1 스플릿 포인트(split point) 및 제2 스플릿 포인트 중 적어도 하나에 대한 정보를 수신하고,상기 제1 스플릿 포인트에 기초하여 제1 스플릿 인퍼런스(split inference)를 수행하고, 제1 중간 데이터(intermediate data)를 생성하고,상기 기지국으로 상기 제1 중간 데이터를 송신하고,상기 기지국으로부터 상기 제2 스플릿 포인트에 기초하여 생성된 제2 중간 데이터를 수신하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트부터 나머지 스플릿 인퍼런스를 수행하고,상기 제2 중간 데이터에 기초하여 상기 제2 스플릿 포인트를 조정하는, 컴퓨터 판독 가능 매체.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP21911297.6A EP4270253A1 (en) | 2020-12-24 | 2021-12-01 | Method and device for adjusting split point in wireless communication system |
KR1020237013061A KR20230074510A (ko) | 2020-12-24 | 2021-12-01 | 무선 통신 시스템에서 스플릿 포인트를 조정하는 방법 및 장치 |
US18/269,437 US20240086743A1 (en) | 2020-12-24 | 2021-12-01 | Method and device for adjusting split point in wireless communication system |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR10-2020-0183638 | 2020-12-24 | ||
KR20200183638 | 2020-12-24 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022139230A1 true WO2022139230A1 (ko) | 2022-06-30 |
Family
ID=82159425
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2021/017979 WO2022139230A1 (ko) | 2020-12-24 | 2021-12-01 | 무선 통신 시스템에서 스플릿 포인트를 조정하는 방법 및 장치 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20240086743A1 (ko) |
EP (1) | EP4270253A1 (ko) |
KR (1) | KR20230074510A (ko) |
WO (1) | WO2022139230A1 (ko) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024011581A1 (zh) * | 2022-07-15 | 2024-01-18 | 华为技术有限公司 | 一种通信方法及装置 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200175361A1 (en) * | 2018-11-30 | 2020-06-04 | Alibaba Group Holding Limited | Partitioning of deep learning inference with dynamic offloading |
KR20200113744A (ko) * | 2019-03-26 | 2020-10-07 | 한국전자통신연구원 | 심층 신경망 분할 방법 및 장치 |
-
2021
- 2021-12-01 US US18/269,437 patent/US20240086743A1/en active Pending
- 2021-12-01 WO PCT/KR2021/017979 patent/WO2022139230A1/ko active Application Filing
- 2021-12-01 EP EP21911297.6A patent/EP4270253A1/en active Pending
- 2021-12-01 KR KR1020237013061A patent/KR20230074510A/ko unknown
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200175361A1 (en) * | 2018-11-30 | 2020-06-04 | Alibaba Group Holding Limited | Partitioning of deep learning inference with dynamic offloading |
KR20200113744A (ko) * | 2019-03-26 | 2020-10-07 | 한국전자통신연구원 | 심층 신경망 분할 방법 및 장치 |
Non-Patent Citations (8)
Title |
---|
3GPP TS 36.212 |
3GPP TS 36.213 |
3GPP TS 36.321 |
3GPPTS 36.211 |
3GPPTS 36.331 |
ANONYMOUS: "3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Study on traffic characteristics and performance requirements for AI/ML model transfer in 5GS (Release 18)", 3GPP STANDARD; TECHNICAL REPORT; 3GPP TR 22.874, no. V0.1.0, 6 October 2020 (2020-10-06), pages 1 - 55, XP051961408 * |
