Communication device, data set providing device, training method of AI / ML model, and method for providing information serving as basis for training AI / ML model

Abstract

[Problem] To provide a communication device, a data set providing device, a method for training an AI / ML model, and a method for providing information serving as a basis for training the AI / ML model, whereby signal processing can be performed regardless of differences in the AI / ML model. [Solution] This communication device is provided with: a signal processing unit that includes an AI / ML model; a receiving unit that receives information serving as a training basis from another communication device; and a control unit that configures the AI / ML model on the basis of the information serving as the training basis.

Description

Technical Field

[0001] The present disclosure relates to a communication device, a data set providing device, a method for training an AI / ML model, and a method for providing information on which learning of an AI / ML model is based. Background Art

[0002] Now, in the 3rd Generation Partnership Project, discussions are underway on 5G and 6G as the next generation mobile communication systems.

[0003] In radio access schemes beyond 5G and 6G, further improvements are expected in enhanced mobile broadband (eMBB), massive machine type communications (mMTC), and ultra-reliable low latency communications (URLLC). To achieve these improvements, the use of artificial intelligence / machine learning (AI / ML) models to process signals sent and received via wireless communications is being discussed.

[0004] Citation List

[0005] Non-patent literature

[0006] Non-patent document 1: R1-2203280, “General aspects of AIPHY”, Ericsson, 3GPP TSG-RAN WG1 Meeting #109-e, Electronic Meeting, May 16th-27th, 2021 Summary of the invention

[0007] Technical issues

[0008] For example, when signal processing is performed between the AI / ML model of the terminal device and the AI / ML model of the base station, if the specifications or implementations of the corresponding AI / ML models are different from each other, signal processing between the corresponding AI / ML models may not be properly performed.

[0009] In addition, for example, when signal processing is performed between the AI / ML model of each of a plurality of terminal devices and the AI / ML model of a base station, if the specifications or implementations of the corresponding AI / ML models of the plurality of terminal devices are different from each other due to, for example, differences in terminal vendors, signal processing between the terminal devices and the base station may not be properly performed.

[0010] The present disclosure solves the above-mentioned problems, and an object of the present disclosure is to provide a communication device, a data set providing device, a method for training an AI / ML model, and a method for providing information on which learning of the AI / ML model is based, which enables signal processing regardless of the differences in the AI / ML model.

[0011] Solution to the problem

[0012] A communication device according to the present disclosure includes: a signal processor including an AI / ML model; a receiver receiving information based on learning from other communication devices; and a controller setting the AI / ML model based on the information based on learning.

[0013] In addition, the dataset providing device according to the present disclosure is a dataset providing device that can be commonly accessed by communication devices of different suppliers, and the dataset providing device includes: a storage device, which stores information on which the learning of the AI / ML model is based; a receiver, which receives a request for providing information on which the learning of the AI / ML model is based from the communication device; and a transmitter, which sends the information on which the learning of the AI / ML model is based to the communication device.

[0014] In addition, the method for training an AI / ML model according to the present disclosure includes: receiving information on which learning is based; and setting the AI / ML model based on the information on which learning is based.

[0015] In addition, the method for providing information based on which learning of an AI / ML model is based according to the present disclosure includes: receiving a request to provide information based on which learning is based; and sending the information based on which learning is based to a sending source that has sent the request for providing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The configuration of the wireless communication system according to the first embodiment is illustrated.

[0017] Figure 2 The diagram illustrates the composition of the management device.

[0018] Figure 3 The diagram illustrates the composition of a base station.

[0019] Figure 4 The diagram illustrates the composition of a relay station.

[0020] Figure 5 The diagram illustrates the composition of the terminal.

[0021] Figure 6 It is a block diagram illustrating the detailed configuration of a terminal device and a base station.

[0022] Figure 7 The diagram illustrates the composition of the data set providing device.

[0023] Figure 8 The diagram illustrates an example of the configuration of a wireless communication network including a data set providing device.

[0024] Fig. 9 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in system information and transmits the system information to each of different terminal devices.

[0025] Fig.10 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in quasi-static information (RRC signaling) and sends the quasi-static information to each of different terminal devices.

[0026] Fig.11 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in each of system information and quasi-static information (RRC signaling), and sends the system information and quasi-static information to each of different terminal devices.

[0027] Fig.12 This is a diagram for describing an example of the following process: a base station receives information on which learning of an AI / ML model is based from a data set providing device; includes the information on which learning of the AI / ML model is based in each of system information and quasi-static information (RRC signaling); and sends the system information and the quasi-static information to each of different terminal devices.

[0028] Fig.13 is a diagram for describing an example of a process in which a base station and a terminal device respectively receive information on which learning of an AI / ML model is based from a dataset providing device.

[0029] Fig.14A is a diagram for describing an example of a process in which a terminal device switches to a different base station.

[0030] Fig. 14B is a diagram for describing an example of a process in which a terminal device switches to a different base station. DETAILED DESCRIPTION

[0031] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In the accompanying drawings, similar or corresponding structural elements will be represented by the same reference numerals, and their detailed description will be appropriately omitted.

[0032] [First embodiment]

[0033] Figure 1The diagram illustrates the configuration of a wireless communication system 1 according to a first embodiment of the present disclosure. The wireless communication system 1 includes a management device 10, a base station 20, a relay station 30, and a terminal device 40. The wireless communication system 1 provides a user with a wireless network that enables mobile communication by cooperatively operating the wireless communication devices included in the wireless communication system 1. The wireless network of the first embodiment includes a radio access network RAN ​​and a core network CN. In the first embodiment, the wireless communication device is a device that includes a function of wireless communication, and the base station 20, the relay station 30, and the terminal device 40 correspond to Figure 1 A wireless communication device in the example shown in the diagram.

[0034] The wireless communication system 1 may include a plurality of management devices 10, a plurality of base stations 20, a plurality of relay stations 30, and a plurality of terminal devices 40. Figure 1 In the example shown in the diagram, the wireless communication system 1 includes management devices 10a and 10b corresponding to the management device 10, and includes base stations 20a, 20b, and 20c corresponding to the base station 20. In addition, the wireless communication system 1 includes relay stations 30a and 30b as relay stations 30, and includes terminal devices 40a, 40b, and 40c as terminal devices 40.

[0035] Figure 1 Each wireless communication device illustrated in the figure can be regarded as a device in a logical sense. In other words, some of the corresponding wireless communication devices can be implemented by virtual machines (VMs), containers, or container engines, and are respectively installed on physically the same hardware.

[0036] The wireless communication system 1 may support a radio access technology (RAT) such as Long Term Evolution (LTE) or New Radio (NR). LTE and NR are each a type of cellular communication technology, and enable mobile communication of the terminal device 40 by arranging a plurality of areas covered by the base station 20 in the form of a cell.

[0037] The radio access scheme of the wireless communication system 1 is not limited to, for example, LTE or NR, and may be other radio access schemes such as Wideband Code Division Multiple Access (W-CDMA) or Code Division Multiple Access 2000 (cdma2000).

[0038] The base station 20 and the relay station 30 included in the wireless communication system 1 may be a ground station or a non-ground station. The non-ground station may be a satellite station or an aircraft station. When the non-ground station is a satellite station, the wireless communication system 1 may be a bent pipe (transparent) mobile satellite communication system.

[0039] In the first embodiment, a ground station (also referred to as a ground base station) refers to a base station (including a "relay station") installed on the ground. Here, the term "ground" has a broad meaning, including not only "on land" but also "underground", "above water" and "underwater". Note that in the following description, the term "ground station" can be replaced by the term "gateway".

[0040] Note that the base station of LTE may be referred to as an evolved Node B (eNodeB) or eNB. In addition, the base station of NR may be referred to as a gNodeB or gNB. In LTE and NR, a terminal device (also referred to as a "mobile station" or "terminal") may be referred to as a user equipment (UE).

[0041] In the first embodiment, the concept of wireless communication device includes not only portable mobile body devices (terminal devices) such as mobile terminals, but also devices installed in structures or mobile bodies. The structure or mobile body itself can be regarded as a wireless communication device. In addition, the concept of wireless communication device includes not only terminal devices 40, but also base stations 20 and relay stations 30. The wireless communication device is a kind of processing device or a kind of information processing device. The wireless communication device can also be selectively expressed as a transmitting device or a receiving device.

[0042] The following specifically describes the configuration of the wireless communication device included in the wireless communication system 1. Note that the configuration of the respective wireless communication devices described below is merely an example. The configuration of the wireless communication device may be different from the configuration described below.

[0043] (Configuration of management device)

[0044] The management device 10 is a device for managing a wireless network. For example, the management device 10 is a device for managing communications performed by the base station 20. When the core network CN is an evolved packet core (EPC), the management device 10 includes, for example, the function of a mobility management entity (MME). When the core network CN is a 5G core network (5GC), the management device 10 includes, for example, the functions of an access and mobility management function (AMF) and / or a session management function (SMF). Note that the functions of the management device 10 are not limited to MME, AMF, and SMF. The management device 10 may be a device including the functions of a network slice selection function (NSSF), an authentication server function (AUSF), or a unified data management (UDM). The management device 10 may be a device including the functions of a home subscriber server (HSS).

[0045] The management device 10 may include the function of a gateway. When the core network CN is an EPC, the management device 10 may include the function of a service gateway (S-GW) or a packet data network gateway (P-GW). When the core network CN is a 5GC, the management device 10 may include the function of a user plane function (UPF). The management device 10 does not necessarily have to be a device included in the core network CN. When the core network CN is a core network of wideband code division multiple access (W-CDMA) or code division multiple access 2000 (cdma2000), the management device 10 may be a device acting as a radio network controller (RNC).

[0046] Figure 2 The diagram illustrates the configuration of the management device 10 according to the first embodiment. The management device 10 includes a communication unit 11, a storage device 12, and a controller 13. Note that Figure 2 The diagram shows the functional components, while the hardware components may differ from Figure 2 The configuration shown in the diagram is different. In addition, the functions of the management device 10 can be statically or dynamically distributed to multiple physically separated configurations. The management device 10 can include multiple server devices.

[0047] The communication unit 11 is a communication interface for communicating with a wireless communication device (such as a base station 20 or a relay station 30). The communication unit 11 may be a network interface or a device connection interface. The communication unit 11 may be a local area network (LAN) interface such as a network interface card (NIC), or may be a USB interface including, for example, a universal serial bus (USB) host controller or a USB port. The communication unit 11 may be a wired interface or a wireless interface. The communication unit 11 serves as a communication device of the management device 10. The communication unit 11 is controlled by a controller 13.

[0048] The storage device 12 is a data readable / writable storage device such as a dynamic random access memory (DRAM), a static random access memory (SRAM), a flash memory, or a hard disk. The storage device 12 acts as a storage device of the management device 10. For example, the storage device 12 stores the connection status of the terminal device 40. The storage device 12 stores therein the status of the radio resource control (RRC) of the terminal device 40 and the status of the EPS connection management (ECM) or the 5G system connection management (CM). The storage device 12 can act as a local memory that stores location information about the location of the terminal device 40.

[0049] The controller 13 is a controller that controls each component of the management device 10. The controller 13 can be implemented by a processor such as a central processing unit (CPU) or a microprocessing unit (MPU). In particular, the controller 13 can be implemented by the processor using a random access memory (RAM) or the like as a work area to execute various programs stored in a storage device located inside the management device 10. The controller 13 can be implemented by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). Any of the CPU, MPU, ASIC, and FPGA can be regarded as a controller.

[0050] (Base station configuration)

[0051] The base station 20 is a wireless communication device that performs wireless communication with the terminal device 40. The base station 20 may perform wireless communication with the terminal device 40 via the relay station 30, or may perform wireless communication with the terminal device 40 directly.

[0052] The base station 20 is a device corresponding to a wireless base station (such as a base station, a node B, an eNB, or a gNB) or a wireless access point (access point). The base station 20 may be a wireless relay station. The base station 20 may be an optical extension device called a remote radio head (RRH). The base station 20 may be a receiving station such as a field pickup unit (FPU). The base station 20 may be an integrated access and backhaul (IAB) donor node or an IAB relay node that provides a wireless access link and a wireless backhaul link using time division multiplexing, frequency division multiplexing, or space division multiplexing.

[0053] The wireless access technology used by the base station 20 may be a cellular communication technology. The wireless access technology used by the base station 20 may be a wireless LAN technology. The wireless access technology used by the base station 20 may be a low power wide area (LPWA) communication technology. Note that the wireless access technology used by the base station 20 is not limited thereto and may be other wireless access technologies. The wireless communication used by the base station 20 may be wireless communication using millimeter waves. The wireless communication used by the base station 20 may be wireless communication using radio waves, or wireless communication using infrared or visible light, that is, optical wireless communication.

[0054] The base station 20 may be capable of non-orthogonal multiple access (NOMA) communications with the terminal device 40 . NOMA communications refer to communications (transmission, reception, or both) using non-orthogonal resources. The base station 20 may be capable of NOMA communications with other base stations 20 .

