Wireless communication system

Abstract

A communication system provided with a first UE (110A) and a second UE (110B) that communicates with the first UE (110A) by means of SL communication, the first UE (110A) transmitting, to the second UE (110B), encoded channel state information obtained by encoding using an encoding model trained for channel state information. The channel state information is information for reporting the channel state of SL communication, and the second UE (110B) decodes the encoded channel state information using a decoding model trained for the channel state information.

Description

Technical Field

[0001] The present disclosure relates to wireless communication systems. Background Art

[0002] In 3GPP (3rd Generation Partnership Project), which is a standardization organization for mobile communication systems, standards for various services and applications using SL (Side Link) communication (also called PC5 communication) are being developed.

[0003] For example, in SL communication, similar to the adaptive transmission technology used in communications between terminal devices and base stations to enable rank adaptation (adaptive control of the number of transmission layers) in adaptive modulation or MIMO (Multipl-Input Multiple-Output) transmission, the terminal device sending data sends SL-CSI (Channel State Information)-RS (Reference Signal), and the terminal device receiving data receives the SL-CSI-RS and analyzes the channel state, and sends the analysis result as an SL-CSI report to the terminal device sending the data.

[0004] Prior art literature

[0005] Non-patent literature

[0006] Non-patent document 1: 3GPP TS38.473 V16.6.0 Summary of the invention

[0007] Problem that the invention aims to solve

[0008] However, in conventional communication systems, there is a problem that bit data of SL-CSI reports occupy radio resources, which limits the radio resources available for user data.

[0009] In view of the above-mentioned problems, one of the objectives of the present disclosure is to apply AI (Artificial Intelligence) to encoding and decoding of data in SL communication to achieve efficient communication system.

[0010] Means used to solve problems

[0011] A wireless communication system of one embodiment of the present invention comprises a first terminal device and a second terminal device communicating with the first terminal device via side link communication, i.e., SL communication, and is characterized in that the first terminal device sends coded channel state information to the second terminal device, the coded channel state information being channel state information encoded using a coding model trained for channel state information, the channel state information being information for reporting the channel state of the SL communication, and the second terminal device decodes the coded channel state information using a decoding model trained for the channel state information.

[0012] Effects of the Invention

[0013] According to one or more aspects of the present disclosure, AI can be applied to encoding and decoding of data in SL communication, thereby achieving high efficiency of the communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a block diagram schematically showing the configuration of a communication system according to Embodiments 1 to 4.

[0015] Figure 2 This is a block diagram schematically showing the configuration of a UE in Embodiments 1 to 4.

[0016] Figure 3 (A) and (B) are block diagrams showing hardware configuration examples.

[0017] Figure 4 This is a block diagram schematically showing the configuration of a base station in Embodiments 1 to 4.

[0018] Figure 5 This is a block diagram schematically showing the configuration of a host device.

[0019] Figure 6 A structural example is shown in which the UE performs training of a coding model and a decoding model in Implementation Example 1 and the UE performs inference using the trained coding model and decoding model.

[0020] Figure 7 This is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 1.

[0021] Figure 8 This is a schematic diagram showing an example of a neural network used in training the encoding model and the decoding model.

[0022] Fig. 9 A structural example is shown in Implementation Example 2 in which the base station trains the encoding model and the decoding model and the UE performs inference using the trained encoding model and the decoding model.

[0023] Fig.10 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 2.

[0024] Fig.11 A structural example is shown in Implementation Example 3 in which the base station trains the encoding model and the decoding model and the UE uses the trained encoding model and the decoding model for inference.

[0025] Fig.12 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 3.

[0026] Fig.13 A structural example is shown in Implementation Example 4 in which the U2N relay trains the encoding model and the decoding model and the UE uses the trained encoding model and the decoding model for inference.

[0027] Fig.14 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 4. DETAILED DESCRIPTION

[0028] Implementation method 1.

[0029] Figure 1 This is a block diagram schematically showing the configuration of the communication system 100 according to the first embodiment.

[0030] The communication system 100 includes a UE (User Equipment) 110 as a terminal device, a base station 130 as a base station device, and a host device 160. The UE 110 here is a mobile terminal device, but the present invention is not limited to this example.

[0031] UE 110 can perform wireless communication with base station 130 and transmit and receive signals through wireless communication. UE 110 can also perform communication between UEs 110 through SL communication and can also perform communication via a relay.

[0032] The base stations 130 constitute a wireless access network.

[0033] Here, the wireless access network is assumed to be a fifth generation (hereinafter also referred to as "5G") wireless access system for illustration, but the present invention is not limited to such an example.

[0034] In the case where the radio access network is a 5G method discussed in 3GPP, the radio access network is called NG-RAN (Next Generation Radio Access Network). In this case, the base station 130 is called gNB (NG-RAN NodeB). The core network is called 5G Core (5G Core). In addition, in the case where the radio access network is an LTE method discussed in 3GPP, the radio access network is called E-UTRAN (Evolved Universal Terrestrial Radio Access Network). In this case, the base station 130 is called eNB (E-UTRAN NodeB).

[0035] Figure 2 is a block diagram schematically showing the structure of UE 110.

[0036] The UE 110 includes a terminal-side communication unit 111 , an application unit (hereinafter referred to as an AP unit) 120 , a protocol processing unit 121 , a control unit 122 , and a storage unit 123 .

[0037] First, the transmission processing of UE 110 is described.

[0038] Control data from the protocol processing unit 121 or user data from the AP unit 120 is provided to the terminal side communication unit 111 .

[0039] The terminal-side communication unit 111 converts the data into a transmission signal for wireless communication, and transmits the transmission signal to a destination using one or more antennas.

[0040] Next, the reception processing of UE 110 is described.

[0041] The terminal side communication unit 111 uses one or more antennas to receive a wireless signal from a transmission source as a reception signal, and generates control data or user data based on the reception signal. Then, the terminal side communication unit 111 sends the control data to the protocol processing unit 121 and sends the user data to the AP unit 120.

[0042] The series of processes in the UE 110 as described above is controlled by the control unit 122. Although not shown in the figure, the control unit 122 is connected to the terminal side communication unit 111, the AP unit 120, and the protocol processing unit 121.

[0043] In addition, the number of antennas used by UE 110 for transmission and the number of antennas used for reception may be the same or different.

[0044] Here, the AP unit 120 in Embodiment 1 functions as a training unit that performs a training function or an inference unit that performs an inference function. The processing here will be described later.

[0045] The storage unit 123 stores programs and data necessary for processing in the UE 110. For example, the storage unit 123 stores at least one of a coding model and a decoding model described later.

[0046] Part or all of the AP unit 120, the protocol processing unit 121, and the control unit 122 described above can be implemented as follows, for example. Figure 3 As shown in (A), it is composed of a memory 10 and a processor 11 such as a CPU (Central Processing Unit) that executes the program stored in the memory 10. Such a program can be provided via a network, or can be provided by recording in a recording medium. That is, such a program can also be provided as a program product, for example.

[0047] In addition, part or all of the AP unit 120, the protocol processing unit 121, and the control unit 122 may also be configured as follows. Figure 3 As shown in (B), the processing circuit 12 is composed of a single circuit, a composite circuit, a processor operated by a program, a parallel processor operated by a program, an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0048] As described above, the AP unit 120, the protocol processing unit 121, and the control unit 122 can be realized by a processing circuit network.

[0049] In addition, the terminal side communication unit 111 can be implemented by a wireless communication interface which is an interface for wireless communication.

[0050] The storage unit 123 can be implemented by a nonvolatile memory, or a storage device such as a HDD (Hard Disk Drive) or an SSD (Sorid State Drive).

[0051] Figure 4 It is a block diagram schematically showing the structure of the base station 130 .

[0052] The base station 130 includes a station-side communication unit 131 , a communication processing unit 140 , a protocol processing unit 144 , a control unit 145 , and a storage unit 146 .

[0053] The communication processing unit 140 includes: a 5GC communication unit 141, which performs data transmission and reception between 5GCs such as the upper device 160; an EPC communication unit 142, which performs data transmission and reception between EPCs (Evolved Packet Cores) such as MME (Mobility Management Entity); and other base station communication units 143 that perform data transmission and reception between other base stations. In addition, the 5GC communication unit 141, the EPC communication unit 142, and the other base station communication units 143 respectively exchange information with the protocol processing unit 144.

[0054] First, the transmission processing of the base station 130 will be described.

[0055] Control data from the protocol processing unit 144, or user data or control data from the 5GC communication unit 141, the EPC communication unit 142 or other base station communication units 143 is provided to the station side communication unit 131.

[0056] The station-side communication unit 131 converts the data into a transmission signal for wireless communication, and transmits the transmission signal to a destination using one or more antennas.

[0057] Next, the reception processing of the base station 130 will be described.

