First communication device, second communication device, third communication device, communication method, and program

Abstract

The purpose of the present invention is to provide a communication device capable of improving the service quality of XRM. A first communication device according to the present disclosure comprises: a management unit for managing a learning model subjected to machine learning using learning data consisting of operation content executed by a remote operator with respect to a remote device and a traffic pattern of communication data generated on the basis of the operation content; a communication unit for receiving first communication data from a remote operation terminal that generated the first communication data on the basis of first operation content indicating remote operation with respect to a first device to be operated remotely; and an estimation unit for estimating the first operation content from the first traffic pattern of the first communication data using the learning model. The communication unit transmits the first operation content to a second communication device that controls the operation of the first device to be operated remotely on the basis of the first operation content.

Description

First communication device, second communication device, third communication device, communication method and program

[0001] The present disclosure relates to a first communication device, a second communication device, a communication method, and a program.

[0002] As a technology for fusing the real world and the virtual world to create new experiences, XRM (Extended Reality and Media service) has been studied.

[0003] Non-Patent Document 1 discloses the contents of studies on QoS (Quality of Service) for providing XRM. Specifically, Non-Patent Document 1 discusses what kind of information from an application in a DN (Data Network) should be notified to a 5G (5th Generation) network when the size of a media frame changes due to a scene change, a user's drag operation, or the like.

[0004] 3GPP TR 23.700-70 V19.0.0 (2024-09)

[0005] In Non-Patent Document 1, it is premised that a server or the like that executes application processing is arranged in a DN located outside the 5G network. That is, after application data is analyzed in the DN, the analysis result is notified to the 5G network. When real-time analysis is required, an increase in the amount of data may cause communication delay. There is a problem in that the occurrence of delay when application data is analyzed in this manner degrades the service quality of XRM.

[0006] An object of the present disclosure is to provide a first communication device, a second communication device, a communication method, and a program capable of preventing a decrease in the service quality of XRM.

[0007] The first communication device according to this disclosure includes a management unit that manages a machine learning model that uses the actions performed by a remote operator on a remote device and the traffic patterns of communication data generated based on the actions as learning data; a communication unit that receives the first communication data from a remote operation terminal that generates the first communication data based on the first actions indicating remote operation on a first remotely controlled device; and an estimation unit that estimates the first actions from the first traffic patterns of the first communication data using the learning model, wherein the communication unit transmits the first actions to a second communication device that controls the operation of the first remotely controlled device based on the first actions.

[0008] The second communication device according to this disclosure includes a management unit that manages a machine learning model that uses as learning data the actions performed by a remote operator on a remote device and the actions that enable the device to perform actions according to the actions; a communication unit that receives the first actions from a first communication device that has estimated the first actions indicating remote operation on a first remotely controlled device; and an estimation unit that uses the learning model to estimate the first actions that enable the device to perform actions according to the first actions, wherein the communication unit transmits the first actions to the first remotely controlled device.

[0009] The communication method performed in the first communication device relating to this disclosure is performed in the second communication device, which receives the first operation content indicating remote operation to a first remotely controlled device from the first communication device which has estimated the first operation content, and uses a machine learning model that has been trained using the operation content performed by a remote operator on a remotely located device and operation parameters that realize the operation in accordance with the operation content on the device as training data to estimate the first operation parameters that realize the operation in accordance with the first operation content from the first operation content and transmits the first operation parameters to the first remotely controlled device.

[0010] A communication method performed in a second communication device according to this disclosure receives a first operation content indicating remote operation to a first remotely controlled device from a first communication device that has estimated the first operation content, and uses a machine learning model that has been trained using the operation content performed by a remote operator on a remotely located device and the operation parameters that realize the operation in accordance with the operation content on the device as training data to estimate a first operation parameter that realizes the operation in accordance with the first operation content from the first operation content, and transmits the first operation parameter to the first remotely controlled device.

[0011] The program relating to this disclosure receives first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device, and uses a machine learning model that has been trained using operation content performed by a remote operator on a remotely located device and the traffic pattern of the communication data generated based on said operation content as training data to estimate the first operation content from the first traffic pattern of the first communication data, and transmits the first operation content to a second communication device that controls the operation of the first remotely controlled device based on the first operation content.