[0055] The base station 20 may be able to communicate with the core network CN through an interface between the base station and the core network CN, such as an S1 interface. The interface may be a wired or wireless interface. The base station 20 may be able to communicate with other base stations through an interface between base stations, such as an X2 interface. The interface may be a wired or wireless interface.

[0056] The concept of a base station (also referred to as a "base station device") includes not only a donor base station but also a relay base station (also referred to as a "relay station"). The concept of a base station includes not only a structure having the function of a base station but also a device installed in the structure.

[0057] Examples of structures include buildings such as high-rise buildings, houses, towers, station facilities, airport facilities, port facilities, and stadiums. The concept of structure includes not only buildings, but also non-building structures such as tunnels, bridges, dams, walls, and iron columns; and facilities such as cranes, gates, and windmills. The concept of structure includes not only structures on land (on the ground in a narrow sense) or underground, but also structures on water such as docks or giant floating bodies, and underwater structures such as ocean observation facilities. A base station may alternatively be described as an information processing device.

[0058] The base station 20 may be a fixed station or a movable wireless communication device, i.e., a mobile station. The base station 20 may be a device installed in a mobile body, or may be the mobile body itself. A relay station having mobility may be regarded as a base station 20 acting as a mobile station. Devices that are inherently mobile and include at least a portion of the functions of a base station, such as vehicles, unmanned aerial vehicles (UAVs) represented by drones, or smart phones, may also be regarded as base stations 20 acting as mobile stations.

[0059] The moving object may be a mobile terminal such as a smart phone or a mobile phone. The moving object may be a moving object moving on land (ground in a narrow sense) (vehicles such as cars, bicycles, buses, trucks, motorcycles, trains, and maglev trains), or a moving object moving underground (e.g., in a tunnel) (such as a subway).

[0060] The moving body may be a moving body that moves on water (such as ships including passenger ships, cargo ships, and hovercraft), or a moving body that moves underwater (such as a submersible including a submersible, a submarine, and an unmanned submarine).

[0061] The mobile object may be a mobile object moving in the Earth's atmosphere (such as an aircraft including an airplane, an airship, and a drone).

[0062] The base station 20 may be a ground base station (ground station) installed on the ground. The base station 20 may be a base station set to a structure on the ground, or may be a base station installed in a mobile body moving on the ground. The base station 20 may be an antenna installed in a structure such as a building and a signal processing device connected to the antenna. The base station 20 may be a structure or a mobile body itself. The term "on the ground" has a broad meaning and includes not only "on land" ("on the ground" in a narrow sense), but also "underground", "on water" and "underwater". The base station 20 is not limited to a ground base station. When the wireless communication system 1 is a satellite communication system, the base station 20 may be an aircraft station. From the perspective of a satellite station, an aircraft station located on the earth is a ground station.

[0063] The base station 20 is not limited to a ground station. The base station 20 may be a non-ground base station (non-ground station) capable of floating in the air or in space. The base station 20 may be an aircraft station or a satellite station.

[0064] A satellite station is a satellite station capable of floating outside the Earth's atmosphere. A satellite station may be a device included in a space mobile body such as an artificial satellite, or may be the space mobile body itself. A space mobile body is a mobile body that moves outside the Earth's atmosphere. Examples of space mobile bodies include artificial celestial bodies such as artificial satellites, spacecraft, space stations, and probes.

[0065] The satellite acting as a satellite station may be any of a low earth orbit (LEO) satellite, a medium earth orbit (MEO) satellite, a geosynchronous orbit (GEO) satellite, and a highly elliptical orbit (HEO) satellite. The satellite station may be a device included in a low earth orbit satellite, a medium earth orbit satellite, a geosynchronous orbit satellite, or a highly elliptical orbit satellite.

[0066] An aircraft station is a wireless communication device capable of floating in the earth's atmosphere, such as an aircraft. The aircraft station may be a device included in, for example, an aircraft, or may be the aircraft itself. The concept of an aircraft includes not only heavy aircraft such as airplanes and gliders, but also light aircraft such as balloons and airships. The concept of an aircraft includes not only heavy aircraft and light aircraft, but also rotorcraft such as helicopters and rotorcraft. An aircraft station or an aircraft including an aircraft station may be an unmanned aircraft such as a drone.

[0067] The concept of unmanned aircraft also includes unmanned aircraft systems (UAS) and tethered UAS. The concept of unmanned aircraft includes lighter-than-air (LTA) UAS and heavier-than-air (HTA) UAS. The concept of unmanned aircraft also includes high-altitude UAS platforms (HAP).

[0068] The coverage of base station 20 may be relatively large like a macro cell, or relatively small like a pico cell. The coverage of base station 20 may be extremely small like a femto cell. Base station 20 may include beamforming functionality. For base station 20, a cell or service area may be formed for each beam.

[0069] Figure 3 The diagram illustrates the configuration of the base station 20 according to the first embodiment. The base station 20 includes a wireless communication unit 21, a storage device 22, a decoder 23, and a controller 25. Note that Figure 3 The diagram shows the functional components, while the hardware components may differ from Figure 3 In addition, the functions of the base station 20 may be distributed to a plurality of physically separate configurations.

[0070] The wireless communication unit 21 is a signal processor for wirelessly communicating with other wireless communication devices (such as a relay station 30, a terminal device 40 or another base station 20). The wireless communication unit 21 is controlled by a controller 25. The wireless communication unit 21 supports at least one radio access scheme. The wireless communication unit 21 can support both NR and LTE. In addition to NR and LTE, the wireless communication unit 21 can also support, for example, W-CDMA and cdma2000. The wireless communication unit 21 can support automatic retransmission technologies such as hybrid automatic repeat request (HARQ).

[0071] The wireless communication unit 21 includes a transmitter 211, a receiver 212, and an antenna 213. The wireless communication unit 21 may include a plurality of transmitters 211, a plurality of receivers 212, and a plurality of antennas 213. When the wireless communication unit 21 supports a plurality of radio access schemes, the corresponding components of the wireless communication unit 21 may be separately configured for each radio access scheme. The transmitter 211 and the receiver 212 may be separately configured for LTE and NR, respectively. The antenna 213 may include a plurality of antenna elements, such as a plurality of patch antennas. The wireless communication unit 21 may include a beamforming function. The wireless communication unit 21 may include a polarization beamforming function using a vertically polarized wave (V-polarized wave) and a horizontally polarized wave (H-polarized wave).

[0072] The transmitter 211 performs a process of transmitting downlink control information and downlink data. For example, first, the transmitter 211 encodes the downlink control information and downlink data input from the controller 24 using a coding scheme such as block coding, convolution coding, or turbo coding. As coding, coding using a polarization code or coding using a low-density parity check (LDPC) code may be performed.

[0073] Next, the transmitter 211 modulates the coded bits according to a specified modulation scheme such as BPSK, QPSK, 16QAM, 64QAM or 256QAM. In this case, the signal points on the constellation do not necessarily have to be equidistant. In other words, the constellation can be a non-uniform constellation (NUC).

[0074] Next, the transmitter 211 multiplexes the modulation symbols and downlink reference signals of each channel, and provides the multiplexed modulation symbols and downlink reference signals to the specified resource elements. Next, the transmitter 211 performs various signal processing on the multiplexed signal. For example, the transmitter 211 performs such processing as conversion to the frequency domain using fast Fourier transform, addition of a guard interval (cyclic prefix), generation of a baseband digital signal, conversion to an analog signal, orthogonal modulation, up-conversion, removal of an extra frequency component, and power amplification. Finally, the signal generated by the transmitter 211 is transmitted via the antenna 213.

[0075] The receiver 212 processes the uplink signal received through the antenna 213. For example, first, the receiver 212 performs down-conversion, removal of unnecessary frequency components, control of amplification level, orthogonal demodulation, conversion into a digital signal, removal of a guard interval (cyclic prefix), extraction of a frequency domain signal using a fast Fourier transform, etc. on the uplink signal.

[0076] Next, the receiver 212 separates uplink channels such as a physical uplink shared channel (PUSCH) and a physical uplink control channel (PUCCH) and an uplink reference signal from the signal to which these processes have been applied. Next, the receiver 212 demodulates the received signal using the modulation symbols of the uplink channel according to a modulation scheme such as binary phase shift keying (BPSK) or quadrature phase shift keying (QPSK). The modulation scheme may be 16-quadrature amplitude modulation (QAM), 64QAM, or 256QAM. In this case, the signal points on the constellation do not necessarily have to be equidistant. In other words, the constellation may be a non-uniform constellation.

[0077] Next, the receiver 212 performs a decoding process on the coded bits of the demodulated uplink channel. Finally, the decoded uplink data and uplink control information are output to the controller 24.

[0078] Antenna 213 is an antenna device that performs conversion between electric current and radio waves. Antenna 213 may include a single antenna element, i.e., for example, a single patch antenna. Antenna 213 may include multiple antenna elements, i.e., for example, multiple patch antennas. When antenna 213 includes multiple antenna elements, wireless communication unit 21 may include a beamforming function. Wireless communication unit 21 may be configured to generate a directional beam by controlling the directivity of a radio signal using multiple antenna elements. Antenna 213 may be a dual-polarized antenna. When antenna 213 is a dual-polarized antenna, wireless communication unit 21 may use vertically polarized waves (V-polarized waves) and horizontally polarized waves (H-polarized waves) when sending radio signals. Wireless communication unit 21 may control the directivity of radio signals sent using vertically polarized waves and horizontally polarized waves.

[0079] The storage device 22 is a data readable / writable storage device such as a DRAM, an SRAM, a flash memory, or a hard disk. The storage device 22 serves as a storage device of the base station 20.

[0080] The decoder 23 includes a function of performing signal processing using an AI / ML model. The AI / ML model is a neural network model obtained by machine learning or deep learning. Examples of neural network models may include a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, and a long short-term memory (LSTM) model. The AI / ML model may be any of these models, or may be obtained by combining these models in series or in parallel.

[0081] The decoder 23 restores and outputs the first data by performing signal processing using an AI / ML model in which the second data is used as an input. The decoder 23 may be implemented by a processor such as a CPU or an MPU. The decoder 23 may be implemented by an integrated circuit such as an ASIC or an FPGA. The configuration of the decoder 23 will be described later. Figure 6 Describe in more detail.

[0082] The controller 25 is a controller that controls each component of the base station 20. The controller 25 can be implemented by a processor such as a CPU or an MPU. In particular, the controller 25 can be implemented by the processor using a RAM or the like as a work area to execute various programs stored in a storage device located inside the base station 20. The controller 25 can be implemented by an integrated circuit such as an ASIC or an FPGA. Any of the CPU, MPU, ASIC, and FPGA can be regarded as a controller. In addition to or instead of the CPU, the controller 25 can also be implemented by a graphics processing unit (GPU).

[0083] Note that in some embodiments, the base station 20 may include a collection of multiple physical or logical devices. For example, the base station 20 according to the present embodiment may include multiple different types of devices, such as a baseband unit (BBU) and a radio unit (RU). The base station 20 may be understood as a collection of multiple different types of devices. In addition, the base station 20 may be one or both of the BBU and the RU. The BBU and the RU may be connected to each other via a specified interface such as an enhanced common public radio interface (eCPRI).

[0084] The RU may alternatively be expressed as a remote radio unit (RRU) or a radio point (RD). The RU may correspond to a gNB distributed unit (gNB-DU) described later. The BBU may correspond to a gNB central unit (gNB-CU) described later. The RU may be a device formed integrally with an antenna. The antenna of the base station 20, that is, for example, an antenna formed integrally with the RU, may adopt an advanced antenna system and may support MIMO or beamforming such as FD-MIMO. The antenna of the base station 20 may include, for example, 64 transmit antenna ports and 64 receive antenna ports.

[0085] The antenna included in the RU may be an antenna panel including at least one antenna element, and the RU may include at least one antenna panel. The RU may include two types of antenna panels, namely, an antenna panel for horizontally polarized waves and an antenna panel for vertically polarized waves. The RU may include two types of antenna panels, namely, an antenna panel for right-hand circularly polarized waves and an antenna panel for left-hand circularly polarized waves. The RU may form an independent beam for each antenna panel and control the formed beam.

[0086] The base stations 20 may be connected to each other. At least one base station 20 may be included in a radio access network (RAN). In this case, the base station 20 may be simply referred to as a RAN, a RAN node, an access network (AN), or an AN node. The RAN in LTE may be referred to as an Enhanced Universal Terrestrial RAN (EUTRAN). The RAN in NR may be referred to as an NGRAN. The RAN in W-CDMA (UMTS) may be referred to as an UTRAN.