[0058] The station side communication unit 131 uses one or more antennas to receive a wireless signal from a transmission source as a received signal, and generates control data or user data based on the received signal. Then, the station side communication unit 131 sends control data to the protocol processing unit 144, the 5GC communication unit 141, the EPC communication unit 142, or other base station communication units 143, and sends user data to the 5GC communication unit 141, the EPC communication unit 142, or other base station communication units 143.

[0059] The above-described series of processes in the base station 130 are controlled by the control unit 145. Although not shown in the figure, the control unit 145 is connected to the station-side communication unit 131, the communication processing unit 140, and the protocol processing unit 144.

[0060] In addition, the number of antennas used for transmission and the number of antennas used for reception of the base station 130 may be the same or different.

[0061] In addition, the number of antennas of UE 110 and the number of antennas of base station 130 may be the same or different.

[0062] The storage unit 146 stores programs and data necessary for processing in the base station 130 .

[0063] Part or all of the communication processing unit 140, the protocol processing unit 144, and the control unit 145 described above can be, for example, Figure 3 As shown in (A), it is composed of a memory 10 and a processor 11 such as a CPU that executes the program stored in the memory 10. Such a program can be provided via a network, or can be provided by recording in a recording medium. That is, such a program can also be provided as a program product, for example.

[0064] In addition, part or all of the communication processing unit 140, the protocol processing unit 144, and the control unit 145 may also be configured as follows. Figure 3 As shown in (B), the processing circuit 12 is composed of a single circuit, a composite circuit, a processor operated by a program, a parallel processor operated by a program, an ASIC, or an FPGA.

[0065] As described above, the communication processing unit 140 , the protocol processing unit 144 , and the control unit 145 can be realized by a processing circuit network.

[0066] In addition, the station-side communication unit 131 can be realized by a wireless communication interface which is an interface for wireless communication.

[0067] The storage unit 126 can be implemented by a nonvolatile memory or a storage device such as a HDD (Hard Disk Drive) or an SSD (Sorid State Drive).

[0068] Figure 5 It is a block diagram schematically showing the configuration of the host device 160 .

[0069] In the present embodiment, the 5G method is used as the wireless access network, and therefore the host device 160 is also referred to as a 5GC (5G Core) unit.

[0070] Figure 5 Shown in Figure 1 The upper device 160 shown includes the structure of AMF (Access and Mobility Management Function), the structure of SMF (Session Management Function), and the structure of UPF (User Plane Function).

[0071] The host device 160 includes a data network communication unit 161 , a base station communication unit 162 , a user plane communication unit 163 , a session management unit 164 , a control plane control unit 165 , and a control unit 169 .

[0072] The data network communication unit 161 performs transmission and reception of data between the host device 160 and the data network.

[0073] The base station communication unit 162 performs transmission and reception of data between the host device 160 and the base station 130 based on the NG interface.

[0074] When the data received from the data network is user data, the user data is transmitted from the data network communication unit 161 to the base station communication unit 162 via the user plane communication unit 163 , and then transmitted from the base station communication unit 162 to the base station 130 .

[0075] When the data received from the base station 130 is user data, the user data is sent from the base station communication unit 162 to the data network communication unit 161 via the user plane communication unit 163 , and is transmitted from the data network communication unit 161 to the data network.

[0076] When the data received from the data network is control data, the control data is sent from the data network communication unit 161 to the session management unit 164 via the user plane communication unit 163 .

[0077] The session management unit 164 transmits control data to the control plane control unit 165 .

[0078] When the data received from the base station 130 is control data, the control data is sent from the base station communication unit 162 to the control plane control unit 165. The control plane control unit 165 sends the control data to the session management unit 164.

[0079] The control plane control unit 165 performs all processing for the control plane (hereinafter sometimes referred to as C-Plane).

[0080] The control plane control unit 165 includes a NAS (Non-Access Stratum) security unit 166 , a PDU session control unit 167 , and an idle state mobility management unit 168 .

[0081] The NAS security unit 166 is responsible for the security of NAS messages, etc.

[0082] The PDU session control unit 167 performs management of the PDU session between the UE 110 and the host device 160 , and the like.

[0083] The idle state mobility management unit 168 performs mobility management in the waiting state, generation and control of paging signals in the waiting state, addition, deletion, update and retrieval of tracking areas of one or more UEs 110 under control, and tracking area list management, etc. In addition, the waiting state is also called an idle state, an RRC_IDLE state, or simply idle.

[0084] A series of processes of the host device 160 are controlled by the control unit 169. Although not shown, the control unit 169 is connected to the data network communication unit 161, the base station communication unit 162, the user plane communication unit 163, the session management unit 164, and the control plane control unit 165.

[0085] For example, part or all of the data network communication unit 161, the base station communication unit 162, the user plane communication unit 163, the session management unit 164, the control plane control unit 165, and the control unit 169 can be configured as follows: Figure 3 As shown in (A), it is composed of a memory 10 and a processor 11 such as a CPU that executes the program stored in the memory 10. Such a program can be provided via a network, or can be provided by recording in a recording medium. That is, such a program can also be provided as a program product, for example.

[0086] In addition, part or all of the data network communication unit 161, the base station communication unit 162, the user plane communication unit 163, the session management unit 164, the control plane control unit 165, and the control unit 169 may also be configured as follows. Figure 3 As shown in (B), the processing circuit 12 is composed of a single circuit, a composite circuit, a processor operated by a program, a parallel processor operated by a program, an ASIC, or an FPGA.

[0087] As described above, the data network communication unit 161 , the base station communication unit 162 , the user plane communication unit 163 , the session management unit 164 , the control plane control unit 165 , and the control unit 169 can be realized by a processing circuit network.

[0088] In implementation mode 1, adaptive transmission technology can also be applied in SL communication. At this time, the data transmission side can also use SL-CSI-RS and SL-CSI report to grasp the channel state between each transmission antenna and each reception antenna. In this case, UE 110 receiving data receives SL-CSI-RS sent from UE 110 on the data transmission side and analyzes the channel state, and sends the analysis result as SL-CSI report to the terminal device on the data transmission side.

[0089] In addition, SL-CSI can reduce the amount of bit data by generating a coding model and a decoding model using a training device and applying them.

[0090] In Embodiment 1, UE 110 that receives data performs training of a coding model and a decoding model. The following describes the structure thereof.

[0091] Figure 6A configuration example is shown in which, in Embodiment 1, UE 110 trains a coding model and a decoding model and UE 110 performs inference using the trained coding model and decoding model.

[0092] In the first embodiment, the second UE 110B trains the encoding model and the decoding model, and the first UE 110A and the second UE 110B communicate using the trained models.

[0093] exist Figure 6 In the example, the first UE 110A is a data receiving-side terminal device that sends SL-CSI, and the second UE 110B is a data sending-side terminal device that receives SL-CSI.

[0094] Figure 7 This is a timing diagram showing the process of training the encoding model and the decoding model and applying the trained models in Implementation 1.

[0095] First, the control unit 122 of the first UE 110A and the control unit 122 of the second UE 110B establish SL communication via the PC5 interface via their respective protocol processing units 121 and terminal-side communication units 111 (S10). As a result, the first UE 110A and the second UE 110B are able to communicate.

[0096] Next, the control unit 145 of the base station 130 transmits the initial coding model as the coding model in the initial state and the initial decoding model as the decoding model in the initial state to each connected UE 110 via the protocol processing unit 144, the communication processing unit 140, and the station side communication unit 131 (S11). Here, the second UE 110B that performs data transmission receives these models. Specifically, the control unit 122 of the second UE 110B obtains the initial coding model and the initial decoding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores the initial coding model and the initial decoding model in the storage unit 123.

[0097] Here, the encoding model data and the decoding model data are respectively constituted by, for example, the following information (1) to (11).

[0098] (1) the category of input values, (2) the number of nodes in the input layer, (3) the depth of the hidden layer, (4) the number of nodes in the hidden layer, (5) the function applied to each node in the hidden layer, (6) the number of nodes in the output layer, (7) the function applied to each node in the output layer, (8) the weight coefficient of the branch, (9) the information related to the nodes connected to the branch, (10) the information related to the default action mode of the training device, and (11) the binary information of the model.

[0099] The "input value category" may be, for example, information indicating that it is a SL-CSI report. The information may also include the details of the SL-CSI report to be input, for example, information related to the resources of the SL-CSI-RS of the measurement object. The resources of the SL-CSI-RS of the measurement object are, for example, time, frequency, or antenna port. UE 110 may also use this information to perform encoding processing of the SL-CSI report. In this way, for example, erroneous actions caused by inconsistent identification between UE 110 can be avoided.

[0100] In addition, the "input value category" may also include information related to the category of the SL-CSI-RS of the measurement object. This information may be, for example, ZP (Zero-Power)-CSI-RS, NZP (Non-Zero-Power)-CSI-RS, or CSI-IM (Channel State Information-Interference Measurement). Thus, for example, the flexibility of channel measurement can be improved.

[0101] In addition, ZP-CSI-RS and NZP-CSI-RS are specified in 3GPP TS38.211, and CSI-IM is specified in 3GPP TS38.214.