[0012] The program relating to this disclosure receives a first operation from a first communication device that has estimated a first operation indicating remote operation to a first remotely controlled device, and causes a computer to perform the following actions: using a machine learning model that has been trained with the operation performed by a remote operator on a remote device and the operation parameters that realize the operation in accordance with the operation as training data, estimate a first operation parameter that realizes the operation according to the first operation from the first operation, and transmit the first operation parameter to the first remotely controlled device.

[0013] This disclosure provides a first communication device, a second communication device, a communication method, and a program that can prevent a deterioration in the service quality of XRM.

[0014] Figure 1 shows an example of the configuration of a communication device. Figure 2 shows the flow of communication processing performed in the communication device. Figure 3 shows an example of the configuration of a communication device. Figure 4 shows the flow of communication processing performed in the communication device. Figure 5 shows an example of the configuration of a communication system. Figure 6 shows the flow of communication processing related to the linking of AI / ML models. Figure 7 shows the flow of communication processing between UEs. Figure 8 shows an example of the configuration of a communication system. Figure 9 shows the flow of communication processing between UEs. Figure 10 shows an example of the configuration of a communication system. Figure 11 is a block diagram showing an example of the configuration of a communication device, UPF, and Computation node.

[0015] Embodiment 1 Figure 1 shows an example of the configuration of a communication device 10. The communication device 10 may be a computer device that operates by having a processor execute a program stored in memory.

[0016] The communication device 10 includes a management unit 11, a communication unit 12, and an estimation unit 13. The management unit 11, the communication unit 12, and the estimation unit 13 may be software or modules whose processing is performed by a processor executing a program stored in memory. Alternatively, the management unit 11, the communication unit 12, and the estimation unit 13 may be hardware such as circuits or chips. For example, the communication device 10 may include, instead of the management unit 11, a generation unit (not shown) for generating a trained model and a storage unit (not shown) for storing the trained model. For example, the communication device 10 may include, instead of the communication unit 12, a receiving unit (not shown) and a transmitting unit (not shown).

[0017] The management unit 11 may be used as a means for managing data. The communication unit 12 may be used as a means for transmitting or receiving data. The estimation unit 13 may be used as a means for estimating processing, operation, etc.

[0018] The management unit 11, the communication unit 12, and the estimation unit 13 may be included in a single communication device 10, or they may be distributed across two or more computer devices. Two or more computer devices in which the management unit 11, the communication unit 12, and the estimation unit 13 are distributed may communicate via a network. Two or more computer devices in which the management unit 11, the communication unit 12, and the estimation unit 13 are distributed may constitute a communication system.

[0019] The management unit 11 manages a machine learning model that uses the actions performed by a remote operator on a remote device and the traffic patterns of communication data generated based on those actions as training data. For example, the management unit 11 generates a learning model based on communication data related to the remote operator's actions and labels indicating the actions of the remote operator. The remote operator may, for example, use XRM to perform actions to operate the remote device. For example, the remote operator may operate the remote device by operating a device displayed as a virtual world on an XR terminal. The XR terminal may be a terminal that provides a world that merges the real world and the virtual world. The XR terminal may be, for example, a head-mounted display that displays 3D (Dimension) images that merge the real world and the virtual world to a user wearing the head-mounted display. XR may be a comprehensive general term that includes, for example, VR (Virtual Reality), AR (Augmented Reality), and MR (Mixed Reality).

[0020] The actions performed by the remote operator may include movements of a part of their body. For example, the actions performed by the remote operator may include moving their hand to operate buttons, levers, handles, etc., on a remote device. In other words, the actions performed by the remote operator may include causing a remote robot to perform actions such as pressing a button, moving a lever, or moving a handle, and the actions performed by the remote operator are not limited to these actions.

[0021] The actions performed by the remote operator are transmitted to the remote device via the network. In other words, the actions performed by the remote operator are included in the communication data and sent to the remote device.

[0022] The traffic pattern of communication data may refer to the pattern of IP (Internet Protocol) packets generated when raw data indicating the operation is sent to a remote device. For example, if raw data indicating the operation is generated within a certain period of time, the frequency of raw data generation will increase as the range of the operation, such as the range of hand movement, increases, and more raw data will be generated as the duration of the operation increases. As the amount of raw data increases or the frequency of raw data generation increases, the packet size of IP packets increases, or the number of IP packets increases. For example, if the amount of raw data increases, the size of a single IP packet may increase, or the number of IP packets may increase due to IP packets being split. Also, as the frequency of raw data generation increases, the transmission interval of IP packets also shortens.