[0087] The base station 20 of LTE may be referred to as an evolved Node B (eNodeB) or an eNB. In this case, the EUTRAN includes at least one eNodeB (eNB). The base station 20 of NR may be referred to as a gNodeB or a gNB. In this case, the NGRAN includes at least one gNB. The EUTRAN may include a gNB (en-gNB) connected to a core network (EPC) in an LTE communication system (EPS). The NGRAN may include an ng-eNB connected to a core network 5GC in a 5G communication system (5GS).

[0088] When the base station 20 is, for example, an eNB or a gNB, the base station 20 may be referred to as a 3GPP access. When the base station 20 is a wireless access point (access point), the base station 20 may be referred to as a non-3GPP access. The base station 20 may be an optical extension device called a remote radio head (RRH). When the base station 20 is a gNB, the base station 20 may be obtained by combining the above-mentioned gNB-CU and gNB-DU, or may be one of the gNB-CU and gNB-DU.

[0089] In order to communicate with the UE, the gNB-CU hosts multiple upper layers in the access layer (such as RRC, SDAP, and PDCP). The gNB-DU hosts multiple lower layers in the access layer (such as RLC, MAC, and PHY). Among the messages / information described later, RRC signaling (quasi-static reporting) can be generated by the gNB-CU, while MAC CE and DCI (dynamic reporting) can be generated by the gNB-DU. Alternatively, among the RRC configuration (quasi-static reporting), part of the configuration such as IE: cellGroupInfig can be generated by the gNB-DU, while the remaining configuration can be generated by the gNB-CU. These configurations can be sent and received through the F1 interface described later.

[0090] The base station 20 may be able to communicate with other base stations. When multiple base stations 20 are obtained by combining eNBs or by combining eNBs and en-gNBs, the base stations 20 in the multiple base stations 20 may be connected to each other using an X2 interface. When multiple base stations 20 are obtained by combining gNBs or by combining gn-eNBs and gNBs, the base stations 20 in the multiple base stations 20 may be connected to each other using an Xn interface. When multiple base stations are obtained by combining gNB-CUs and gNB-DUs, the base stations 20 in the multiple base stations 20 may be connected to each other through the above-mentioned F1 interface. The messages / information described later (such as RRC signaling, MAC control elements (MACCE) or DCI) may be sent between multiple base stations 20 through, for example, an X2 interface, an Xn interface or an F1 interface.

[0091] The cell provided by the base station 20 may be referred to as a serving cell. The concept of a serving cell includes a primary cell (PCell) and a secondary cell (SCell). When a dual connection is provided to a terminal device 40, a PCell provided by a master node (MN) and zero SCells or at least one SCell may be referred to as a primary cell group. Examples of dual connections include EUTRA-EUTRA dual connection, EUTRA-NR dual connection (ENDC), EUTRA-NR dual connection with 5GC, NR-EUTRA dual connection (NEDC), and NR-NR dual connection.

[0092] The serving cell may include a primary secondary cell or a primary SCG cell (PSCell). When dual connectivity is provided to the terminal device 40, the PSCell provided by the secondary node (SN) and zero SCell or at least one SCell may be referred to as a secondary cell group (SCG). Unless special settings have been made (such as PUCCH on the SCell), the physical uplink control channel (PUCCH) is sent in the PCell and PSCell, but not in the SCell. Radio link failure is detected in the PCell and PSCell, but not in the SCell (no need to detect). As described above, the PCell and PSCell play a special role in the serving cell and are therefore also referred to as special cells (SpCells).

[0093] A single downlink component carrier and a single uplink component carrier may be associated with a single cell. The system bandwidth corresponding to a single cell may be divided into a plurality of bandwidth parts (BWPs). In this case, at least one BWP may be set for the terminal device 40, and the terminal device 40 may use one BWP as an active BWP. Radio resources such as frequency bands, parameter sets (subcarrier spacing), and time slot formats (time slot configurations) that may be used by the terminal device 40 may be different for the respective cells, for the respective component carriers, or for the respective BWPs.

[0094] (Relay Station Configuration)

[0095] The relay station 30 is a wireless communication device that acts as a relay device of the base station 20. The relay station 30 is a type of base station. The relay station 30 is a type of information processing device. The relay station may alternatively be expressed as a relay base station.

[0096] The relay station 30 may be capable of NOMA communication with the terminal device 40. The relay station 30 relays the communication between the base station 20 and the terminal device 40. The relay station 30 may be capable of wireless communication with other relay stations 30 and the base station 20. The relay station 30 may be a ground station device or a non-ground station device. The relay station 30 forms a radio access network RAN ​​together with the base station 20.

[0097] The relay station 30 may be a fixed device, a movable device or a floatable device. The size of the coverage area of ​​the relay station 30 is not limited to a specific size. The cell covered by the relay station 30 may be a macro cell, a micro cell or a small cell.

[0098] The relay station 30 may be included in any device in which the relay function of the relay station 30 is satisfied. The relay station 30 may be included in a terminal device such as a smart phone, may be included in, for example, a car, a train, or a rickshaw, may be included in, for example, a balloon, an airplane, or a drone, or may be included in a home appliance such as a television, a game console, an air conditioner, a refrigerator, or a lighting fixture.

[0099] The relay station 30 may have a configuration similar to that of the above-mentioned base station 20. As in the case of the above-mentioned base station 20, the relay station 30 may be a device installed in a mobile body, or may be the mobile body itself. As described above, the mobile body may be a mobile terminal such as a smart phone or a mobile phone. The mobile body may be a mobile body that moves on land (on the ground in a narrow sense), or a mobile body that moves underground. The mobile body may be a mobile body that moves on water, or a mobile body that moves underwater. The mobile body may be a mobile body that moves in the Earth's atmosphere, or a mobile body that moves outside the Earth's atmosphere. The relay station 30 may be a ground station device or a non-ground station device. The relay station 30 may be, for example, an aircraft station or a satellite station.

[0100] As in the case of the base station 20, the coverage of the relay station 30 can be as large as a macro cell, or as small as a pico cell. The coverage of the relay station 30 can be as small as a femto cell. The relay station 30 may include a beamforming function. For the relay station 30, a cell or service area may be formed for each beam.

[0101] Figure 4 The diagram illustrates the configuration of the relay station 30 according to the first embodiment. The relay station 30 includes a wireless communication unit 31, a storage device 32, a decoder 33, and a controller 35. Note that Figure 4 The diagram shows the functional components, while the hardware components may differ from Figure 4 In addition, the functions of the relay station 30 may be distributed to a plurality of physically separate structures.

[0102] The wireless communication unit 31 is a signal processor for wirelessly communicating with other wireless communication devices (such as the base station 20, the terminal device 40 or other relay stations 30). The wireless communication unit 31 supports at least one radio access scheme. The wireless communication unit 31 can support both NR and LTE. In addition to NR and LTE, the wireless communication unit 31 can also support W-CDMA and cdma3000.

[0103] The wireless communication unit 31 includes a transmitter 311, a receiver 312, and an antenna 313. The wireless communication unit 31 may include a plurality of transmitters 311, a plurality of receivers 312, and a plurality of antennas 313. When the wireless communication unit 31 supports a plurality of radio access schemes, the corresponding components of the wireless communication unit 31 may be separately configured for each radio access scheme. The transmitter 311 and the receiver 312 may be separately configured for LTE and NR, respectively. The transmitter 311, the receiver 312, and the antenna 313 respectively have configurations similar to the corresponding configurations of the transmitter 211, the receiver 212, and the antenna 213 of the above-mentioned base station 20. As in the case of the wireless communication unit 21 of the base station 20, the wireless communication unit 31 may include a beamforming function.

[0104] The storage device 32 is a data readable / writable storage device such as a DRAM, an SRAM, a flash memory, or a hard disk. The storage device 32 serves as a storage device of the relay station 30.

[0105] The decoder 33 includes a function of performing signal processing using an AI / ML model. The AI / ML model is a neural network model obtained by machine learning or deep learning. Examples of neural network models may include a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, and a long short-term memory (LSTM) model. The AI / ML model may be any of these models, or may be obtained by combining these models in series or in parallel.

[0106] The decoder 33 restores and outputs the first data by performing signal processing using an AI / ML model using the second data as input, wherein the second data is received from a terminal device 40 described later. The decoder 33 may have a configuration and function similar to those of the decoder 23 of the above-described base station 20, respectively.

[0107] The controller 34 is a controller that controls each component of the relay station 30. The controller 34 can be implemented by a processor such as a CPU or an MPU. Specifically, the controller 34 can be implemented by the processor using a RAM or the like as a work area to execute various programs stored in a storage device located inside the relay station 30. The controller 34 can be implemented by an integrated circuit such as an ASIC or an FPGA. All of the CPU, MPU, ASIC, and FPGA can be regarded as controllers. In addition to or instead of the CPU, the controller 34 can also be implemented by a GPU.

[0108] Note that the relay station 30 may be an IAB relay node. The relay station 30 serves as an IAB mobile terminal (IAB-MT) for an IAB donor node that is provided with backhaul, and serves as an IAB distributed unit (IAB-DU) for a terminal device 40 that is provided with access. The IAB donor node may be, for example, a base station 20, and serves as an IAB central unit (IAB-CU).

[0109] (Terminal Device Configuration)

[0110] The terminal device 40 is a wireless communication device that wirelessly communicates with other wireless communication devices (such as the base station 20, the relay station 30 or other terminal devices 40). Examples of the terminal device 40 may include a mobile phone, a smart device (a smart phone or a tablet), a personal digital assistant (PDA), and a personal computer. The terminal device 40 may be a device such as a camera for commercial use that includes a communication function. The terminal device 40 may be, for example, a motorcycle or a mobile relay vehicle that includes a communication device such as a field pickup unit (FPU). The terminal device 40 may be, for example, a machine-to-machine (M2M) device or an Internet of Things (IoT) device.

[0111] The terminal device 40 may be able to perform NOMA communication with the base station 20. The terminal device 40 may be able to use automatic retransmission technology such as HARQ when communicating with the base station 20. The terminal device 40 may be able to perform sidelink communication with other terminal devices 40. The terminal device 40 may be able to use automatic retransmission technology such as HARQ when performing sidelink communication. The terminal device 40 may be able to perform NOMA communication in sidelink communication with other terminal devices 40. The terminal device 40 may be able to perform LPWA communication with other wireless communication devices such as the base station 20. The wireless communication used by the terminal device 40 may be wireless communication using millimeter waves. The wireless communication used by the terminal device 40 may be wireless communication using radio waves including sidelink communication, or may be wireless communication using infrared rays or visible light, that is, optical wireless communication.

[0112] The terminal device 40 may be a movable wireless communication device, i.e., a mobile device. The terminal device 40 may be a wireless communication device installed in a mobile body, or may be the mobile body itself. The terminal device 40 may be a vehicle moving on a road, such as a car, a bus, a truck, or a motorcycle, or may be a wireless communication device included in a vehicle. The mobile body may be a mobile terminal, or a mobile body moving on land (on the ground in a narrow sense), underground, on water, or underwater. The mobile body may be a mobile body moving in the Earth's atmosphere, such as a drone or a helicopter, or may be a mobile body moving outside the Earth's atmosphere, such as an artificial satellite.

[0113] The terminal device 40 may be able to communicate by connecting to multiple base stations 20 or multiple cells at the same time. When a single base station 20 supports a communication area through multiple cells (such as pCell and sCell), the base station 20 and the terminal device 40 can communicate with each other by aggregating multiple cells together using, for example, carrier aggregation (CA), dual connectivity (DC), or multi-connectivity (MC). Alternatively, the terminal device 40 can also communicate with different base stations 20 using coordinated multi-point transmission and reception (CoMP) through cells of different base stations 20.

[0114] Figure 5 The diagram illustrates the configuration of the terminal device 40 according to the first embodiment. The terminal device 40 includes a wireless communication unit 41, a storage device 42, an encoder 43, and a controller 45. Note that Figure 5 The diagram shows the functional components, while the hardware components may differ from Figure 5 The configuration shown in the diagram is different. In addition, the functions of the terminal device 40 may be distributed to a plurality of physically separated configurations.

[0115] The wireless communication unit 41 is a signal processor for wirelessly communicating with other wireless communication devices (such as the base station 20, the relay station 30 or other terminal devices 40). The wireless communication unit 41 is controlled by the controller 45. The wireless communication unit 41 includes a transmitter 411, a receiver 412 and an antenna 413. The wireless communication unit 41, the transmitter 411, the receiver 412 and the antenna 413 respectively have configurations similar to the corresponding configurations of the wireless communication unit 21, the transmitter 211, the receiver 212 and the antenna 213 of the base station 20. As in the case of the wireless communication unit 21 of the base station 20, the wireless communication unit 41 may include a beamforming function.

[0116] The storage device 42 is a data readable / writable storage device such as a DRAM, an SRAM, a flash memory, or a hard disk. The storage device 42 serves as a storage device of the terminal device 40.