[0102] In addition, the "input value category" may also be information related to other input values. For example, the "input value category" may also be information related to the positioning signal reception result. Thus, for example, encoding or decoding may also be performed for other input values, and as a result, the efficiency in the communication system may be improved.

[0103] The "number of nodes in the input layer" may be, for example, the number of nodes in the input layer itself, or an identifier indicating the number of nodes in the input layer. The correspondence between the number of nodes in the input layer and the identifier may be determined in advance according to a standard, or may be determined by the base station 130 and notified to the UE 110, or may be determined by a core NW (Network) device (not shown) having an AMF and notified to at least one of the base station 130 and the UE 110. The UE 110 may also use this information to construct an input layer of a coding model. Thus, for example, a coding model may be constructed in the UE 110.

[0104] The “depth of the hidden layer” may be information related to the depth of the hidden layer in the coding unit (not shown). The UE 110 may also construct a coding model using this information. Thus, for example, the UE 110 may construct a coding model.

[0105] The "number of nodes in the hidden layer" may also include information related to the number of nodes in each hidden layer. Instead of the information of the number of nodes itself, an identifier representing the number of nodes may also be used. The correspondence between the number of nodes in the hidden layer and the identifier may be determined in advance according to a standard, or may be determined by the base station 130 and notified to the UE 110, for example, may be determined by a core NW device (not shown) having an AMF and notified to at least one of the base station 130 and the UE 110. The UE 110 may also use this information to construct a hidden layer of a coding model. Thus, for example, a coding model can be constructed in the UE 110.

[0106] The "function applied to each node of the hidden layer" may also include information related to the function applied to the input value of each hidden layer. The function may be, for example, a Sigmoid function or a hyperbolic tangent function. As another example, the function may be a linear function or no function may be applied, in other words, the input value and output value for the hidden layer are equal. UE110 may also use this information to construct the hidden layer of the coding model. By using the information of the "function applied to each node of the hidden layer", for example, the flexibility of at least one of the coding model and the decoding model can be improved.

[0107] In addition, the "function applied to each node of the hidden layer" may also include information related to the offset applied to the input value of each hidden layer. The offset may be, for example, a value applied to the value obtained by adding the product of the weighting coefficient input to each hidden layer and the value of the previous layer for each branch. With this information, for example, the flexibility of at least one of the encoding model and the decoding model can be improved.

[0108] In addition, information related to "functions applied to each node of the hidden layer" can also be provided for each node. Thus, for example, the flexibility of at least one of the encoding model and the decoding model can be improved. As another example, the information can also be provided for each hidden layer. In other words, the information can also be shared between nodes in the same layer. Thus, for example, the signaling amount of the information can be reduced. As another example, the information can also be shared between nodes. Thus, for example, the signaling amount of the information can be further reduced.

[0109] The "number of nodes in the output layer" may also include information related to the number of nodes in the output layer. An identifier representing the number of nodes may also be used instead of the information about the number of nodes itself. The correspondence between the number of nodes in the output layer and the identifier may be determined in advance according to a standard, or may be determined by the base station 130 and notified to the UE 110, for example, may be determined by a core NW device (not shown) having an AMF and notified to at least one of the base station 130 and the UE 110. The UE 110 may also use this information to construct an output layer of a coding model. Thus, for example, a coding model can be constructed in the UE 110.

[0110] The “function applied to each node of the output layer” may be the same information as the “function applied to each node of the hidden layer.” This can achieve the same effect as the aforementioned “function applied to each node of the hidden layer.”

[0111] Information about the "function applied to each node of the output layer" can also be provided for each node. This can, for example, improve the flexibility of at least one of the encoding model and the decoding model. As another example, this information can also be shared between nodes. This can, for example, also reduce the amount of signaling of this information.

[0112] The information of “branch weight coefficient” may be provided for each branch. UE 110 may use this information to configure branches between nodes. Thus, for example, a coding model may be constructed in UE 110.

[0113] The "information related to the node connected to the branch" may also include information related to the nodes at both ends of the branch. The information related to the node may be provided, for example, using an identifier of the entire node, or may be provided using a combination of information identifying a layer and information identifying a node in the layer. The UE 110 may also use this information to form a branch between nodes. Thus, for example, a coding model can be constructed in the UE 110.

[0114] In addition, regarding "information related to the node connected to the branch", both ends of the branch can also span one or more layers. As another example, both ends of the branch can be nodes of the same layer or connected to the same node. This can, for example, improve the flexibility of the coding model.

[0115] The "information related to the default operation mode of the training device" may also include information related to the operation status of the encoding unit (not shown) of the UE 110 after the notification. The information may be, for example, information indicating that encoding is to be performed or information indicating that encoding is not to be performed. The UE 110 may also use the information to perform encoding or not to perform encoding. Thus, for example, it is possible to prevent the operation mode of the training device between the UEs 110 from being inconsistent, and as a result, it is possible to prevent erroneous operation in the communication system 100.

[0116] The "binary information of the model" may be notified, for example, when base station 130 and UE 110 use the same platform. UE 110 may also use this information to construct a coding model. This may reduce the amount of signaling between UEs 110, for example.

[0117] The initial coding model and the initial decoding model may be models specific to the base station 130, or may be shared by all base stations 130. Alternatively, a unique initial coding model and initial complex model may be allocated to each terminal.

[0118] In addition, here, the initial encoding model and the initial decoding model are stored in the storage unit 146 of the base station 130.

[0119] Next, the control unit 122 of the second UE 110B decides to perform training and applies the coding model and the decoding model to the SL-CSI feedback from the first UE 110A. Alternatively, the base station may decide and instruct the second UE 100B in step S11. Then, the control unit 122 forwards the initial coding model received from the base station 130 in step S11 to the first UE 110A via the protocol processing unit 121 and the terminal side communication unit 111 (S12). The control unit 122 of the first UE 110A receives the initial coding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores it in the storage unit 123.

[0120] Next, the control unit 122 of the second UE 110B instructs the first UE 110A to apply the received coding model via the protocol processing unit 121 and the terminal side communication unit 111 (S13).

[0121] The control unit 122 of the first UE 110A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the coding model stored in the storage unit 123 according to the instruction (S14).

[0122] Furthermore, in the second UE 110B, the control unit 122 also applies the decoding model stored in the storage unit 123 ( S15 ).

[0123] Next, in order to perform training, the control unit 122 of the second UE 110B sends the SL-CSI-RS via the protocol processing unit 121 and the terminal side communication unit 111, and requests the first UE 110A to send the SL-CSI report for training (S16). At this time, the control unit 122 may also specify the number of times N (N is an integer greater than 1) to send the SL-CSI report for training in the request.

[0124] The control unit 122 of the first UE 110A receives the request via the terminal side communication unit 111 and the protocol processing unit 121. Then, the control unit 122 transmits the SL-CSI report for training via the protocol processing unit 121 and the terminal side communication unit 111 in response to the request (S17).

[0125] In the second UE 110B that has received the SL-CSI report, the control unit 122 receives the SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121, and performs training of the encoding model and the decoding model (S18).

[0126] For example, the control unit 122 of the second UE 110B trains the encoding model and the decoding model by so-called supervised training according to a known neural network model. Here, supervised training refers to a method in which a set of data of an input and a label representing the result is provided as training data to a training device, and the training device trains the features existing in the training data and infers the result based on the input. Here, the control unit 122 functions as a training device. When training the encoding model and the decoding model, the input data itself may be used as the result data.

[0127] The neural network is composed of an input layer including multiple neurons, an intermediate layer including multiple neurons, and an output layer including multiple neurons. The intermediate layer can also be one or more layers, also called a hidden layer. As another example, there may be no intermediate layer. Thus, for example, the amount of processing in the communication system 100 can be reduced.

[0128] Figure 8 This is a schematic diagram showing an example of a neural network used in training the encoding model and the decoding model.

[0129] exist Figure 8 In the example shown, the neural network consists of an encoding model M1 and a decoding model M2.

[0130] like Figure 8As shown, when multiple input values, such as SL-CSI reports in different frequency bands, are input to input layers X1 to X4, the values ​​are multiplied by values ​​w11 to w43 as weights W1 and then input to intermediate layers Y1 to Y3, and the results are further multiplied by values ​​v11 to v32 as weights V1 and then input to intermediate layers Z1 to Z2. The results change according to the values ​​of weights W1 and V1.

[0131] Furthermore, the values ​​of the intermediate layers Z1 to Z2 are multiplied by the values ​​v'11 to v'23 as weights V'1 and input to the intermediate layers Y'1 to Y'3, and the results are further multiplied by the values ​​w'11 to w'34 as weights W'1 and input to the output layers X'1 to X'4. The output results change according to the values ​​of weights W1, V1, V'1, and W'1.

[0132] For the values ​​input to the intermediate layers Y1 to Y3, the intermediate layers Z1 to Z2, the intermediate layers Y'1 to Y'3, and the output layers X'1 to X'4, operations based on predetermined functions are performed. The operation results in the intermediate layers are multiplied by weight V1, weight V'1, or weight W'1. For the predetermined functions, for example, the Sigmoid function is sometimes used, the hyperbolic tangent function is sometimes used, and other functions are sometimes used.