YANG KAI; SHI YUANMING; YU WEI; DING ZHI: "Energy-Efficient Processing and Robust Wireless Cooperative Transmission for Edge Inference", IEEE INTERNET OF THINGS JOURNAL, vol. 7, no. 10, 10 March 2020 (2020-03-10), USA , pages 9456 - 9470, XP011813820, DOI: 10.1109/JIOT.2020.2979523 * |
ZHOU ZHI; CHEN XU; LI EN; ZENG LIEKANG; LUO KE; ZHANG JUNSHAN: "Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing", PROCEEDINGS OF THE IEEE, vol. 107, no. 8, 1 August 2019 (2019-08-01), US , pages 1738 - 1762, XP011738425, ISSN: 0018-9219, DOI: 10.1109/JPROC.2019.2918951 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024011581A1 (zh) * | 2022-07-15 | 2024-01-18 | 华为技术有限公司 | 一种通信方法及装置 |
Also Published As
Publication number | Publication date |
---|---|
US20240086743A1 (en) | 2024-03-14 |
KR20230074510A (ko) | 2023-05-30 |
EP4270253A1 (en) | 2023-11-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022139230A1 (ko) | 무선 통신 시스템에서 스플릿 포인트를 조정하는 방법 및 장치 | |
WO2022250221A1 (ko) | 무선 통신 시스템에서 신호 전송 방법 및 장치 | |
WO2022092859A1 (ko) | 무선 통신 시스템에서 스플릿 포인트 조정을 위한 장치 및 방법 | |
WO2023017881A1 (ko) | 무선 통신 시스템에서 측정 갭을 이용한 측정 결과에 기반하여 핸드오버를 수행하기 위한 장치 및 방법 | |
WO2022050565A1 (ko) | 무선 통신 시스템에서 핸드오버를 위한 장치 및 방법 | |
WO2023013857A1 (ko) | 무선 통신 시스템에서 데이터 학습을 수행하는 방법 및 장치 | |
WO2023017882A1 (ko) | 무선 통신 시스템에서 배터리 효율을 고려하여 무선 접속 기술을 선택하기 위한 장치 및 방법 | |
WO2024010103A1 (ko) | 무선 통신 시스템에서 자원 요청을 수행하는 방법 및 장치 | |
WO2023042937A1 (ko) | 무선 통신 시스템에서 메타렌즈 인공지능 시스템에 기초하여 신호를 송수신하는 방법 및 장치 | |
WO2022119424A1 (ko) | 무선 통신 시스템에서 신호 전송을 위한 장치 및 방법 | |
WO2023286884A1 (ko) | 무선 통신 시스템에서 신호를 송수신하는 방법 및 장치 | |
WO2022260189A1 (ko) | 무선 통신 시스템에서 신호를 송수신하는 방법 및 장치 | |
WO2022092905A1 (ko) | 무선 통신 시스템에서 신호 전송을 위한 장치 및 방법 | |
WO2023120781A1 (ko) | 무선 통신 시스템에서 신호 전송을 위한 장치 및 방법 | |
WO2023120768A1 (ko) | 무선 통신 시스템에서 신호 전송을 위한 장치 및 방법 | |
WO2023042938A1 (ko) | 무선 통신 시스템에서 반사판을 통해 신호를 전송하는 방법 및 장치 | |
WO2023101304A1 (ko) | 무선 통신 시스템에서 통신을 수행하는 방법 및 장치 | |
WO2022231084A1 (ko) | 무선 통신 시스템에서 신호 전송 방법 및 장치 | |
WO2022124729A1 (ko) | 무선 통신 시스템에서 신호 전송을 위한 장치 및 방법 | |
WO2023022251A1 (ko) | 무선 통신 시스템에서 신호 전송 방법 및 장치 | |
WO2023027311A1 (ko) | 무선 통신 시스템에서 핸드오버를 수행하기 위한 장치 및 방법 | |
WO2024122667A1 (ko) | 무선 통신 시스템에서 앙상블 모델 기반의 수신기에 대한 학습을 수행하기 위한 장치 및 방법 | |
WO2024071459A1 (ko) | 무선 통신 시스템에서 신호 송수신 방법 및 장치 | |
WO2024034695A1 (ko) | 무선 통신 시스템에서 송수신 신호를 생성하기 위한 장치 및 방법 | |
WO2023277218A1 (ko) | 무선 통신 시스템에서 반사판을 통해 신호를 전송하는 방법 및 장치 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21911297 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 20237013061 Country of ref document: KR Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18269437 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2021911297 Country of ref document: EP Effective date: 20230724 |