[0117] The encoder 43 includes a function of performing signal processing using an AI / ML model. The AI / ML model is a neural network model obtained by machine learning or deep learning. Examples of neural network models may include a convolutional neural network (CNN) model, a recurrent neural network (RNN) model, and a long short-term memory (LSTM) model. The AI / ML model may be any of these models, or may be obtained by combining these models in series or in parallel.

[0118] The encoder 43 generates and outputs the second data by performing signal processing using an AI / ML model in which the given first data is used as an input. The encoder 43 may be implemented by a processor such as a CPU or an MPU. The encoder 43 may be implemented by an integrated circuit such as an ASIC or an FPGA. The configuration of the encoder 43 will be described later. Figure 6 Describe in more detail.

[0119] The controller 45 is a controller that controls each component of the terminal device 40. The controller 45 can be implemented by a processor such as a CPU or an MPU. In particular, the controller 45 can be implemented by the processor using a RAM or the like as a work area to execute various programs stored in a storage device located inside the terminal device 40. The controller 45 can be implemented by an integrated circuit such as an ASIC or an FPGA. Any of the CPU, MPU, ASIC, and FPGA can be regarded as a controller. In addition to or instead of the CPU, the controller 45 can also be implemented by a GPU.

[0120] (Configuration of terminal device and base station)

[0121] The following is based on an example in which signal processing based on an AI / ML model is performed between a terminal device 40 including an encoder 43 serving as a signal processor including an AI / ML model and a base station 20 including a decoder 23 serving as a signal processor including an AI / ML model, and describes the technology according to the present disclosure. However, the terminal device 40 and the base station 20 may be replaced with each other. In other words, when signal processing is performed between any two communication devices including an AI / ML model, the technology according to the present disclosure may be applied.

[0122] In the present disclosure, the terms "encoding" and "decoding" have the concept of including any signal processing in pairs, respectively, and each term can represent not only signal processing related to compression and decompression of data, but also other signal processing. In the present disclosure, the terms "encoding" and "decoding" can be referred to by terms other than "encoding" or "decoding". The terms "encoding", "encoder", "decoding" and "decoder" are used below to describe the technology according to the present disclosure. However, terms other than the above terms can be used to describe the technology according to the present disclosure.

[0123] like Figure 6 As shown in the diagram, the wireless communication system 1 according to the first embodiment includes a terminal device 40 and a base station 20 .

[0124] The terminal device 40 includes an encoder 43 that functions as a signal processor including an AI / ML model. The encoder 43 encodes given first data using the AI / ML model in which the first data is used as an input. Thus, the encoder 43 generates second data.

[0125] The second data generated by the encoder 43 is transmitted by the transmitter 411 to the base station 20 via the uplink, and is received by the receiver 212 of the base station 20 .

[0126] The base station 20 includes a decoder 23 serving as a signal processor including an AI / ML model. The decoder 23 decodes the second data received from the terminal device 40 using the AI / ML model in which the second data is used as an input. Thus, the decoder 23 restores the first data.

[0127] (Base Station Operation)

[0128] In the first embodiment, before performing the signal processing corresponding to the above encoding based on the AI / ML model and the signal processing corresponding to the above decoding based on the AI / ML model, the base station 20 sends the information on which the learning of the AI / ML model is based to the terminal device 40. The terminal device 40 sets the AI / ML model included in the encoder 43 of the terminal device 40 based on the information on which the learning of the AI / ML model is based received from the base station 20. Thereafter, the terminal device 40 trains the AI / ML model included in the encoder 43.

[0129] For example, the information on which the learning of the AI / ML model is based may be included in system information, radio resource control (RRC) signaling, MAC control element (MAC CE), or downlink control information (DCI), and the system information, RRC signaling, MAC CE, or DCI may be transmitted. Note that in the following description, the term "transmit" has a concept including "report" or "configuration".

[0130] Examples of information on which learning of AI / ML models is based include information about the type of AI / ML model (such as CNN model, RNN model, and LSTM model), the number of input nodes (the number of neurons in the input layer), the number of output nodes (the number of neurons in the output layer), layer composition (such as the number of intermediate layers, the number of neurons in each intermediate layer, and the connection relationships), the values ​​of weighting coefficients (the values ​​of weighting coefficients between neurons), the layers in which learning is performed (such as learning is performed only in the fully connected layer in the final stage), and learning data (such as training data, validation data, and test data).

[0131] The information on which learning is based may also be expressed as, for example, "initial value of learning". When the information on which learning of the AI / ML model is based is defined in advance, this makes it possible to reduce the performance difference between the models due to subsequent learning in the corresponding terminal devices. Therefore, the base station 20 can perform signal processing with each terminal device regardless of the difference between the models of the corresponding terminal devices.

[0132] For example, when the number of input nodes of the AI / ML model and the number of output nodes of the AI / ML model are predetermined as information on which the learning of the AI / ML model is based, this makes it possible to unify the number of input nodes of the AI / ML model of the corresponding terminal device and the number of output nodes of the AI / ML model. This results in preventing the AI / ML models of the corresponding terminal devices from having different numbers of input nodes and different numbers of output nodes. Therefore, the base station 20 can send and receive data for signal processing to and from each terminal device regardless of the differences in the number of input nodes and output nodes of the AI / ML model between the corresponding terminal devices.

[0133] For example, when the values ​​(initial values) of the weighting coefficients of the AI / ML model are predetermined as information on which the learning of the AI / ML model is based, this makes it possible to reduce the performance differences between the models due to the learning subsequently performed in the corresponding terminal devices. This results in that for terminal devices with low learning capabilities, it is possible to ensure minimum performance.

[0134] A portion of the information on which the learning of the AI / ML model is based may be predetermined as a specification and stored in the storage device 42 of the terminal device 40. For example, the type of the AI / ML model, the number of input nodes, the number of output nodes, and the layer configuration may be predetermined as a specification and stored in the storage device 42 of the terminal device 40, and the information on which the learning is based that is transmitted from the base station 20 to the terminal device 40 may include only, for example, the value of the weighting coefficient and the learning data.

[0135] The information on which the learning of the AI / ML model is based can be determined based on the frequency band used by the terminal device 40. Examples of frequency bands include uplink frequency bands, downlink frequency bands, sidelink frequency bands, auxiliary uplink frequency bands, component carriers, bandwidth parts (BWPs), resource blocks, subcarriers, frequency bands below 6 GHz in FR1, frequency bands from 24 GHz to 54.26 GHz in FR2, millimeter wave bands, and terahertz wave bands.

[0136] For example, when the terminal device 40 and the base station 20 communicate with each other using frequency band A, the base station 20 can send information A based on learning to the terminal device 40, and the information A is determined to be suitable for frequency band A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using frequency band B, the base station 20 can send information B based on learning to the terminal device 40, and the information B is determined to be suitable for frequency band B.

[0137] The information on which the learning of the AI / ML model is based may be determined based on the cell in which the terminal device 40 is located. For example, when the terminal device 40 is located in cell A, the base station 20 may transmit information A on which the learning is based to the terminal device 40, and the information A is determined to be suitable for communication in cell A. Furthermore, for example, when the terminal device 40 is located in cell B, the base station 20 may transmit information B on which the learning is based to the terminal device 40, and the information B is determined to be suitable for communication in cell B.

[0138] The information on which the learning of the AI / ML model is based may be determined based on the beam used by the terminal device 40. For example, when the terminal device 40 uses beam A, the base station 20 may transmit information A to the terminal device 40, and the information A is determined to be suitable for beam A. In addition, for example, when the terminal device 40 uses beam B, the base station 20 may transmit information B on which the learning is based, and the information B is determined to be suitable for beam B, to the terminal device 40.

[0139] The information on which the learning of the AI / ML model is based can be determined based on the synchronization signal block (SSB) used by the terminal device 40. For example, when the base station 20 and the terminal device 40 communicate with each other using SSB_A, the base station 20 can send information A on which the learning is based to the terminal device 40, and the information A is determined to be suitable for SSB_A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using SSB_B, the base station 20 can send information B on which the learning is based to the terminal device 40, and the information B is determined to be suitable for SSB_B.

[0140] The information on which the learning of the AI / ML model is based can be determined based on the polarized wave used by the terminal device 40. For example, when the base station 20 and the terminal device 40 communicate with each other using right-hand circularly polarized waves, the base station 20 can send information R based on which the learning is based to the terminal device 40, and the information R is determined to be suitable for right-hand circularly polarized waves. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using left-hand circularly polarized waves, the base station 20 can send information L based on which the learning is based to the terminal device 40, and the information L is determined to be suitable for left-hand circularly polarized waves.

[0141] The information on which the learning of the AI / ML model is based may be determined based on a network slice identifier (Single Network Slice Selection Assistance Information (S-NSSAI)) used by the terminal device 40. For example, when the base station 20 and the terminal device 40 communicate with each other using network slice A, the base station 20 may transmit information A on which the learning is based to the terminal device 40, and the information A is determined to be suitable for network slice A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using network slice B, the base station 20 may transmit information B on which the learning is based to the terminal device 40, and the information B is determined to be suitable for network slice B.

[0142] In the above case, the base station 20 can make settings for each network slice regarding whether to perform signal processing based on the AI / ML model between the base station 20 and the terminal device 40. For example, when the base station 20 and the terminal device 40 communicate with each other using network slice C, the base station 20 can make settings so that signal processing based on the AI / ML model is performed. In addition, for example, when the base station 20 and the terminal device 40 communicate with each other using network slice D, the base station 20 can make settings so that signal processing based on the AI / ML model is not performed.

[0143] The base station 20 may transmit information based on which the learning of the AI / ML model is based according to the radio resource control state (RRC state) of the terminal device 40. For example, when the terminal device 40 is in the RRC idle state or the RRC inactive state, the base station 20 does not necessarily have to transmit information based on which the learning of the AI / ML model is based to the terminal device 40. In this case, depending on the implementation of the terminal device 40, the terminal device 40 may separately set and train the AI / ML model included in the encoder 43 of the terminal device 40.

[0144] The base station 20 may transmit information based on which the learning of the AI / ML model is to be performed to the terminal device 40 at the timing of updating the tracking area (TA), the registration area (RA), or the notification area (RNA) based on the RAN.

[0145] For example, when the terminal device 40 is in the RRC idle state or the RRC inactive state, the base station 20 may include information on which learning of the AI / ML model is based in the system information, and may send the system information to the terminal device 40. In this case, the terminal device 40 may set and train the AI / ML model included in the encoder 43 of the terminal device 40 based on the received information on which learning is based.

[0146] For example, when the terminal device 40 is in the RRC connected state, the base station 20 may include information on which learning of the AI / ML model is based in RRC signaling, MAC CE, DCI, etc., and may send the RRC signaling, MAC CE, DCI, etc. to the terminal device 40. In this case, the terminal device 40 may set and train the AI / ML model included in the encoder 43 of the terminal device 40 based on the received information on which learning is based.

[0147] The information on which the learning of the AI / ML model is based may include only a single piece of information on which the learning is based, or may include a plurality of pieces of information on which the learning is based.

[0148] For example, when the information on which the learning of the AI / ML model received from the base station 20 is based only includes a single information on which the learning is based, the terminal device 40 can set and train the AI / ML model included in the encoder 43 of the terminal device 40 based on the single information on which the learning is based.

[0149] For example, when the information on which the learning of the AI / ML model received from the base station 20 is based includes multiple information on which the learning is based, the terminal device 40 can select one of the multiple information on which the learning is based, and then the terminal device 40 can set and train the AI / ML model included in the encoder 43 of the terminal device 40.

[0150] When the information based on which the learning of the AI / ML model is based is included in the system information transmitted from the base station 20 to each terminal device, the information based on which the learning of the AI / ML model is based can be shared by the cell. Only when each terminal device has the ability to perform signal processing based on the AI / ML model can the terminal device obtain the information based on which the learning of the AI / ML model is based included in the system information.

[0151] For example, it is assumed that system information for an AI / ML model is defined. The terminal device 40 refers to scheduling information about other system information included in, for example, block 1 of the received system information. When the system information for the AI / ML model is scheduled, the terminal device 40 can obtain information based on which learning of the AI / ML model is based on the system information for the AI / ML model.

[0152] The information on which the learning of the AI / ML model is based may include information on whether additional learning is allowed. When the information on whether additional learning is allowed is set to "not allowed", after the AI / ML model is set based on the information on which the learning of the AI / ML model is based received from the base station 20, the terminal device 40 is not allowed to additionally train the AI / ML model included in the encoder 43 of the terminal device 40.

[0153] Alternatively, separately from the information on which the learning of the AI / ML model is based, the base station 20 may transmit information on whether additional learning is allowed to the terminal device 40. When the information on whether additional learning is allowed is set to "not allowed", after the AI / ML model is set based on the information on which the learning of the AI / ML model is based received from the base station 20, the terminal device 40 is not allowed to additionally train the AI / ML model included in the encoder 43 of the terminal device 40.