[0133] In the first embodiment, the neural network trains the encoding model and the decoding model through so-called supervised training according to the training data generated based on the SL-CSI report transmitted from the first UE 110A.

[0134] Here, the values ​​of the output layers X'1 to X'4 train the encoding model and the decoding model, for example, weight W1, weight V1, weight V'1, and weight W'1, in the same manner as the input layers X1 to X4.

[0135] In addition, the SL-CSI report can be used as input, and the data obtained by encoding the SL-CSI report can be used as the label of the result, thereby serving as training data for the encoding model. In addition, the data obtained by encoding the SL-CSI report can be used as input, and the SL-CSI report can be used as the label of the result, thereby serving as training data for the decoding model.

[0136] Return to Figure 7, the control unit 122 of the second UE 110B repeats the processing of steps S17 and S18, and when the number of transmissions of the SL-CSI report reaches the preset number of transmissions N, or when it is determined that the data required for training is obtained, it instructs the first UE 110A to stop sending the training data via the protocol processing unit 121 and the terminal side communication unit 111 (S19). Thus, the control unit 122 ends the training.

[0137] Then, the control unit 122 of the second UE 110B generates an updated encoding model and decoding model based on the trained model obtained through training, and applies the generated decoding model to its own inference device (S20). Here, the control unit 122 of the second UE 110B functions as an inference device.

[0138] In addition, the trained decoding model may be a model obtained by changing the numerical values ​​of the elements or the combination of the constituent information itself according to the information (1) to (11) in the initial model.

[0139] Next, the control unit 122 of the second UE 110B transmits the coding model of the training result generated in the second UE 110B to the first UE 110A using SL data communication via the protocol processing unit 121 and the terminal side communication unit 111 (S21). The control unit 122 of the first UE 110A receives such a coding model via the terminal side communication unit 111 and the protocol processing unit 121. Then, the control unit 122 of the first UE 110A stores the coding model in the storage unit 123.

[0140] Furthermore, the control unit 122 of the second UE 110B instructs the first UE 110A to apply the coding model transmitted in step S21 via the protocol processing unit 121 and the terminal side communication unit 111 ( S22 ).

[0141] The control unit 122 of the first UE 110A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the trained coding model stored in the storage unit 123 (S23).

[0142] Then, the control unit 122 of the first UE 110A encodes the SL-CSI report using the trained coding model, and transmits the encoded SL-CSI report to the second UE 110B via the protocol processing unit 121 and the terminal side communication unit 111 ( S24 ).

[0143] Furthermore, the control unit 122 of the second UE 110B receives and decodes the encoded SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121. Then, the control unit 122 analyzes the obtained SL-CSI signal and applies it to actual PHY control (S25).

[0144] In addition, the control unit 122 of the second UE 110B may analyze the encoded SL-CSI report received in step S24, retrain and continuously update the encoding model and the decoding model, and apply them as appropriate.

[0145] Here, when implementing retraining, the control unit 122 of the second UE 110B can use the current coding model and decoding model, in other words, the trained coding model and decoding model for retraining, or use the initial model for retraining, or train from scratch and generate the coding model and decoding model.

[0146] In addition, when the result of the analysis is that the error increases due to the use of the trained coding model and decoding model, the control unit 122 of the second UE 110B may also extend the signal indicating the application of the coding model in step S13, send an instruction such as not to use the coding model to the first UE 110A, and stop the processing using the coding model and decoding model. As an indicator of the error, for example, a bit error rate, a block error rate, a throughput, a delay time, or the number of retransmissions may also be used. In addition, the base station may also instruct the second UE 110B which indicator and threshold to use.

[0147] As described above, in Embodiment 1, the communication system 100 as a wireless communication system includes a first UE 110A and a second UE 110B that communicates with the first UE 110A through SL communication. The first UE 110A sends coded channel state information to the second UE 110B, the coded channel state information being channel state information encoded using a coding model trained for channel state information, the channel state information being information for reporting the channel state of SL communication. The second UE 110B decodes the coded channel state information using a decoding model trained for channel state information. Here, the coding model and the decoding model are trained by the second UE 110B using the channel state information for training sent by the first UE 110A. Thus, in Embodiment 1, the SL-CSI report as channel state information can be sent by reducing the amount of data through the coding model, and such data can be decoded using the decoding model.

[0148] In the above-described first embodiment, the control unit 122 of the second UE 110B functions as a training device, but the first embodiment is not limited to such an example. For example, the encoder unit 113 or the decoder unit 118 of the second UE 110B may also function as a training device. In the first embodiment, the part that functions as a training device is referred to as a training unit, and the storage unit 123 that stores the trained model is referred to as a trained model storage unit.

[0149] In addition, in Embodiment 1, the control unit 122 of the first UE 110A performs encoding using the encoding model that is a trained model, but Embodiment 1 is not limited to such an example. For example, the encoder unit 113 of the first UE 110A may also perform encoding using the encoding model. In Embodiment 1, as an inference device, the part that performs encoding using the encoding model is referred to as a model encoding unit.

[0150] In addition, in Embodiment 1, the control unit 122 of the second UE 110B performs decoding using the decoding model that is a trained model, but Embodiment 1 is not limited to such an example. For example, the decoder unit 118 of the second UE 110B may also perform decoding using the decoding model. In Embodiment 1, as an inference device, the part that performs decoding using the decoding model is also referred to as a model decoding unit.

[0151] Implementation method 2.

[0152] In Embodiment 1, the encoding model and the decoding model are trained by the second UE 110B, but it is assumed that the training function does not exist or cannot be implemented in the second UE 110B because the computational processing load of the training is large and it cannot be executed in the UE 110. In Embodiment 2, an example is described in which the training is not performed by the UE but by the base station.

[0153] like Figure 1 As shown, the communication system 200 according to the second embodiment includes a UE 210 , a base station 230 , and a host device 160 .

[0154] The host device 160 of the communication system 200 according to the second embodiment is the same as the host device 160 of the communication system 100 according to the first embodiment.

[0155] like Figure 2 As shown, the UE 210 in Embodiment 2 includes a terminal-side communication unit 111 , an AP unit 120 , a protocol processing unit 121 , a control unit 222 , and a storage unit 123 .

[0156] The terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 210 in implementation mode 2 are the same as the terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 110 in implementation mode 1.

[0157] The control unit 222 controls a series of processes performed by the UE 210 .

[0158] Furthermore, the control unit 222 performs processing related to training in the base station 230. This processing will be described later.

[0159] like Figure 4 As shown, the base station 230 in Embodiment 2 includes a station-side communication unit 131 , a communication processing unit 140 , a protocol processing unit 144 , a control unit 245 , and a storage unit 146 .

[0160] The station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 230 in implementation mode 2 are the same as the station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 130 in implementation mode 1.

[0161] The control unit 245 controls a series of processes in the base station 230 .

[0162] Furthermore, the control unit 245 trains the encoding model and the decoding model. This process will be described later.

[0163] In Embodiment 2, base station 230 performs training of the encoding model and the decoding model. This configuration will be described below.

[0164] Fig. 9 A configuration example is shown in Embodiment 2 in which the base station 230 trains the encoding model and the decoding model and the UE 210 performs inference using the trained encoding model and the decoding model.

[0165] In the second embodiment, the base station 230 performs training of the encoding model and the decoding model, and the first UE 210A and the second UE 210B perform communication using the trained models.

[0166] Here, the first UE 210A is a data receiving terminal device that transmits SL-CSI, and the second UE 210B is a data transmitting terminal device that receives SL-CSI. The second UE 210B is located within the range of the base station 230 and is in a state of being able to communicate.

[0167] Fig.10 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 2.

[0168] When the second UE 210B enters the range of the base station 230 and performs handover, or when the second UE 210B starts up within the range of the base station 230, the control unit 222 of the second UE 210B requests registration and permission for SL communication from the base station 230 via the protocol processing unit 121 and the terminal side communication unit 111. At this time, the control unit 222 of the second UE 210B requests the base station 230 to implement training related to encoding and decoding of SL-SL-CSI (S30).

[0169] The control unit 245 of the base station 230 receives the request via the station side communication unit 131, the communication processing unit 140 and the protocol processing unit 144. In response to the request, if the base station 230 has a training function, the control unit 245 responds to the second UE 210B via the protocol processing unit 144, the communication processing unit 140 and the station side communication unit 131 to indicate permission (S31).

[0170] The control unit 222 of the first UE 210A and the control unit 222 of the second UE 210B establish SL communication via the PC5 interface via the respective protocol processing units 121 and the terminal side communication unit 111 (S32). As a result, the first UE 210A and the second UE 210B are able to communicate.