[0023] Traffic patterns may be defined by the packet size, number, transmission interval, etc., of IP packets, which are the communication data. Alternatively, traffic patterns may be defined by the frame size, number, transmission interval, etc., of frames at layers lower than IP packets. Furthermore, for example, traffic patterns may indicate the time variation of the amount of data (or communication volume, etc.) or the number of data (or the number of packets, etc.) of the communication data.

[0024] The learning model managed by the management unit 11 may be generated using AI (Artificial Intelligence). Specifically, the learning model may be a pre-trained learning model generated using a general-purpose machine learning algorithm with operation content and traffic patterns of communication data generated based on the operation content as training data. Individual operation content may be distinguished, for example, by an operation name assigned to each operation content. For example, the training data may consist of a label indicating the operation name and communication data related to the operation. Alternatively, the learning model may be generated using a general-purpose machine learning algorithm with operation content, communication data, and traffic patterns of the communication data as training data. The training data may also be called learning data or teacher data. The pre-trained learning model causes the computer device to function by accepting data as input, performing calculations on the input data based on the parameters of the pre-trained learning model, and outputting the calculation results. The machine learning algorithm may be, for example, an algorithm that performs deep learning using a neural network.

[0025] The communication unit 12 receives first communication data from a remote control terminal that generates first communication data based on first operation content indicating remote operation to the first remotely controlled device. The first remotely controlled device may be a device that operates by remote operation performed by a remote operator. Specifically, the first remotely controlled device may be a robot, an unmanned vehicle, etc. The first remotely controlled device may be equipped with a computer device that controls the operation of the first remotely controlled device by receiving communication data related to remote operation.

[0026] The remote control terminal detects the first action performed by the remote operator. The communication unit 12 may receive an IP packet or frame as first communication data generated in conjunction with the generation of raw data indicating the first action at the remote control terminal.

[0027] The estimation unit 13 uses a learning model to estimate the first operation content from the first traffic pattern of the first communication data. The learning model may also cause the communication device 10 to function by receiving data indicating the traffic pattern, performing calculations on the received data, and then outputting the operation content. In other words, the trained learning model may be a program or software that causes a computer device to function.

[0028] The communication unit 12 transmits the first operation details (estimated result) to the communication device 20, which controls the operation of the first remotely controlled device based on the first operation details. The communication device 20 may be a device that determines or identifies the operation parameters for operating the first remotely controlled device. The communication device 20 may be connected to the first remotely controlled device via a network, or it may be mounted on the first remotely controlled device.

[0029] Figure 2 shows the flow of communication processing performed in the communication device 10. For example, the communication device 10 manages a machine learning model that uses the actions performed by a remote operator on a remote device and the traffic patterns of communication data generated based on those actions as training data.

[0030] First, the communication unit 12 receives first communication data from a remote control terminal that has generated first communication data based on first operation content indicating remote control of the first remotely controlled device (S11). Next, the estimation unit 13 estimates the first operation content from the first traffic pattern of the first communication data using a learning model (S12). Next, the communication unit 12 transmits the first operation content (estimated result) to a second communication device that controls the operation of the first remotely controlled device based on the first operation content (S13).

[0031] Figure 3 shows an example configuration of the communication device 20. The communication device 20 may be a computer device that operates by having a processor execute a program stored in memory.

[0032] The communication device 20 includes a management unit 21, a communication unit 22, and an estimation unit 23. The management unit 21, the communication unit 22, and the estimation unit 23 may be software or modules whose processing is performed by a processor executing a program stored in memory. Alternatively, the management unit 21, the communication unit 22, and the estimation unit 23 may be hardware such as circuits or chips. For example, the communication device 20 may have a generation unit (not shown) for generating trained models and a storage unit (not shown) for storing the trained models instead of the management unit 21. For example, the communication device 20 may have a receiving unit (not shown) and a transmitting unit (not shown) instead of the communication unit 22.

[0033] The management unit 21 may be used as a means for managing data. The communication unit 22 may be used as a means for transmitting or receiving data. The estimation unit 23 may be used as a means for estimating processing, operation, etc.

[0034] The management unit 21, the communication unit 22, and the estimation unit 23 may be included in a single communication device 20, or they may be distributed across two or more computer devices. Two or more computer devices in which the management unit 21, the communication unit 22, and the estimation unit 23 are distributed may communicate via a network. Two or more computer devices in which the management unit 21, the communication unit 22, and the estimation unit 23 are distributed may constitute a communication system.