[0154] (Operation of terminal device)

[0155] In the first embodiment, after the terminal device 40 sets and trains the AI / ML model included in the encoder 43 of the terminal device 40 based on the information on which the learning of the AI / ML model is based received from the base station 20, the terminal device 40 can feedback information about the learned AI / ML model (such as the type of AI / ML model, the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) to the base station 20.

[0156] The base station 20 may send information for additional learning of the AI / ML model (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) to the terminal device 40 based on the information about the learned AI / ML model fed back by the terminal device 40. The terminal device 40 may additionally train the AI / ML model included in the encoder 43 of the terminal device 40 based on the information for additional learning of the AI / ML model received from the base station 20.

[0157] When the AI / ML model included in the encoder 43 of the terminal device 40 has been learned, the terminal device 40 can send information about the learned AI / ML model to the base station 20 before receiving information on which the learning of the AI / ML model is based from the base station 20, wherein the information about the learned AI / ML model is similar to the above-mentioned information about the learned AI / ML model.

[0158] For example, when the terminal device 40 performs signal processing based on the AI / ML model together with multiple base stations, the terminal device 40 needs to train the AI / ML models corresponding to the respective base stations, which results in a very heavy load on the terminal device 40.

[0159] When the AI / ML model included in the encoder 43 of the terminal device 40 has been learned, the terminal device 40 can send information about its own learned AI / ML model (such as the type of AI / ML model, the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) to the base station 20 to which the terminal device 40 is newly connected, so as to reduce the above-mentioned load.

[0160] When the base station 20 receives information about the AI / ML model that has been learned from the terminal device 40, the base station 20 can determine the information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) based on the information about the AI / ML model that has been learned, and can send the determined information to the terminal device 40.

[0161] As capability information about the capabilities of the terminal device 40, the terminal device 40 may send the type of AI / ML model supported by the terminal device 40, the number of input nodes, the number of output nodes, the layer composition, etc. to the base station 20. Alternatively, as capability information about the capabilities of the terminal device 40, the terminal device 40 may send information about its own learned AI / ML model to the base station 20.

[0162] The base station 20 may pre-prepare a list of AI / ML models supported by the base station 20, and may send the list to the terminal device 40 as static information such as a specification, or may include the list in system information to send the system information to the terminal device 40. The terminal device 40 may select an AI / ML model that the terminal device 40 can support from the list, and may include the selected AI / ML model in the capability information about the capability of the terminal device 40 to send the capability information to the base station 20.

[0163] For example, as capability information about the capability of the terminal device 40, the terminal device 40 may send the type of AI / ML model supported by the terminal device 40, the number of input nodes, the number of output nodes, the layer composition, etc. to the base station 20. When the base station 20 receives the capability information from the terminal device 40, the base station 20 may determine the information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) based on the capability information, and may send the determined information to the terminal device 40.

[0164] In this case, when the terminal device 40 has already made the AI / ML model learn with other base stations, the terminal device 40 may send limited capability information about the learned AI / ML model, such as the number of input nodes, the number of output nodes, and the layer composition. Alternatively, the terminal device 40 may send information about the learned AI / ML model separately from the capability information to the base station 20.

[0165] In this case, as capability information about the capability of the terminal device 40, the terminal device 40 may send information about an AI / ML model that has been learned or information about an AI / ML model that has not yet been learned to the base station 20. The base station 20 may determine information on which the learning of the AI / ML model is based based on the capability information, and may send the determined information to the terminal device 40.

[0166] For example, as capability information about the capabilities of the terminal device 40, the terminal device 40 may send information about, for example, supported AI / ML models, capabilities of a CPU or GPU, whether it is a device supporting reduced capabilities (RedCap), whether it is an IoT terminal such as NB-IoT or eMTC, and supported bandwidth to the base station 20. The base station 20 may determine information on which learning of the AI / ML model is based based on the capability information, and may send the determined information to the terminal device 40.

[0167] As capability information about the capability of the terminal device 40, the terminal device 40 may transmit information about, for example, the remaining processing capability and the maximum value of the processing capability to the base station 20. For example, such information may be included in capability information, RRC signaling, MAC CE, DCI, etc. at the time of initial access, and capability information, RRC signaling, MAC CE, DCI, etc. may be transmitted. In addition, such information may be transmitted when the base station 20 requests the information, or may be transmitted periodically. When transmitting information periodically, the base station 20 may transmit instructions to the terminal device 40 about, for example, at which timing and at what frequency to transmit. Alternatively, the terminal device 40 may transmit at, for example, a predetermined frequency and a predetermined timing.

[0168] When the terminal device 40 sets and trains multiple AI / ML models in the encoder 43 of the terminal device 40 based on the information on which the learning is based determined according to the frequency band, the terminal device 40 can select one of the multiple AI / ML models according to the frequency band used by the terminal device 40, and can use the selected AI / ML model.

[0169] For example, when the terminal device 40 and the base station 20 communicate with each other using frequency band A, the terminal device 40 can use the AI / ML model set and trained based on the information based on the learning determined to be suitable for frequency band A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using frequency band B, the terminal device 40 can use the AI / ML model set and trained based on the information based on the learning determined to be suitable for frequency band B.

[0170] When the terminal device 40 sets and trains multiple AI / ML models in the encoder 43 of the terminal device 40 based on the information on which the learning is based determined according to the cell, the terminal device 40 can select one of the multiple AI / ML models according to the cell in which the terminal device 40 is located, and can use the selected AI / ML model.

[0171] For example, when the terminal device 40 is located in cell A, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning in cell A. In addition, for example, when the terminal device 40 is located in cell B, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning in cell B.

[0172] When the terminal device 40 sets and trains multiple AI / ML models in the encoder 43 of the terminal device 40 based on information on which learning is based determined according to the beam, the terminal device 40 can select one of the multiple AI / ML models according to the beam used by the terminal device 40, and can use the selected AI / ML model.

[0173] For example, when the terminal device 40 and the base station 20 communicate with each other using beam A, the terminal device 40 can use an AI / ML model set and trained based on information based on learning determined to be suitable for beam A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using beam B, the terminal device 40 can use an AI / ML model set and trained based on information based on learning determined to be suitable for beam B.

[0174] When the terminal device 40 sets and trains multiple AI / ML models in the encoder 43 of the terminal device 40 based on the information on which the learning is based determined according to the SSB, the terminal device 40 can select one of the multiple AI / ML models according to the SSB used by the terminal device 40, and can use the selected AI / ML model.

[0175] For example, when the terminal device 40 and the base station 20 communicate with each other using SSB_A, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning based on SSB_A. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using SSB_B, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning based on SSB_B.

[0176] When the terminal device 40 sets and trains multiple AI / ML models in the encoder 43 of the terminal device 40 based on the information on which the learning is based determined according to the polarized wave, the terminal device 40 can select one of the multiple AI / ML models according to the polarized wave used by the terminal device 40, and can use the selected AI / ML model.

[0177] For example, when the terminal device 40 and the base station 20 communicate with each other using right-hand circularly polarized waves, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning based on right-hand circularly polarized waves. In addition, for example, when the terminal device 40 and the base station 20 communicate with each other using left-hand circularly polarized waves, the terminal device 40 can use an AI / ML model set and trained based on information determined to be suitable for learning based on left-hand circularly polarized waves.

[0178] The terminal device 40 may select one of a plurality of AI / ML models according to the RRC state of the terminal device 40 , and may use the selected AI / ML model.

[0179] For example, when the terminal device 40 is in the RRC idle state, the terminal device 40 may use an AI / ML model that is individually set and trained according to the implementation of the terminal device 40. Alternatively, when the terminal device 40 is in the RRC idle state, the terminal device 40 may use an AI / ML model that is set based on information on which learning is based included in system information received from the base station 20, or the terminal device 40 may use an AI / ML model that is set and trained based on information on which learning is based.

[0180] For example, when the terminal device 40 is in the RRC inactive state, the terminal device 40 may use an AI / ML model set based on the information based on the learning received during the RRC connection, or the terminal device 40 may use an AI / ML model set and trained based on the information based on the learning. Alternatively, when the terminal device 40 is in the RRC inactive state, the terminal device 40 may use an AI / ML model set based on the information based on the learning included in the system information received from the base station 20, or the terminal device 40 may use an AI / ML model set and trained based on the information based on the learning.

[0181] For example, when the terminal device 40 is in an RRC connected state, the terminal device 40 may use an AI / ML model set based on information on which learning is based, for example, included in RRC signaling, MAC CE, or DCI, or the terminal device 40 may use an AI / ML model set and trained based on information on which learning is based.

[0182] (Dataset providing device)

[0183] Figure 7The diagram illustrates the configuration of a data set providing device 50 according to the first embodiment. The data set providing device 50 is a device that provides information on which the learning of the AI / ML model is based, and can be commonly accessed by terminal devices of different vendors and base stations of different vendors. The data set providing device 50 can be a device having a physical entity, or can be defined as, for example, a single network function (NF).

[0184] The data set providing device 50 includes a storage device 51, a receiver 52, a transmitter 53, and a controller 54. The storage device 51 stores therein information on which the learning of the AI / ML model is based. The receiver 52 receives a request for providing information on which the learning of the AI / ML model is based from the base station 20 or the terminal device 40. The transmitter 53 transmits the information on which the learning of the AI / ML model is based to the base station 20 or the terminal device 40 that has transmitted the provision request. The controller 54 controls the storage device 51, the receiver 52, and the transmitter 53.

[0185] The base station 20 or the terminal device 40 accesses the entity of the data set providing device 50 before the training of the AI / ML model, and receives in advance the information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, the value of the weighting coefficient, the layer in which the learning is performed, and the learning data). This enables the base station 20 and the terminal device 40 to jointly obtain a piece of information on which the learning is based from the data set providing device 50, which does not depend on the specifications or implementation of a specific vendor. This makes it possible to set and train AI / ML models that are interoperable between different vendors.

[0186] The data set providing device 50 may be defined as a first data set providing device for a base station and a second data set providing device for a terminal device, and the first and second data set providing devices are defined separately from each other. Alternatively, only one of the first data set providing device for a base station and the second data set providing device for a terminal device may be defined.

[0187] For example, when two data set providing devices are defined separately from each other, the base station 20 receives information based on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, the value of the weighting coefficient, the layer in which the learning is performed, and the learning data) from the first data set providing device in advance. In addition, the terminal device 40 receives information based on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, the value of the weighting coefficient, the layer in which the learning is performed, and the learning data) from the second data set providing device in advance.

[0188] The base station 20 or the terminal device 40 may send a request to provide information on which the learning of the AI / ML model is based to the dataset providing device 50. When the dataset providing device 50 receives the provision request, the dataset providing device 50 may resend the information on which the learning of the AI / ML model is based to the base station 20 or the terminal device 40 that sent the provision request.

[0189] As capability information about its own capabilities, the base station 20 or the terminal device 40 may send, for example, information about the currently held AI / ML model that has been learned, supported AI / ML models, capabilities of the CPU or GPU, whether it is a device that supports reduced capabilities (RedCap), whether it is an IoT terminal such as NB-IoT or eMTC, and supported bandwidth to the dataset providing device 50. The dataset providing device 50 may determine information on which the learning of the AI / ML model is based based on the capability information, and may send the determined information to the base station 20 or the terminal device 40. In addition, based on the information about the currently held AI / ML model that has been learned received from the base station 20 or the terminal device 40, the dataset providing device 50 may update the information on which the learning of the AI / ML model is based, and the updated information is maintained by the dataset providing device 50.

[0190] The base station 20 may transmit information about access to the dataset providing apparatus 50 to the terminal device 40. For example, the base station 20 may transmit information such as an IP address or a MAC address required for the terminal device 40 to access the dataset providing apparatus 50 to the terminal device 40.

[0191] Figure 8 The diagram illustrates an example of the composition of a wireless communication network including a data set providing device 50. For example, as illustrated in the figure, the data set providing device 50 can be arranged under a data network outside the core network. For example, the data set providing device 50 can be included in a server (such as a mobile edge computing (MEC) server) connected to a user plane function (UPF). For example, the data set providing device 50 can be included in a server (such as a MEC server) connected to the base station 20. For example, the data set providing device 50 can be defined as a new network function (NF). For example, the data set providing device 50 can be defined as an additional function of an existing NF (such as a model training logic function (MTLF)). In addition, not only a single data set providing device 50 arranged at a single location can be adopted, but also multiple data set providing devices 50 arranged at multiple locations can be adopted, or not only a single data set providing device 50 can be defined, but also multiple data set providing devices 50 can be defined.

[0192] (Operations performed during switching)

[0193] When the terminal device 40 switches from a certain base station (source base station) to another base station (target base station), the source base station may send information about the AI / ML model that has been learned (such as the type of AI / ML model, the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient), information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient), capability information about the capabilities of the AI / ML model (such as the type of AI / ML model supported, the number of input nodes, the number of output nodes, and the layer composition), and other information to the target base station. When the target base station receives this information from the source base station, the target base station may set and train the AI / ML model included in the decoder 23 of the target base station based on the received information.