[0171] Next, in preparation for starting data transmission from the second UE 210B to the first UE 210A, the control unit 222 of the second UE 210B transmits a request for acquiring an initial model for SL-CSI encoding and decoding to the base station 230 via the protocol processing unit 121 and the terminal side communication unit 111 (S33).

[0172] The control unit 245 of the base station 230 receives the request via the station side communication unit 131, the communication processing unit 140 and the protocol processing unit 144. If the base station 230 can cope with it, it responds to the second UE 210B indicating the situation via the protocol processing unit 144, the communication processing unit 140 and the station side communication unit 131 (S34).

[0173] The control unit 222 of the second UE 210B receives the response via the terminal side communication unit 111 and the protocol processing unit 121, and starts sending the SL-CSI-RS without encoding. The control unit 222 of the first UE 210A parses the SL-CSI information received via the terminal side communication unit 111 and the protocol processing unit 121, and sends the SL-CSI report to the second UE 210B via the protocol processing unit 121 and the terminal side communication unit 111 (S35).

[0174] The control unit 222 of the second UE 210B receives the SL-CSI report via the station side communication unit 131 and the protocol processing unit 144, and directly transfers the SL-CSI report to the base station 230 via the protocol processing unit 144 and the station side communication unit 131 (S36).

[0175] The control unit 245 of the base station 230 receives the SL-CSI report via the station-side communication unit 131, the communication processing unit 140 and the protocol processing unit 144, and estimates the radio wave propagation state from each transmitting antenna of the second UE 210B to each receiving antenna of the first UE 210A based on the SL-CSI report, and determines the optimal model for the transmission path from the second UE 210B to the first UE 210A from the pre-prepared initial coding model and initial decoding model, or from the coding model and decoding model applied in other terminal devices as the initial model (S37).

[0176] For example, in Embodiment 2, a plurality of pairs of encoding models and decoding models are stored in the storage unit 146, and the control unit 245 determines an initial encoding model and an initial decoding model to be used from these pairs.

[0177] Specifically, the storage unit 146 stores a plurality of pairs of encoding models and decoding models that differ in the number of intermediate nodes, the number of levels of intermediate nodes, the number of branches connected to the intermediate nodes, and the number of final nodes.

[0178] Then, the control unit 245 can grasp the strength of the received power or the antenna separation capability of MIMO based on the obtained SL-CSI report, and therefore determine the initial coding model and initial decoding model to be used. The higher the strength or the antenna separation capability, the smaller at least any one of the number of intermediate nodes, the number of levels of intermediate nodes, the number of branches connected to the intermediate nodes, and the number of final nodes.

[0179] Then, the control unit 245 of the base station 230 transmits the determined initial coding model and initial decoding model to the second UE 210B via the protocol processing unit 144, the communication processing unit 140, and the station-side communication unit 131 (S38). The control unit 222 of the second UE 210B receives the initial coding model and initial decoding model via the terminal-side communication unit 111 and the protocol processing unit 121, and stores the initial coding model and initial decoding model in the storage unit 123.

[0180] Here, the encoding model data and the decoding model data are respectively constituted by, for example, the above-mentioned information (1) to (11).

[0181] Next, the control unit 222 of the second UE 210B forwards the initial coding model received from the base station 230 in step S38 to the first UE 210A via the protocol processing unit 121 and the terminal side communication unit 111 (S39). The control unit 222 of the first UE 210A receives the initial coding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores the initial coding model in the storage unit 123.

[0182] Next, the control unit 222 of the second UE 210B instructs the first UE 210A to apply the received coding model via the protocol processing unit 121 and the terminal side communication unit 111 (S40).

[0183] The control unit 222 of the first UE 210A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the coding model stored in the storage unit 123 (S41).

[0184] Furthermore, in the second UE 210B, the control unit 222 also applies the decoding model stored in the storage unit 123 (S42).

[0185] Next, in order to perform training, the control unit 222 of the second UE 210B sends the SL-CSI-RS via the protocol processing unit 121 and the terminal side communication unit 111, and requests the first UE 210A to send the SL-CSI report for training (S43). At this time, the control unit 222 may also specify the number of times N of sending the SL-CSI report for training in the request.

[0186] The control unit 222 of the first UE 210A receives the request via the terminal side communication unit 111 and the protocol processing unit 121, and transmits a training SL-CSI report via the protocol processing unit 121 and the terminal side communication unit 111 in response to the request (S44).

[0187] The control unit 222 of the second UE 210B receives the SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121, and transfers the SL-CSI report to the base station 230 via the protocol processing unit 121 and the terminal side communication unit 111 (S45).

[0188] Next, the control unit 245 of the base station 230 receives the transferred SL-CSI report via the station-side communication unit 131 , the communication processing unit 140 , and the protocol processing unit 144 , and performs training of the encoding model and the decoding model ( S46 ).

[0189] The control unit 245 of the base station 230 repeats the processing of steps S44, S45, and S46, and when the number of transmissions of the SL-CSI report reaches the preset number of transmissions N, or when it is determined that the data required for training is obtained, it instructs the second UE 210B to stop transmitting the training data via the protocol processing unit 144, the communication processing unit 140, and the station-side communication unit 131 (S47). Thus, the control unit 245 ends the training.

[0190] The control unit 222 of the second UE 210B receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and transfers the instruction to the first UE 210A via the terminal side communication unit 111 and the protocol processing unit 121 (S48).

[0191] Then, the control unit 245 of the base station 230 generates an updated encoding model and decoding model based on the trained model obtained through training (S49). In addition, the trained decoding model can also be a model obtained by changing the numerical value of the elements or the combination of the constituent information itself from the above-mentioned information (1) to (11) in the initial model.

[0192] Furthermore, the control unit 245 of the base station 230 transmits the generated coding model and decoding model to the second UE 210B via the protocol processing unit 144, the communication processing unit 140, and the station side communication unit 131 (S50).

[0193] The control unit 222 of the second UE 210B receives the coding model and decoding model from the base station 230 via the terminal side communication unit 111 and the protocol processing unit 121, stores the decoding model in the storage unit 123, and applies the decoding model (S51). Here, the control unit 222 of the second UE 210B functions as an inference device.

[0194] Next, the control unit 222 of the second UE 210B transmits the coding model of the training result received from the base station 230 to the first UE 210A via the protocol processing unit 121 and the terminal side communication unit 111 using SL data communication (S52). The control unit 222 of the first UE 210A receives the coding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores the coding model in the storage unit 123.

[0195] Furthermore, the control unit 222 of the second UE 210B instructs the first UE 210A to apply the coding model transmitted in step S52 via the protocol processing unit 121 and the terminal side communication unit 111 ( S53 ).

[0196] The control unit 222 of the first UE 210A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the trained coding model stored in the storage unit 123 (S54).

[0197] Then, the control unit 222 of the first UE 210A encodes the SL-CSI report using the trained coding model, and starts transmitting the SL-CSI report to the second UE 210B (S55).

[0198] In addition, the control unit 222 of the second UE 210B receives the encoded SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121, forwards the encoded SL-CSI report to the base station 230 via the protocol processing unit 121 and the terminal side communication unit 111 (S56), and decodes the encoded SL-CSI report, analyzes the obtained SL-CSI signal and uses it for actual PHY control (S57).

[0199] In addition, the control unit 245 of the base station 230 may also analyze the encoded SL-CSI report received in step S56, perform retraining, continuously update the encoding model and the decoding model, and apply them in a timely manner.

[0200] Here, the determination and request for retraining may be made by the second UE 210B or by the base station 230. Alternatively, the second UE 210B or the base station 230 may determine the retraining cycle in advance.

[0201] In addition, when retraining is performed, the current encoding model and decoding model may be used for retraining, the initial encoding model and initial decoding model may be used for retraining, or the encoding model and decoding model may be trained from scratch. The choice of updating the model or resetting the model may be determined by the second UE 210B or by the base station 230.

[0202] In addition, when the result of the analysis shows that the error increases due to the use of the coding model and the decoding model, the control unit 245 of the base station 230 may extend the signal indicating the application of the coding model in step S40, and the base station 230 may send an instruction not to use the coding model to the second UE 210B, and the second UE 210B may further forward the same instruction to the first UE 210A, thereby stopping the use of the coding model and the decoding model. As an indicator of the error, for example, a bit error rate, a block error rate, a throughput, a delay time, or the number of retransmissions may be used.

[0203] As described above, in Embodiment 2, the communication system 200 includes the base station 230 that communicates with the second UE 210B through SL communication, and the base station 230 trains the coding model and the decoding model using the channel state information for training sent by the first UE 210A. In other words, in Embodiment 2, the coding model and the decoding model can be trained in the base station 230 instead of the second UE 210B, and are distributed to the second UE 210B and the first UE 210A. As a result, the amount of data in the SL-CSI report can be reduced.

[0204] Furthermore, in Embodiment 2, the second UE 210B receives the channel state information for training transmitted by the first UE 210A, and transfers the channel state information for training to the base station 230 .

[0205] Implementation method 3.