[0194] For example, based on capability information about the capabilities of the terminal device 40 or the base station in terms of the AI / ML model, or about the AI / ML model that has been learned, the base station to which the terminal device 40 is switched can be limited or determined. In other words, when switching the terminal device 40, if the AI / ML model that has been learned is not allowed to be used at the base station to which the terminal device 40 is switched, it will be necessary to re-set and train the AI / ML model. This results in an increase in the processing burden imposed on the terminal device 40 and the base station. Therefore, the terminal device 40 can be switched to a base station where the AI / ML model that has been learned can be used.

[0195] (Operations for each use case)

[0196] Generally, the following are three use cases in which terminal devices and base stations jointly perform AI / ML model-based signal processing.

[0197] 1. Only base stations use AI / ML models.

[0198] 2. Only end devices use AI / ML models.

[0199] 3. Both the terminal device and the base station use the AI / ML model (the case of the first embodiment).

[0200] In the case of "1" above, only the base station uses the AI / ML model. Therefore, it is not a problem even if the AI / ML models of different base stations are different. For example, when there are multiple base station vendors, the multiple vendors can set and train AI / ML models separately according to the implementation.

[0201] In the case of "2" above, only the terminal device uses the AI / ML model. Therefore, it is not a problem even if the AI / ML models of different terminal devices are different. For example, when there are multiple terminal vendors, the multiple terminal vendors can set and train AI / ML models separately according to the implementation.

[0202] However, there may be differences in processing capabilities between terminal devices, and there may be terminal devices that cannot properly set and train AI / ML models. To address this problem, the base station can send information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) to the terminal device. In addition, the terminal device can send information about the capabilities of the terminal device with respect to the AI / ML model to the base station (such as the type of AI / ML model supported, the number of input nodes, the number of output nodes, the layer composition, and information about the CPU or GPU maintained).

[0203] In the case of "3" above, both the terminal device and the base station use the AI / ML model. Therefore, the difference between the terminal device and the base station needs to be considered. In order to cope with this problem, the base station can send information on which the learning of the AI / ML model is based (such as the number of input nodes, the number of output nodes, the layer composition, and the value of the weighting coefficient) to the terminal device. In addition, the terminal device can send information about the capabilities of the terminal device with respect to the AI / ML model to the base station (such as the type of AI / ML models supported, the number of input nodes, the number of output nodes, the layer composition, and information about the CPU or GPU maintained).

[0204] (Communication between terminal devices and communication between base stations)

[0205] The technology according to the present disclosure can also be applied to communication between terminal devices such as sidelink and communication between base stations such as integrated access backhaul (IAB).

[0206] For example, when a parent terminal device and a child terminal device exist in communication between terminal devices, the parent terminal device can serve the function of the above-mentioned base station, and the child terminal device can serve the function of the above-mentioned terminal device.

[0207] For example, when a parent base station and a child base station exist in communication between base stations, the parent base station can serve the function of the above-mentioned base station, and the child base station can serve the function of the above-mentioned terminal device.

[0208] (Example of process)

[0209] Six examples of processes in which the technology according to the present disclosure is applied to channel state information (CSI) feedback are described below. However, the scope in which the technology according to the present disclosure can be applied is not limited thereto. For example, the technology according to the present disclosure can also be applied to, for example, beam management, beam positioning, or mobility management.

[0210] (First example)

[0211] Fig. 9 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in system information and transmits the system information to each of different terminal devices.

[0212] In step S101, terminal devices A and B respectively receive synchronization signals sent by the base station and perform downlink synchronization. In addition, terminal devices A and B respectively receive system information sent by the base station to receive information required for cell connection. Here, the system information includes information on which the learning of the AI / ML model is based.

[0213] In step S102, the terminal devices A and B may respectively additionally train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the system information received from the base station. The base station may additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based included in the system information sent by the base station.

[0214] Terminal devices A and B may determine whether to perform additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform additional training separately, terminal devices A and B may obtain different training results (learning results).

[0215] In step S103, terminal devices A and B respectively perform a random access process and establish a connection with the base station. Terminal devices A and B respectively achieve uplink synchronization through the random access process and complete the connection with the base station. The random access process can be performed after the additional training is completed or during the additional training.

[0216] In step S104, terminal devices A and B respectively send capability information about the capability of a corresponding one of terminal devices A and B to the base station. The capability information may include capability information about the capability of the AI / ML model. In step S105, the base station sends quasi-static information to each of terminal devices A and B. The quasi-static information may be, for example, RRC signaling.

[0217] In step S106, the base station transmits a reference signal for estimating the downlink channel state to the terminal apparatus B. In step S107, the terminal apparatus B performs a process of estimating the downlink channel state, and generates downlink channel state information.

[0218] In step S108, the terminal device B performs signal processing based on the AI / ML model included in the encoder of the terminal device B to encode the downlink channel state information. In step S109, the terminal device B feeds back the encoded downlink channel state information to the base station.

[0219] The above-mentioned downlink channel state information may be an uplink control signal defined by uplink control information (UCI) (such as channel quality indication (CQI), precoding matrix indication (PMI) or resource indication (RI)), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0220] In step S110 , the base station performs signal processing based on an AI / ML model included in a decoder of the base station to decode the encoded downlink channel state information received from the terminal device B, thereby acquiring the downlink channel state information.

[0221] In step S111, the base station sends a downlink control signal. The downlink control signal may be, for example, downlink control information (DCI). The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0222] In step S112, the base station sends downlink data. The downlink data may be, for example, a physical downlink shared channel (PDSCH). In step S113, terminal device B sends information about retransmission control to the base station according to the decoding result of the downlink data. For example, the information about retransmission control may be an acknowledgment (ACK) / negative acknowledgment (NACK), or may be a "hybrid automatic repeat request-acknowledgement" (HARQ-ACK).

[0223] (Second example)

[0224] Fig.10 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in quasi-static information (RRC signaling) and sends the quasi-static information to each of different terminal devices.

[0225] In step S201, terminal devices A and B respectively receive synchronization signals transmitted by the base station and perform downlink synchronization. In addition, terminal devices A and B respectively receive system information transmitted by the base station to receive information required for cell connection.

[0226] In step S202, terminal devices A and B respectively perform a random access process and establish a connection with the base station. Terminal devices A and B respectively achieve uplink synchronization through the random access process and complete the connection with the base station. The random access process can be performed after the additional training described later is completed or during the additional training.

[0227] In step S203, the terminal devices A and B respectively transmit capability information about the capability of a corresponding one of the terminal devices A and B to the base station. The capability information may include capability information about the capability of the AI / ML model.

[0228] In step S204, the base station sends quasi-static information to each of the terminal devices A and B. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information on which the learning of the AI / ML model is based.

[0229] In step S205, the terminal devices A and B may respectively additionally train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the quasi-static information received from the base station. The base station may additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based included in the quasi-static information sent by the base station.

[0230] Terminal devices A and B may determine whether to perform additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform additional training separately, terminal devices A and B may obtain different training results (learning results).

[0231] In step S206, the base station transmits a reference signal for estimating the downlink channel state to the terminal apparatus B. In step S207, the terminal apparatus B performs a process of estimating the downlink channel state, and generates downlink channel state information.

[0232] In step S208, the terminal device B performs signal processing based on the AI / ML model included in the encoder of the terminal device B to encode the downlink channel state information. In step S209, the terminal device B feeds back the encoded downlink channel state information to the base station.

[0233] The above-mentioned downlink channel state information may be an uplink control signal defined by UCI (such as CQI, PMI or RI), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0234] In step S210, the base station performs signal processing based on an AI / ML model included in a decoder of the base station to decode the encoded downlink channel state information received from the terminal device B, thereby acquiring the downlink channel state information.

[0235] In step S211, the base station sends a downlink control signal. The downlink control signal may be, for example, a DCI. The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0236] In step S212, the base station transmits downlink data. The downlink data may be, for example, PDSCH. In step S213, terminal device B transmits information on retransmission control to the base station according to the decoding result of the downlink data. For example, the information on retransmission control may be ACK / NACK, or may be HARQ-ACK.

[0237] (Third example)

[0238] Fig.11 is a diagram for describing an example of a process in which a base station includes information on which learning of an AI / ML model is based in each of system information and quasi-static information (RRC signaling), and sends the system information and quasi-static information to each of different terminal devices.

[0239] In step S301, terminal devices A and B respectively receive synchronization signals sent by the base station and perform downlink synchronization. In addition, terminal devices A and B respectively receive system information sent by the base station to receive information required for cell connection. Here, the system information includes information on which the learning of the AI / ML model is based.

[0240] In step S302, the terminal devices A and B may respectively additionally train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the system information received from the base station. The base station may additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based included in the system information sent by the base station.

[0241] Terminal devices A and B may determine whether to perform additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform additional training separately, terminal devices A and B may obtain different training results (learning results).

[0242] In step S303, terminal devices A and B respectively perform a random access process and establish a connection with the base station. Terminal devices A and B respectively achieve uplink synchronization through the random access process and complete the connection with the base station. The random access process can be performed after the additional training is completed or during the additional training.

[0243] In step S304, the terminal devices A and B respectively transmit capability information about the capability of a corresponding one of the terminal devices A and B to the base station. The capability information may include capability information about the capability of the AI / ML model.

[0244] In step S305, the base station sends quasi-static information to each of the terminal devices A and B. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information on which the learning of the AI / ML model is based.

[0245] In step S306, the terminal devices A and B may further respectively train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the quasi-static information received from the base station. The base station may also further train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based included in the quasi-static information sent by the base station.

[0246] Terminal devices A and B may determine whether to perform further additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform further additional training separately, terminal devices A and B may obtain different training results (learning results).

[0247] In step S307, the base station transmits a reference signal for estimating the downlink channel state to the terminal apparatus B. In step S308, the terminal apparatus B performs a process of estimating the downlink channel state, and generates downlink channel state information.

[0248] In step S309, the terminal device B performs signal processing based on the AI / ML model included in the encoder of the terminal device B to encode the downlink channel state information. In step S310, the terminal device B feeds back the encoded downlink channel state information to the base station.

[0249] The above-mentioned downlink channel state information may be an uplink control signal defined by UCI (such as CQI, PMI or RI), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0250] In step S311, the base station performs signal processing based on the AI / ML model included in the decoder of the base station to decode the encoded downlink channel state information received from the terminal device B, thereby acquiring the downlink channel state information.

[0251] In step S312, the base station sends a downlink control signal. The downlink control signal may be, for example, a DCI. The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0252] In step S313, the base station transmits downlink data. The downlink data may be, for example, PDSCH. In step S314, terminal device B transmits information on retransmission control to the base station according to the decoding result of the downlink data. For example, the information on retransmission control may be ACK / NACK, or may be HARQ-ACK.

[0253] (Fourth example)

[0254] Fig.12 This is a diagram for describing an example of the following process: a base station receives information on which learning of an AI / ML model is based from a data set providing device; includes the information on which learning of the AI / ML model is based in each of system information and quasi-static information (RRC signaling); and sends the system information and the quasi-static information to each of different terminal devices.

[0255] In step S401, the base station sends a request to the data set providing device to provide information on which the learning of the AI / ML model is based. In step S402, the data set providing device sends the information on which the learning of the AI / ML model is based to the base station.

[0256] In step S403 , the base station may additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based received from the dataset providing device.

[0257] In step S404, terminal devices A and B respectively receive synchronization signals sent by the base station and perform downlink synchronization. In addition, terminal devices A and B respectively receive system information sent by the base station to receive information required for cell connection. Here, the system information includes information on which the learning of the AI / ML model is based.

[0258] In step S405 , the terminal devices A and B may respectively additionally train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the system information received from the base station.

[0259] Terminal devices A and B may determine whether to perform additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform additional training separately, terminal devices A and B may obtain different training results (learning results).

[0260] In step S406, terminal devices A and B respectively perform a random access process and establish a connection with the base station. Terminal devices A and B respectively achieve uplink synchronization through the random access process and complete the connection with the base station. The random access process can be performed after the additional training is completed or during the additional training.

[0261] In step S407, the terminal devices A and B respectively transmit capability information about the capability of a corresponding one of the terminal devices A and B to the base station. The capability information may include capability information about the capability of the AI / ML model.

[0262] In step S408, the base station sends quasi-static information to each of the terminal devices A and B. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information on which the learning of the AI / ML model is based.

[0263] In step S409, the terminal devices A and B may further respectively train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based included in the quasi-static information received from the base station. The base station may also additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based included in the quasi-static information sent by the base station.

[0264] Terminal devices A and B may determine whether to perform further additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform further additional training separately, terminal devices A and B may obtain different training results (learning results).

[0265] In step S410, the base station transmits a reference signal for estimating the downlink channel state to the terminal apparatus B. In step S411, the terminal apparatus B performs a process of estimating the downlink channel state, and generates downlink channel state information.

[0266] In step S412, the terminal device B performs signal processing based on the AI / ML model included in the encoder of the terminal device B to encode the downlink channel state information. In step S413, the terminal device B feeds back the encoded downlink channel state information to the base station.

[0267] The above-mentioned downlink channel state information may be an uplink control signal defined by UCI (such as CQI, PMI or RI), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0268] In step S414, the base station performs signal processing based on the AI / ML model included in the decoder of the base station to decode the encoded downlink channel state information received from the terminal device B, thereby acquiring the downlink channel state information.

[0269] In step S415, the base station sends a downlink control signal. The downlink control signal may be, for example, a DCI. The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0270] In step S416, the base station transmits downlink data. The downlink data may be, for example, PDSCH. In step S417, terminal device B transmits information on retransmission control to the base station according to the decoding result of the downlink data. For example, the information on retransmission control may be ACK / NACK, or may be HARQ-ACK.

[0271] (Fifth example)

[0272] Fig.13 is a diagram for describing an example of a process in which a base station and a terminal device respectively receive information on which learning of an AI / ML model is based from a dataset providing device.

[0273] In step S501, the base station sends a request to the data set providing device to provide information on which the learning of the AI / ML model is based. In step S502, the data set providing device sends the information on which the learning of the AI / ML model is based to the base station.

[0274] In step S503 , the base station may additionally train the AI / ML model included in the decoder of the base station based on the information on which the learning of the AI / ML model is based received from the dataset providing device.

[0275] In step S504, terminal devices A and B respectively receive synchronization signals transmitted by the base station and perform downlink synchronization. In addition, terminal devices A and B respectively receive system information transmitted by the base station to receive information required for cell connection. Here, the system information includes information about the data set providing device.

[0276] In step S505, terminal devices A and B respectively send a request to the dataset providing device to provide information based on which learning of the AI / ML model is based, based on information about the dataset providing device included in the system information received from the base station. In step S506, the dataset providing device sends the information based on which learning of the AI / ML model is based to each of terminal devices A and B.

[0277] In step S507 , the terminal devices A and B may respectively additionally train the AI / ML model included in the encoder of the corresponding one of the terminal devices A and B based on the information on which the learning of the AI / ML model is based received from the dataset providing device.

[0278] Terminal devices A and B may determine whether to perform additional training according to implementation or based on instructions sent by the base station. Since terminal devices A and B perform additional training separately, terminal devices A and B may obtain different training results (learning results).

[0279] In step S508, terminal devices A and B respectively perform a random access process and establish a connection with the base station. Terminal devices A and B respectively achieve uplink synchronization through the random access process and complete the connection with the base station. The random access process can be performed after the additional training is completed or during the additional training.

[0280] In step S509, the terminal devices A and B respectively transmit capability information about the capability of a corresponding one of the terminal devices A and B to the base station. The capability information may include capability information about the capability of the AI / ML model.

[0281] In step S510, the base station transmits quasi-static information to each of the terminal devices A and B. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information about the technology according to the present disclosure.

[0282] In step S511, the base station transmits a reference signal for estimating the downlink channel state to the terminal apparatus B. In step S512, the terminal apparatus B performs a process of estimating the downlink channel state, and generates downlink channel state information.

[0283] In step S513, the terminal device B performs signal processing based on the AI / ML model included in the encoder of the terminal device B to encode the downlink channel state information. In step S514, the terminal device B feeds back the encoded downlink channel state information to the base station.

[0284] The above-mentioned downlink channel state information may be an uplink control signal defined by UCI (such as CQI, PMI or RI), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0285] In step S515, the base station performs signal processing based on the AI / ML model included in the decoder of the base station to decode the encoded downlink channel state information received from the terminal device B, thereby acquiring the downlink channel state information.

[0286] In step S516, the base station sends a downlink control signal. The downlink control signal may be, for example, a DCI. The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0287] In step S517, the base station transmits downlink data. The downlink data may be, for example, PDSCH. In step S518, terminal device B transmits information on retransmission control to the base station according to the decoding result of the downlink data. For example, the information on retransmission control may be ACK / NACK, or may be HARQ-ACK.

[0288] (Sixth Example)

[0289] FIG. 14 is a diagram for describing an example of a process in which a terminal device is handed over to a different base station.

[0290] In step S601, the terminal device receives a synchronization signal transmitted by the base station C and performs downlink synchronization. In addition, the terminal device receives system information transmitted by the base station C to receive information required for cell connection. Here, the system information includes information on which the learning of the AI / ML model is based.

[0291] In step S602, the terminal device may additionally train the AI / ML model included in the encoder of the terminal device based on the information on which the learning of the AI / ML model is based included in the system information received from the base station C. The base station C may additionally train the AI / ML model included in the decoder of the base station C based on the information on which the learning of the AI / ML model is based included in the system information sent by the base station C. The terminal device may determine whether to perform additional training according to implementation or based on an instruction sent by the base station C.

[0292] In step S603, the terminal device performs a random access process and establishes a connection with the base station C. The terminal device achieves uplink synchronization through the random access process and completes the connection with the base station C. The random access process may be performed after the additional training is completed or during the additional training.

[0293] In step S604, the terminal device sends capability information about the capability of the terminal device to the base station C. The capability information may include capability information about the capability of the AI / ML model.

[0294] In step S605, the base station C sends quasi-static information to the terminal device. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information on which the learning of the AI / ML model is based.

[0295] In step S606, the terminal device may also additionally train the AI / ML model included in the encoder of the terminal device based on the information on which the learning of the AI / ML model is based included in the quasi-static information received from the base station C. The base station C may also additionally train the AI / ML model included in the decoder of the base station C based on the information on which the learning of the AI / ML model is based included in the quasi-static information sent by the base station C. The terminal device may determine whether to perform further additional training based on the implementation or based on the instruction sent by the base station C.

[0296] In step S607, the base station C transmits a reference signal for estimating the downlink channel state to the terminal device. In step S608, the terminal device performs a process of estimating the downlink channel state and generates downlink channel state information.

[0297] In step S609, the terminal device performs signal processing based on the AI / ML model included in the encoder of the terminal device to encode the downlink channel state information. In step S610, the terminal device feeds back the encoded downlink channel state information to the base station C.

[0298] The above-mentioned downlink channel state information may be an uplink control signal defined by UCI (such as CQI, PMI or RI), may be the channel response information itself, or may be eigenvalue information about the eigenvalue of the channel.

[0299] In step S611, the base station C performs signal processing based on the AI / ML model included in the decoder of the base station C to decode the encoded downlink channel state information received from the terminal device, thereby acquiring the downlink channel state information.

[0300] In step S612, base station C sends a downlink control signal. The downlink control signal may be, for example, a DCI. The downlink control signal may include information about the technology according to the present disclosure. In addition, the downlink control signal may be partially determined based on the downlink channel state information fed back above.

[0301] In step S613, the base station C sends downlink data. The downlink data may be, for example, PDSCH. In step S614, the terminal device sends information on retransmission control to the base station C according to the decoding result of the downlink data. For example, the information on retransmission control may be ACK / NACK, or may be HARQ-ACK.

[0302] In step S615, the base station C (source cell) and the terminal device perform a measurement process related to the connected cell. In step S616, the base station C determines whether a handover is required based on the result of the measurement process, and determines that there is a need for a handover.

[0303] In step S617, base station C sends a handover request to base station D which is a candidate target cell. In step S618, base station D performs admission control. In step S619, base station D sends a report (handover request confirmation) to base station C on whether the handover is permitted.

[0304] In step S620, base station C sends information on which the learning of the AI / ML model maintained by base station C is based to base station D. In addition, base station C can send information about other learned AI / ML models to base station D. In step S621, base station D sends an ACK to base station C.

[0305] In step S622, base station D may also additionally train the AI / ML model included in the decoder of base station D based on the information on which the learning of the AI / ML model is based received from base station C.

[0306] In step S623, the base station C performs RCC configuration on the terminal device and sends information about the handover to the terminal device. The information about the handover includes information on which the learning of the AI / ML model is based. In step S624, the terminal device detaches from the base station C.

[0307] In step S625, the terminal device performs a random access procedure and establishes a connection with the base station D. The terminal device achieves uplink synchronization through the random access procedure and completes the connection with the base station D. The random access procedure may be performed after the additional training is completed or during the additional training.

[0308] In step S626, the terminal device sends capability information about the capability of the terminal device to the base station D. The capability information may include capability information about the capability of the AI / ML model.

[0309] In step S627, the base station D sends quasi-static information to the terminal device. The quasi-static information may be, for example, RRC signaling. The quasi-static information includes information on which the learning of the AI / ML model is based.

[0310] In step S628, the terminal device may also additionally train the AI / ML model included in the encoder of the terminal device based on the information on which the learning of the AI / ML model is based included in the quasi-static information received from the base station D. The base station D may also additionally train the AI / ML model included in the decoder of the base station D based on the information on which the learning of the AI / ML model is based included in the quasi-static information sent by the base station D. The terminal device may determine whether to perform further additional training based on the implementation or based on the instruction sent by the base station D.

[0311] (in conclusion)

[0312] As described above, the communication device (terminal device 40 or base station 20) according to the first embodiment of the present disclosure includes a signal processor (encoder 43 or decoder 23), a receiver, and a controller, wherein the signal processor includes an AI / ML model, the receiver receives information based on learning from other communication devices (base station 20 or terminal device 40), and the controller sets the AI / ML model based on the information based on learning. Such characteristics enable the communication device (terminal device 40 or base station 20) according to the first embodiment of the present disclosure to perform signal processing regardless of the differences between AI / ML models.

[0313] In addition, the dataset providing device 50 according to the first embodiment of the present disclosure is a dataset providing device that can be commonly accessed by communication devices of different vendors. The dataset providing device 50 includes a storage device 51, a receiver 52, and a transmitter, wherein the storage device 51 stores information on which the learning of the AI / ML model is based, the receiver 52 receives a request from the communication device to provide information on which the learning of the AI / ML model is based, and the transmitter sends the information on which the learning of the AI / ML model is based to the communication device. Such characteristics enable the base station 20 and the terminal device 40 to jointly obtain a piece of information on which learning is based from the dataset providing device 50, which information does not depend on the specifications or implementation of a specific vendor. This makes it possible to set and train AI / ML models that are interoperable between different vendors.

[0314] The technology according to the present disclosure is not limited to a specific standard, and any modifications may be made to the settings shown in the diagrams as appropriate. Note that the embodiments described above are examples for embodying the technology according to the present disclosure, and the technology according to the present invention may be embodied in various other embodiments. For example, various modifications, substitutions, omissions, or combinations thereof may be applied without departing from the scope of the present disclosure. Embodiments obtained by applying, for example, such modifications, substitutions, omissions, or combinations are also included within the scope of the present disclosure, as well as within the scope of the claimed embodiments of the present disclosure and their equivalents.

[0315] The processing procedures described in the present disclosure may be considered as a method including a series of processes. Alternatively, the processing procedures described in the present disclosure may be considered as a program for causing a computer to perform the series of processes, or a recording medium in which the program is stored. In addition, the above processing may be performed by a processor such as a CPU of a computer. In addition, the type of recording medium is not particularly limited, because the embodiments of the present disclosure are not affected by the type of recording medium.

[0316] In this disclosure Figure 2 to Figure 7The structural elements illustrated in the diagram may be provided by software, or may be implemented by hardware. For example, each structural element may be a software module provided by software such as a microprogram, and may be provided by a processor that executes the software module. Alternatively, each structural element may be implemented by a circuit block on a semiconductor chip (bare die), i.e., for example, an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). In addition, the number of structural elements does not necessarily have to be equal to the number of hardware that implements the structural elements. For example, multiple structural elements may be implemented by a single processor or circuit. On the contrary, a single structural element may be implemented by multiple processors or circuits.

[0317] The type of processor described in the present disclosure is not limited. For example, the processor may be a CPU, a microprocessing unit (MPU), or a graphics processing unit (GPU).

[0318] Furthermore, the present disclosure may also adopt the following configurations.

[0319] [1] A communication device, comprising:

[0320] a signal processor, the signal processor comprising an AI / ML model;

[0321] a receiver that receives information on which the learning is based from other communication devices; and

[0322] A controller that sets the AI / ML model based on the information on which the learning is based.

[0323] [2] The communication device according to [1], wherein

[0324] The information on which the learning is based includes at least one of the type of AI / ML model, the number of input nodes, the number of output nodes, the layer composition, the value of the weighting coefficient, the layer for learning, and the learning data.

[0325] [3] The communication device according to [1] or [2], wherein

[0326] The information on which the learning is based is determined based on at least one of a cell in which the communication device is located, a frequency band used by the communication device, and a synchronization signal block (SSB) used by the communication device.