[0206] In Embodiment 2, in order to train the SL-CSI report received from the first UE 210A, the second UE 210B needs to forward it to the base station 230 each time. For example, assuming that the first UE 210A and the second UE 210B are both within the communication range of the base station 230 in a flat, sparsely populated area or a place where the state of the transmission path does not change much, such as at sea, the base station 230 directly receives the SL-CSI report sent from the first UE 210A and performs training, thereby reducing the increase in processing volume or processing time caused by forwarding.

[0207] like Figure 1 As shown, the communication system 300 according to the third embodiment includes a UE 310 , a base station 330 , and a host device 160 .

[0208] The host device 160 of the communication system 300 according to the third embodiment is the same as the host device 160 of the communication system 100 according to the first embodiment.

[0209] like Figure 2 As shown, the UE 310 in Embodiment 3 includes a terminal side communication unit 111, an AP unit 120, a protocol processing unit 121, a control unit 322, and a storage unit 123.

[0210] The terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 310 in Implementation Example 3 are the same as the terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 110 in Implementation Example 1.

[0211] The control unit 322 controls a series of processes performed in the UE 310 .

[0212] Furthermore, the control unit 322 performs processing related to training in the base station 330. This processing will be described later.

[0213] like Figure 4 As shown, the base station 330 in Embodiment 3 includes a station-side communication unit 131 , a communication processing unit 140 , a protocol processing unit 144 , a control unit 345 , and a storage unit 146 .

[0214] The station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 330 in implementation mode 3 are the same as the station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 130 in implementation mode 1.

[0215] The control unit 345 controls a series of processes in the base station 130 .

[0216] Furthermore, the control unit 345 trains the encoding model and the decoding model. This process will be described later.

[0217] In Embodiment 3, base station 330 performs training of the encoding model and the decoding model. This structure will be described below.

[0218] Fig.11 A configuration example is shown in Embodiment 3 in which the base station 330 trains the encoding model and the decoding model and the UE 310 performs inference using the trained encoding model and the decoding model.

[0219] In Embodiment 3, the base station 330 performs training of the encoding model and the decoding model, and the first UE 310A and the second UE 310B perform communication using the trained models.

[0220] Here, the first UE 310A is a data receiving terminal device that transmits SL-CSI, and the second UE 310B is a data transmitting terminal device that receives SL-CSI. In Embodiment 3, both the first UE 310A and the second UE 310B are located within the range of the base station 330 and are in a state where they can communicate.

[0221] Fig.12 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 3.

[0222] When the second UE 310B enters the range of the base station 330 and performs handover, or when the second UE 310B starts up within the range of the base station 330, the control unit 322 of the second UE 310B requests registration and permission for SL communication from the base station 330 via the protocol processing unit 121 and the terminal side communication unit 111. At this time, the control unit 322 of the second UE 310B requests the base station 330 to perform training related to encoding and decoding of SL-CSI (S60).

[0223] The control unit 345 of the base station 330 receives the request via the station-side communication unit 131, the communication processing unit 140, and the protocol processing unit 144. In response to the request, if the base station 330 has a training function, the control unit 345 responds to the second UE 310B indicating permission via the protocol processing unit 144, the communication processing unit 140, and the station-side communication unit 131 (S61). At this time, the control unit 345 causes the response to include information indicating that the base station 330 has a function of directly receiving the SL-CSI and performing training.

[0224] The control unit 322 of the first UE 310A and the control unit 322 of the second UE 310B establish SL communication via the PC5 interface via their respective protocol processing units 121 and terminal-side communication units 111 (S62). As a result, the first UE 310A and the second UE 310B are able to communicate.

[0225] Next, as a preparation for starting data transmission from the second UE 310B to the first UE 310A, the control unit 322 of the second UE 310B transmits a request for obtaining an initial model of SL-CSI encoding and decoding to the base station 330 via the protocol processing unit 121 and the terminal side communication unit 111 (S63). Here, the control unit 322 causes the request to include information indicating that the first UE 310A transmits an SL-CSI report.

[0226] The control unit 345 of the base station 330 receives the request via the station-side communication unit 131 and the protocol processing unit 144. If the base station 330 can cope with it, it responds to the second UE 310B via the protocol processing unit 144, the communication processing unit 140 and the station-side communication unit 131 to indicate the situation (S64). Here, the base station 330 directly receives the SL-CSI from the first UE 310A when the first UE 310A is within its range and can communicate. In this case, the control unit 345 adds information indicating a direct reception mode to the response, and the direct reception mode is a mode for directly receiving the SL-CSI report from the first UE 310A.

[0227] The control unit 322 of the second UE 310B receives the response via the terminal side communication unit 111 and the protocol processing unit 121, and starts sending the SL-CSI-RS without encoding. The control unit 322 of the first UE 310A parses the received SL-CSI information via the terminal side communication unit 111 and the protocol processing unit 121, and sends the SL-CSI report (S65) via the protocol processing unit 121 and the terminal side communication unit 111.

[0228] The control unit 322 of the second UE 310B receives the SL-CSI report transmitted from the first UE 310A via the station side communication unit 131 and the protocol processing unit 144 , but does not forward the SL-CSI report to the base station 330 when the direct reception mode is instructed.

[0229] On the other hand, the control unit 345 of the base station 330 receives the SL-CSI report from the first UE 310A via the station-side communication unit 131, the communication processing unit 140 and the protocol processing unit 144, and estimates the radio wave propagation state from each transmitting antenna of the second UE 210B to each receiving antenna of the first UE 210A based on the SL-CSI report, and determines the optimal model for the transmission path from the second UE 310B to the first UE 310A from the pre-prepared initial coding model and initial decoding model, or from the coding model and decoding model applied in other terminal devices as the initial model (S66).

[0230] Then, the control unit 345 of the base station 330 transmits the determined initial coding model and initial decoding model to the second UE 310B via the protocol processing unit 144, the communication processing unit 140, and the station-side communication unit 131 (S67). The control unit 322 of the second UE 310B receives the initial coding model and initial decoding model via the terminal-side communication unit 111 and the protocol processing unit 121, and stores the initial coding model and initial decoding model in the storage unit 123.

[0231] Here, the encoding model data and the decoding model data are respectively constituted by, for example, the above-mentioned information (1) to (11).

[0232] Next, the control unit 322 of the second UE 310B forwards the initial coding model received from the base station 330 in step S67 to the first UE 310A via the protocol processing unit 121 and the terminal side communication unit 111 (S68). The control unit 322 of the first UE 310A receives the initial coding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores it in the storage unit 123.

[0233] Next, the control unit 322 of the second UE 310B instructs the first UE 310A to apply the received coding model via the protocol processing unit 121 and the terminal side communication unit 111 (S69).

[0234] The control unit 322 of the first UE 310A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the coding model stored in the storage unit 123 (S70).

[0235] Furthermore, in the second UE 310B, the control unit 322 also applies the decoding model stored in the storage unit 123 ( S71 ).

[0236] Next, in order to perform training, the control unit 322 of the second UE 310B sends the SL-CSI-RS via the protocol processing unit 121 and the terminal side communication unit 111, and requests the first UE 310A to send the SL-CSI report for training (S72). At this time, the control unit 322 can also specify the number of times N of sending the SL-CSI report for training in the request.

[0237] The control unit 322 of the first UE 310A receives the request via the terminal side communication unit 111 and the protocol processing unit 121, and transmits a SL-CSI report for training via the protocol processing unit 121 and the terminal side communication unit 111 in response to the request (S73).

[0238] The control unit 322 of the second UE 310B receives the SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121 , but does not transfer the SL-CSI report to the base station 330 in the case of the direct reception mode.

[0239] Meanwhile, the control unit 345 of the base station 330 receives the SL-CSI report transmitted from the first UE 310A via the station side communication unit 131, the communication processing unit 140, and the protocol processing unit 144, and trains the encoding model and the decoding model (S74).

[0240] The control unit 345 of the base station 330 repeats the processing of steps S73 and S74, and when the number of transmissions of the SL-CSI report reaches the preset number of transmissions N, or when it is determined that the data required for training is obtained, it instructs the second UE 310B to stop transmitting the training data via the protocol processing unit 144, the communication processing unit 140, and the station-side communication unit 131 (S75). Thus, the control unit 345 ends the training.

[0241] The control unit 322 of the second UE 310B receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and transfers the instruction to the first UE 310A via the terminal side communication unit 111 and the protocol processing unit 121 (S76).

[0242] Then, the control unit 345 of the base station 330 generates an updated encoding model and decoding model based on the trained model obtained through training (S77). In addition, the trained decoding model may also be a model obtained by changing the numerical value of the elements or the combination of the constituent information from the information (1) to (11) in the initial model.

[0243] Furthermore, the control unit 345 of the base station 330 transmits the generated coding model and decoding model to the second UE 310B via the protocol processing unit 144, the communication processing unit 140, and the station side communication unit 131 (S78).