[0327] [4] The communication device according to any one of [1] to [3], wherein

[0328] The information on which the learning is based is included in one of the following items: system information, radio resource control (RRC) signaling, MAC control element (MAC CE) and downlink control information (DCI) sent by the other communication device.

[0329] [5] The communication device according to [4], wherein

[0330] When the communication device is in a radio resource control (RRC) idle state or in a radio resource control (RRC) inactive state, the information on which the learning is based is included in the system information.

[0331] [6] The communication device according to [5], wherein

[0332] Only when the communication device has the capability of performing signal processing based on the AI / ML model, the controller acquires the information on which the learning is based included in the system information.

[0333] [7] The communication device according to [4], wherein

[0334] When the communication device is in a radio resource control (RRC) connected state, the information on which the learning is based is included in RRC signaling, MAC CE or DCI.

[0335] [8] The communication device according to any one of [1] to [7], wherein

[0336] The receiver also receives information from the other communication device regarding whether additional learning is permitted, and

[0337] When the information on whether additional learning is allowed is set to "not allowed", the controller does not additionally train the AI / ML model after setting the AI / ML model based on the information on which the learning is based.

[0338] [9] The communication device according to any one of [1] to [7], wherein

[0339] The controller trains the AI / ML model after setting the AI / ML model based on the information on which the learning is based.

[0340]

[10] The communication device according to [9] further includes

[0341] a transmitter, the transmitter sending a signal to the other communication device, wherein

[0342] After training the AI / ML model, the controller feeds back information about the learned AI / ML model to the other communication devices through the transmitter.

[0343]

[11] The communication device according to any one of [1] to

[10] , further comprising

[0344] a transmitter, the transmitter sending a signal to the other communication device, wherein

[0345] When the AI / ML model has been learned, the controller transmits information about the AI / ML model that has been learned to the other communication device through the transmitter before receiving information on which the learning is based from the other communication device.

[0346]

[12] The communication device according to any one of [1] to

[11] , further comprising

[0347] a transmitter, the transmitter sending a signal to the other communication device, wherein

[0348] The controller transmits capability information about the capability of the communication device to the other communication device through the transmitter.

[0349]

[13] The communication device according to

[12] , wherein

[0350] The capability information includes information about AI / ML models that the communication device can support.

[0351]

[14] The communication device according to

[12] or

[13] , wherein

[0352] The capability information includes information about an AI / ML model that is included in the communication device and has been learned.

[0353]

[15] A communication device according to any one of

[12] to

[14] , wherein

[0354] The capability information includes information about a remaining processing capability of the communication device and / or information about a maximum value of the processing capability of the communication device.

[0355]

[16] The communication device according to

[15] , wherein

[0356] When the other communication device requests the capability information, the capability information is sent.

[0357]

[17] The communication device according to

[15] , wherein

[0358] The capability information is periodically transmitted based on instructions regarding frequency and timing received from the other communication device, or is periodically transmitted based on a predetermined frequency and a predetermined timing.

[0359]

[18] A communication device according to any one of [1] to

[17] , wherein

[0360] The signal processor includes multiple AI / ML models, and

[0361] The controller selects one of the multiple AI / ML models based on at least one of the cell where the communication device is located, the frequency band used by the communication device, and the SSB used by the communication device, and uses the selected one of the multiple AI / ML models.

[0362]

[19] A communication device according to any one of [1] to

[18] , wherein

[0363] The signal processor includes multiple AI / ML models, and

[0364] The controller selects one of the plurality of AI / ML models according to a radio resource control (RRC) state of the communication device and uses the selected one of the plurality of AI / ML models.

[0365]

[20] The communication device according to

[19] , wherein

[0366] When the communication device is in an RRC idle state,

[0367] The communication device uses an AI / ML model set based on information on which learning is based included in the system information received from the other communication device, or

[0368] The communication device uses an AI / ML model that is set and trained based on the learning-based information included in the system information received from the other communication devices.

[0369]

[21] The communication device according to

[19] or

[20] , wherein

[0370] When the communication device is in an RRC inactive state,

[0371] The communication device uses an AI / ML model set based on information on which learning is based received from the other communication device during the RRC connection, or

[0372] The communication device uses an AI / ML model that is set and trained based on information received from the other communication device during the RRC connection, or

[0373] The communication device uses an AI / ML model set based on information on which learning is based included in the system information received from the other communication device, or

[0374] The communication device uses an AI / ML model that is set and trained based on the learning-based information included in the system information received from the other communication devices.

[0375]

[22] A communication device according to any one of [1] to

[21] , wherein

[0376] The further communication device receives information on which the learning is based from a data set providing device which is accessible jointly by communication devices of different vendors.

[0377]

[23] A communication device according to any one of [1] to

[22] , wherein

[0378] The receiver also receives from the other communication devices information required to access a data set providing device, the data set providing device being commonly accessible by communication devices of different vendors.

[0379]

[24] A communication device according to any one of [1] to

[23] , wherein

[0380] The other communication device is a terminal device,

[0381] The communication device is a base station that establishes wireless communication with the terminal device, and

[0382] The communication device also includes a transmitter, which sends at least one of information about an AI / ML model that has been learned, information on which the learning of the AI / ML model is based, and capability information about the capabilities of the AI / ML model to other base stations, and the sending is performed when the terminal device switches from the base station to the other base stations.

[0383]

[25] A data set providing device capable of being accessed by communication devices of different suppliers, the data set providing device comprising:

[0384] A storage device that stores information on which the learning of the AI / ML model is based;

[0385] a receiver that receives a request from a communication device for providing information upon which learning of the AI / ML model is based; and

[0386] A transmitter that sends information on which the learning of the AI / ML model is based to the communication device.

[0387]

[26] The data set providing device according to

[25] , wherein

[0388] The data set providing means is defined as a network function.

[0389]

[27] A method for training an AI / ML model, the method comprising:

[0390] receive information on which learning is based; and

[0391] Set up the AI / ML model based on the information it is learning from.

[0392]

[28] A method for providing information on which learning of an AI / ML model is based, the method comprising:

[0393] receiving a request for providing information upon which the learning is based; and

[0394] The information on which the learning is based is sent to the transmission source which has sent the request for provision.

[0395] Reference numerals list

[0396] 1 Wireless Communication System

[0397] 10 Management Device

[0398] 11 Ministry of Communications

[0399] 12 Storage Devices

[0400] 13 Controller

[0401] 20 Base station (communication device, other communication device)

[0402] 21 Wireless Communications Department

[0403] 211 Transmitter

[0404] 212 Receiver

[0405] 213 Antenna

[0406] 22 Storage Devices

[0407] 23 Decoder (Signal Processor)

[0408] 25 Controller

[0409] 30 Relay station (communication device, other communication device)

[0410] 31 Wireless Communications Department

[0411] 311 Transmitter

[0412] 312 Receiver

[0413] 313 Antenna

[0414] 32 Storage Devices

[0415] 33 Decoder (Signal Processor)

[0416] 35 Controller

[0417] 40 Terminal device (communication device, other communication device)

[0418] 41 Wireless Communications Department

[0419] 411 Transmitter

[0420] 412 Receiver

[0421] 413 Antenna

[0422] 42 Storage Devices

[0423] 43 Decoder (Signal Processor)

[0424] 45 Controller

[0425] 50 Dataset providing device

[0426] 51 Storage device

[0427] 52 Receiver

[0428] 53 Transmitter

Claims

1. A communication device, comprising: a signal processor, the signal processor comprising an AI / ML model; a receiver that receives information on which the learning is based from other communication devices; as well as A controller that sets the AI / ML model based on the information on which the learning is based.

2. The communication device according to claim 1, wherein The information on which the learning is based includes at least one of the type of AI / ML model, the number of input nodes, the number of output nodes, the layer composition, the value of the weighting coefficient, the layer for learning, and the learning data.

3. The communication device according to claim 1, wherein The information on which the learning is based is determined based on at least one of a cell in which the communication device is located, a frequency band used by the communication device, and a synchronization signal block SSB used by the communication device.

4. The communication device according to claim 1, wherein The information on which the learning is based is included in one of the following items: system information, radio resource control RRC signaling, MAC control element MAC CE and downlink control information DCI sent by the other communication device.

5. The communication device according to claim 4, wherein When the communication device is in a radio resource control RRC idle state or in a radio resource control RRC inactive state, the information on which the learning is based is included in the system information.

6. The communication device according to claim 5, wherein Only when the communication device has the capability of performing signal processing based on the AI / ML model, the controller acquires the information on which the learning is based included in the system information.

7. The communication device according to claim 4, wherein When the communication device is in a radio resource control RRC connected state, the information based on which the learning is based is included in RRC signaling, MAC CE or DCI.

8. The communication device according to claim 1, wherein The receiver also receives information from the other communication device regarding whether additional learning is permitted, and When the information about whether additional learning is allowed is set to “not allowed”, the controller does not additionally train the AI / ML model after setting the AI / ML model based on the information on which the learning is based.

9. The communication device according to claim 1, wherein The controller trains the AI / ML model after setting the AI / ML model based on the information on which the learning is based.

10. The communication device according to claim 9, further comprising: a transmitter, the transmitter sending a signal to the other communication device, wherein After training the AI / ML model, the controller feeds back information about the learned AI / ML model to the other communication devices through the transmitter.

11. The communication device according to claim 1, further comprising: a transmitter, the transmitter sending a signal to the other communication device, wherein When the AI / ML model has been learned, the controller transmits information about the AI / ML model that has been learned to the other communication device through the transmitter before receiving information on which the learning is based from the other communication device.

12. The communication device according to claim 1, further comprising: a transmitter, the transmitter sending a signal to the other communication device, wherein The controller transmits capability information about the capability of the communication device to the other communication device through the transmitter.

13. The communication device according to claim 12, wherein The capability information includes information about AI / ML models that the communication device can support.

14. The communication device according to claim 12, wherein The capability information includes information about an AI / ML model that is included in the communication device and has been learned.

15. The communication device according to claim 12, wherein The capability information includes information about a remaining processing capability of the communication device and / or information about a maximum value of the processing capability of the communication device.

16. The communication device according to claim 15, wherein When the other communication device requests the capability information, the capability information is sent.

17. The communication device according to claim 15, wherein The capability information is periodically transmitted based on instructions regarding frequency and timing received from the other communication device, or is periodically transmitted based on a predetermined frequency and a predetermined timing.

18. The communication device according to claim 1, wherein The signal processor includes a plurality of AI / ML models, and The controller selects one of the multiple AI / ML models based on at least one of the cell where the communication device is located, the frequency band used by the communication device, and the SSB used by the communication device, and uses the selected one of the multiple AI / ML models.

19. The communication device according to claim 1, wherein The signal processor includes a plurality of AI / ML models, and The controller selects one of the plurality of AI / ML models according to a radio resource control (RRC) state of the communication device, and uses the selected one of the plurality of AI / ML models.

20. The communication device according to claim 19, wherein When the communication device is in an RRC idle state, The communication device uses an AI / ML model set based on information on which learning is based included in the system information received from the other communication device, or The communication device uses an AI / ML model that is set and trained based on the learning-based information included in the system information received from the other communication devices.

21. The communication device according to claim 19, wherein When the communication device is in an RRC inactive state, The communication device uses an AI / ML model set based on information on which learning is based received from the other communication device during the RRC connection, or The communication device uses an AI / ML model that is set and trained based on information received from the other communication device during the RRC connection, or The communication device uses an AI / ML model set based on information on which learning is based included in the system information received from the other communication device, or The communication device uses an AI / ML model that is set and trained based on the learning-based information included in the system information received from the other communication devices.

22. The communication device according to claim 1, wherein The further communication device receives information on which the learning is based from a data set providing device which is accessible jointly by communication devices of different vendors.

23. The communication device according to claim 1, wherein The receiver also receives from the other communication devices information required to access a data set providing device, the data set providing device being commonly accessible by communication devices of different vendors.

24. The communication device according to claim 1, wherein The other communication device is a terminal device, The communication device is a base station that establishes wireless communication with the terminal device, and The communication device also includes a transmitter, which sends at least one of information about an AI / ML model that has been learned, information on which the learning of the AI / ML model is based, and capability information about the capabilities of the AI / ML model to other base stations, and the sending is performed when the terminal device switches from the base station to the other base stations.

25. A data set providing device capable of being accessed by communication devices of different suppliers, the data set providing device comprising: A storage device that stores information on which the learning of the AI / ML model is based; a receiver that receives from a communication device a request for providing information on which learning of the AI / ML model is based; as well as A transmitter that sends information on which the learning of the AI / ML model is based to the communication device.

26. The data set providing device according to claim 25, wherein The data set providing means is defined as a network function.

27. A method for training an AI / ML model, the method comprising: receive the information upon which learning is based; as well as Set up the AI / ML model based on the information it is learning from.

28. A method for providing information upon which learning of an AI / ML model is based, the method comprising: receiving requests for information upon which learning is based; as well as The information on which the learning is based is sent to the transmission source which has sent the request for provision.

Images