[0244] The control unit 322 of the second UE 310B receives the coding model and decoding model from the base station 330 via the terminal side communication unit 111 and the protocol processing unit 121, stores the decoding model in the storage unit 123, and applies the decoding model (S79). Here, the control unit 322 of the second UE 310B functions as an inference device.

[0245] Next, the control unit 322 of the second UE 310B transmits the coding model of the training result received from the base station 330 to the first UE 310A via the protocol processing unit 121 and the terminal side communication unit 111 using SL data communication (S80). The control unit 322 of the first UE 310A receives the coding model via the terminal side communication unit 111 and the protocol processing unit 121. Then, the control unit 322 of the first UE 310A stores the coding model in the storage unit 123.

[0246] Furthermore, the control unit 322 of the second UE 310B transmits a signal to the first UE 310A via the protocol processing unit 121 and the terminal side communication unit 111, instructing the first UE 310A to apply the coding model transmitted in step S80 (S81).

[0247] In the first UE 310A that has received the instruction, the control unit 322 receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the trained coding model stored in the storage unit 123 (S82).

[0248] Then, the control unit 322 of the first UE 310A encodes the SL-CSI report using the trained coding model and starts transmitting the SL-CSI report ( S83 ).

[0249] In addition, in the second UE 310B, the control unit 322 also receives the received data via the terminal side communication unit 111 and the protocol processing unit 121, decodes it, analyzes the obtained SL-CSI signal and uses it for actual PHY control (S84). In addition, in the direct reception mode, the encoded SL-CSI report is not forwarded.

[0250] In addition, the control unit 345 of the base station 330 may analyze the encoded SL-CSI report received in step S83, retrain and continuously update the encoding model and the decoding model, and apply them in a timely manner.

[0251] Here, the determination and request for retraining may be made by the second UE 310B or by the base station 330. Alternatively, the second UE 310B or the base station 330 may determine the retraining cycle in advance.

[0252] In addition, when retraining is performed, the current encoding model and decoding model may be used for retraining, the initial encoding model and initial decoding model may be used for retraining, or the encoding model and decoding model may be trained from scratch. The choice of updating the model or resetting the model may be determined by the second UE 310B or by the base station 330.

[0253] In addition, when the result of the analysis shows that the error increases due to the use of the coding model and the decoding model, the control unit 345 of the base station 330 may extend the signal indicating the application of the coding model in step S69, and the base station 330 may send an instruction not to use the coding model to the second UE 310B, and the second UE 310B may forward the same instruction to the first UE 310A, thereby stopping the use of the coding model and the decoding model. As an indicator of the error, for example, a bit error rate, a block error rate, a throughput, a delay time, or the number of retransmissions may be used.

[0254] As described above, in Embodiment 3, the base station 330 can receive the SL-CSI report from the first UE 310A instead of the second UE 310B, train the encoding model and the decoding model, and distribute them to the second UE 310B and the first UE 310A. This can reduce the amount of data in the SL-CSI report.

[0255] Implementation method 4.

[0256] In SL communication, communication between UE and NW (Network) via relay is proposed instead of direct communication between UEs. Hereinafter, the relay between UE and NW is sometimes referred to as UE-to-NW relay or UE-NW relay. In addition, the UE that implements the relay between UE and NW is sometimes referred to as U2N relay.

[0257] Embodiment 4 shows an example in which a UE having a U2N relay function performs training on behalf of other UEs.

[0258] like Figure 1 As shown, the communication system 400 according to the fourth embodiment includes a UE 410 , a base station 430 , and a host device 160 .

[0259] The host device 160 of the communication system 400 according to the fourth embodiment is the same as the host device 160 of the communication system 100 according to the first embodiment.

[0260] like Figure 2 As shown, the UE 410 in Embodiment 4 includes a terminal side communication unit 111, an AP unit 120, a protocol processing unit 121, a control unit 422, and a storage unit 123.

[0261] The terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 410 in implementation mode 4 are the same as the terminal side communication unit 111, AP unit 120, protocol processing unit 121 and storage unit 123 of UE 110 in implementation mode 1.

[0262] The control unit 422 controls a series of processes performed in the UE 410 .

[0263] Furthermore, the control unit 422 performs a process related to training. This process will be described later.

[0264] like Figure 4 As shown, the base station 430 in Embodiment 4 includes a station-side communication unit 131 , a communication processing unit 140 , a protocol processing unit 144 , a control unit 445 , and a storage unit 146 .

[0265] The station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 430 in implementation mode 4 are the same as the station-side communication unit 131, communication processing unit 140, protocol processing unit 144 and storage unit 146 of the base station 130 in implementation mode 1.

[0266] The control unit 445 controls a series of processes in the base station 430 .

[0267] Furthermore, the control unit 445 causes the U2N repeater to train the encoding model and the decoding model. This process will be described later.

[0268] In Embodiment 4, the U2N repeater performs training of the encoding model and the decoding model. This structure is described below.

[0269] Fig.13 A configuration example is shown in which the U2N relay performs training of the encoding model and the decoding model in Embodiment 4 and the UE 410 performs inference using the trained encoding model and the decoding model.

[0270] In Embodiment 4, the U2N relay trains the encoding model and the decoding model, and the first UE 410A and the second UE 410B communicate using the trained models.

[0271] Here, the first UE 410A is a MIMO receiving terminal device that sends SL-CSI, the second UE 410B is a MIMO transmitting terminal device that receives SL-CSI, and the third UE 410C is a U2N repeater. In Embodiment 4, the first UE 410A and the second UE 410B are outside the range of the base station 430, but are in a state where they can communicate with the base station 430 via the third UE 410C.

[0272] Fig.14 It is a timing diagram showing the steps of training the encoding model and the decoding model and applying the trained models in Implementation 4.

[0273] When the third UE 410C enters the range of the base station 430 and performs a handover, or when the third UE 410C is activated within the range of the base station 430, the control unit 422 of the third UE 410C requests registration and permission for SL communication and U2N repeater utilization from the base station 430 via the protocol processing unit 121 and the terminal side communication unit 111. At this time, the control unit 422 of the third UE 410C requests permission from the base station 430 to perform training related to encoding and decoding of SL-CSI on behalf of other UEs 410 (S90).

[0274] The control unit 445 of the base station 430 receives the request via the station side communication unit 131, the communication processing unit 140 and the protocol processing unit 144, and responds to the request via the protocol processing unit 144, the communication processing unit 140 and the station side communication unit 131, indicating that the third UE 410C is allowed to perform the relay function (S91).

[0275] Next, the control unit 422 of the first UE 410A, the control unit 422 of the second UE 410B, and the control unit 422 of the third UE 410C establish SL communication (S92 to S94) through the PC5 interface via their respective protocol processing units 121 and terminal-side communication units 111. As a result, the first UE 410A, the second UE 410B, and the third UE 410C are able to communicate with each other.

[0276] The control unit 422 of the second UE 410B inquires of the third UE 410C, which is a U2N repeater, via the protocol processing unit 121 and the terminal side communication unit 111 whether the third UE 410C has a function to replace the training function (S95).

[0277] The control unit 422 of the third UE 410C receives the inquiry via the terminal side communication unit 111 and the protocol processing unit 121, and responds to the second UE 410B via the protocol processing unit 121 and the terminal side communication unit 111, indicating that the second UE 410B has such a function and is executing the training function instead (S96).

[0278] Next, in preparation for starting data transmission from the second UE 410B to the first UE 410A, the control unit 422 of the second UE 410B transmits a request for obtaining an initial model for SL-CSI encoding and decoding to the third UE 410C via the protocol processing unit 121 and the terminal side communication unit 111 (S97).

[0279] The control unit 422 of the third UE 410C receives the request via the terminal side communication unit 111 and the protocol processing unit 121, and if the third UE 410C can cope with it, it responds to the second UE 410B via the protocol processing unit 121 and the terminal side communication unit 111 to that effect (S98).

[0280] The control unit 422 of the second UE 410B receives the response via the terminal side communication unit 111 and the protocol processing unit 121, and starts to send the SL-CSI-RS without encoding. The control unit 422 of the first UE 410A analyzes the SL-CSI information received via the terminal side communication unit 111 and the protocol processing unit 121, and sends the SL-CSI report via the protocol processing unit 121 and the terminal side communication unit 111 (S99). Here, an example is shown in which the first UE 410A directly sends the SL-CSI report to the third UE 410C, but the second UE 410B may also forward the report as in Embodiment 2.

[0281] The control unit 422 of the third UE 410C receives the SL-CSI report from the first UE 410A via the terminal side communication unit 111 and the protocol processing unit 121, estimates the radio wave propagation state from each transmission antenna of the second UE 410B to each reception antenna of the first UE 410A based on the SL-CSI report, and determines the model that is optimal for the transmission path from the second UE 410B to the first UE 410A from among the initial coding model and initial decoding model prepared in advance, or from the coding model and decoding model applied in other terminal devices as the initial model (S100). In addition, the storage unit 123 of the third UE 410C in the fourth embodiment stores a plurality of pairs of initial coding models and initial decoding models.

[0282] Then, the control unit 422 of the third UE 410C sends the determined initial coding model and initial decoding model to the second UE 410B via the protocol processing unit 121 and the terminal side communication unit 111 (S101). The control unit 422 of the second UE 410B receives the initial coding model and initial decoding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores the initial coding model and initial decoding model in the storage unit 123.

[0283] Here, the encoding model data and the decoding model data are respectively constituted by, for example, the above-mentioned information (1) to (11).

[0284] Next, the control unit 422 of the second UE 410B forwards the initial coding model received from the third UE 410C in step S101 to the first UE 410A via the protocol processing unit 121 and the terminal side communication unit 111 (S102). The control unit 422 of the first UE 410A receives the initial coding model via the terminal side communication unit 111 and the protocol processing unit 121, and stores it in the storage unit 123.

[0285] Next, the control unit 422 of the second UE 410B instructs the first UE 410A to apply the received coding model via the protocol processing unit 121 and the terminal side communication unit 111 (S103).

[0286] The control unit 422 of the first UE 410A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the coding model stored in the storage unit 123 according to the instruction (S104).

[0287] Furthermore, in the second UE 410B, the control unit 422 also applies the decoding model stored in the storage unit 123 (S105).

[0288] Next, in order to perform training, the control unit 422 of the second UE 410B sends the SL-CSI-RS via the protocol processing unit 121 and the terminal side communication unit 111, and requests the first UE 410A to send the SL-CSI report for training (S106). At this time, the control unit 422 may also specify the number of times N of sending the SL-CSI report for training in the request.

[0289] The control unit 422 of the first UE 410A receives the request via the terminal side communication unit 111 and the protocol processing unit 121, and transmits a SL-CSI report for training via the protocol processing unit 121 and the terminal side communication unit 111 in response to the request (S107).

[0290] The control unit 422 of the third UE 410C receives the SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121, and performs training of the encoding model and the decoding model (S108).

[0291] The control unit 422 of the third UE 410C repeats the processing of steps S107 and S108, and when the number of transmissions of the SL-CSI report reaches the preset number of transmissions N, or when it is determined that the data required for training is obtained, it instructs the second UE 410B to stop transmitting the training data via the protocol processing unit 121 and the terminal side communication unit 111 (S109). Thus, the control unit 422 ends the training.

[0292] The control unit 322 of the second UE 410B receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and transfers the instruction to the first UE 410A via the terminal side communication unit 111 and the protocol processing unit 121 (S110).

[0293] Then, the control unit 422 of the third UE 410C generates an updated encoding model and decoding model based on the trained model obtained through training (S111). In addition, the trained decoding model can also be a model obtained by changing the numerical value of the elements or the combination of the constituent information from the information (1) to (11) in the initial model.

[0294] Furthermore, the control unit 422 of the third UE 410C transmits the generated encoding model and decoding model to the second UE 410B via the protocol processing unit 121 and the terminal side communication unit 111 (S112).

[0295] The control unit 422 of the second UE 410B receives the encoding model and decoding model from the third UE 410C via the terminal side communication unit 111 and the protocol processing unit 121, stores the decoding model in the storage unit 123, and applies the decoding model (S113). Here, the control unit 422 of the second UE 410B functions as an inference device.

[0296] Next, the control unit 422 of the second UE 410B transmits the coding model of the training result received from the third UE 410C to the first UE 410A via the protocol processing unit 121 and the terminal side communication unit 111 using SL data communication (S114). Such a coding model is received by the terminal side communication unit 111 of the first UE 410A and provided to the control unit 422 via the protocol processing unit 121. Then, the control unit 422 of the first UE 410A stores the coding model in the storage unit 123.

[0297] Furthermore, the control unit 422 of the second UE 410B transmits a signal to the first UE 410A via the protocol processing unit 121 and the terminal side communication unit 111, instructing the first UE 410A to apply the coding model transmitted in step S114 (S115).

[0298] The control unit 422 of the first UE 410A receives the instruction via the terminal side communication unit 111 and the protocol processing unit 121, and applies the trained coding model stored in the storage unit 123 according to the instruction (S116).

[0299] Then, the control unit 422 of the first UE 410A encodes the SL-CSI report using the trained coding model and starts transmitting the SL-CSI report ( S117 ).

[0300] In addition, the control unit 422 of the second UE 310B receives the encoded SL-CSI report via the terminal side communication unit 111 and the protocol processing unit 121, decodes the encoded SL-CSI report, analyzes the obtained CSI signal and uses it for actual PHY control (S118). In the direct reception mode, the encoded SL-CSI report is not forwarded.

[0301] In addition, the control unit 422 of the third UE 410C may analyze the encoded SL-CSI report received via the terminal side communication unit 111 and the protocol processing unit 121, retrain and continuously update the encoding model and the decoding model, and apply them in a timely manner.

[0302] Here, the determination and request for retraining may be made by the second UE 410B or the third UE 410C. Alternatively, the second UE 410B or the third UE 410C may predetermine the retraining cycle.

[0303] In addition, when retraining is implemented, the current encoding model and decoding model can be used for retraining, the initial encoding model and initial decoding model can be used for retraining, or the encoding model and decoding model can be trained from scratch. The choice of updating the model or resetting the model can be determined by the second UE 410B or the third UE 410C.

[0304] In addition, when the result of the analysis is that the error increases due to the use of the coding model and the decoding model, the control unit 422 of the third UE 410C may also extend the signal indicating the application of the coding model in step S103, and the third UE 410C may send an instruction not to use the coding model to the second UE 410B, and the second UE 410B may forward the same instruction to the first UE 410A, thereby stopping the use of the coding model and the decoding model. As an indicator of the error, for example, a bit error rate, a block error rate, a throughput, a delay time, or the number of retransmissions may also be used. In addition, the base station may also instruct the third UE 410C which indicator and threshold to use.

[0305] As described above, in Embodiment 4, the communication system 400 includes a third UE 410C that communicates with the first UE 410A and the second UE 410B through SL communication, and the third UE 410 trains the coding model and the decoding model using the channel state information for training sent by the first UE 410A. Thus, in Embodiment 4, the U2N repeater can perform training processing instead of other UEs 410, generate coding models and decoding models, and apply them to the UE 410 performing SL communication, thereby reducing the amount of data reported by the SL-CSI.

[0306] Here, in Embodiment 4, the second UE 410B may receive the channel state information for training sent by the first UE 410A, and the second UE 410B may forward the channel state information for training to the third UE 410C. Alternatively, the third UE 410C may directly receive the channel state information for training sent by the first UE 410A.

[0307] Description of Reference Numerals

[0308] 100, 200, 300, 400 communication system, 110, 210, 310, 410 UE, 111 terminal side communication unit, 120 AP unit, 121 protocol processing unit, 122, 222, 322, 422 control unit, 123 storage unit, 130, 230, 330, 430 base station, 131 station side communication unit, 140 communication processing unit, 144 protocol processing unit, 145 control unit, 146 storage unit, 160 upper device, 161 data network communication unit, 162 base station communication unit, 163 user plane communication unit, 164 session management unit, 165 control plane control unit, 169 control unit.

Claims

1. A wireless communication system comprising a first terminal device and a second terminal device communicating with the first terminal device through side link communication (SL communication), characterized in that: The first terminal device sends coded channel state information to the second terminal device, the coded channel state information being channel state information obtained by encoding using a coding model trained for channel state information, the channel state information being information for reporting a channel state of the SL communication, The second terminal device decodes the encoded channel state information using a decoding model trained for the channel state information.

2. The wireless communication system according to claim 1, characterized in that: The second terminal device trains the encoding model and the decoding model using the channel state information for training transmitted by the first terminal device.

3. The wireless communication system according to claim 1, characterized in that: The wireless communication system further includes a base station for communicating with the second terminal device through the SL communication. The base station trains the encoding model and the decoding model using the channel state information for training transmitted by the first terminal device.

4. The wireless communication system according to claim 3, characterized in that: The first terminal device sends the channel state information for training, The second terminal device receives the channel state information for training transmitted by the first terminal device and transfers it to the base station. The base station receives the channel state information for training transferred by the second terminal device.

5. The wireless communication system according to claim 3, characterized in that: The first terminal device transmits the channel state information for training, The base station receives the channel state information for training transmitted by the first terminal device.

6. The wireless communication system according to claim 1, characterized in that: The wireless communication system further includes a third terminal device, which communicates with the first terminal device and the second terminal device through the SL communication. The third terminal device trains the encoding model and the decoding model using the channel state information for training transmitted by the first terminal device.

7. The wireless communication system according to claim 6, characterized in that: The first terminal device sends the channel state information for training, The second terminal device receives the channel state information for training transmitted by the first terminal device and transfers it to the third terminal device. The third terminal device receives the channel state information for training transferred by the second terminal device.

8. The wireless communication system according to claim 6, characterized in that: The first terminal device transmits the channel state information for training, The third terminal device receives the channel state information for training transmitted by the first terminal device.

Images