[0035] The management unit 21 manages a machine learning model that uses the actions performed by a remote operator on a remote device and the action parameters that enable the remotely operated device to perform actions according to those actions as training data. The action parameters may be parameters that specify, for example, which part of the remote device to move and to what extent. For example, for the action of raising the right arm of a robot, the action parameters may be parameters that specify that the target of the action is the robot's right arm, that it should move forward A degrees, and at an angular velocity of B radians per second. For example, the management unit 21 may use labels indicating the operator's actions to operate the remote device and the action parameters for operating the remote device as training data to generate a trained model for estimating the action parameters of the remote device from data related to the operator's actions.

[0036] The learning model managed by the management unit 21 may be generated using AI. Specifically, the learning model may be a pre-trained learning model generated using a general-purpose machine learning algorithm with the operation content and the operation parameters that realize the operation according to the operation content as training data.

[0037] The communication unit 22 receives information (estimation result) indicating the first operation content from the communication device 10, which has estimated the first operation content indicating remote operation of the first remotely controlled device.

[0038] The estimation unit 23 uses a learning model to estimate first operation parameters that realize an operation according to the first operation content (estimation result) from the first operation content. The learning model may function the communication device 20 to receive data indicating the operation content, perform calculations on the received data, and then output the operation parameters. In other words, the trained learning model may be a program or software that functions a computer device.

[0039] The communication unit 22 transmits the first operation parameter to the first remotely controlled device. The first remotely controlled device performs the operation specified in the first operation parameter.

[0040] Figure 4 shows the flow of communication processing performed in the communication device 20. The communication device 20 manages a machine learning model that uses the actions performed by a remote operator on a remote device and the action parameters that enable the device to perform actions according to those actions as training data.

[0041] First, the communication unit 22 receives a first operation content from the communication device 10, which has estimated a first operation content indicating remote operation of the first remotely controlled device (S21). Next, the estimation unit 23 uses a learning model to estimate a first operation parameter that realizes the operation according to the first operation content (S22). Next, the communication unit 22 transmits the first operation parameter to the first remotely controlled device (S23).

[0042] As explained above, the communication device 10 can identify the operation content based on the traffic pattern of the communication data without analyzing the operation content contained in the communication data. In other words, the communication device 10 does not need to analyze the application data in order to identify the operation content contained in the communication data as application data. Therefore, the communication device 10 can identify the operation content of the first remotely controlled device earlier than when analyzing the application data, and thus can reduce the delay in communication between the remote control terminal and the first remotely controlled device. As a result, the communication system including the communication device 10 can improve the quality of XRM services.

[0043] Embodiment 2 Fig. 5 shows a configuration example of a communication system. The communication system in Fig. 5 includes a UE (User Equipment) 31, a UPF (User Plane Function) 32, a UE 41, a UPF 42, an SMF (Session Management Function) 50, and an AF (Application Function) 60. UE 31, UPF 32, UE 41, UPF 42, SMF 50, and AF 60 may be computer devices that operate when a processor executes programs stored in memory. UPF, SMF, and AF may be nodes. A node may correspond to an entity (apparatus), or may correspond to a function.

[0044] UE 31 and UE 41 are generic terms for communication terminals used in 3GPP (registered trademark) (3rd Generation Partnership Project). For example, UE 41 may be a remotely operated robot. Alternatively, UE 41 may be a communication terminal incorporated in a remotely operated robot. UE 31 is a remote control terminal that remotely controls UE 41, and may be, for example, an XR terminal. Alternatively, UE 31 may be a communication terminal installed with an XR application used for remotely controlling UE 41.

[0045] UPF 32 and UPF 42 are apparatuses that relay User Plane Data transmitted between UE 31 and UE 41. Furthermore, UPF 32 and UPF 42 may be communication apparatuses that perform communication between UE 31 or UE 41 and a data network (not shown). In other words, UPF 32 and UPF 42 may be gateway apparatuses between UE 31 or UE 41 and the data network. UPF 32 corresponds to the communication apparatus 10 in Fig. 1. UPF 42 corresponds to the communication apparatus 20 in Fig. 3.

[0046] A UPF that serves as an anchor point between UE 31 or UE 41 and a data network may be referred to as a PSA (Protocol Data Unit Session Anchor) UPF. In FIG. 5, UPF 32 and UPF 42 may be PSA UPFs. Here, a UPF arranged between UE 31 or UE 41 and a PSA UPF may be referred to as an I (Intermediary)-UPF. That is, an I-UPF may be arranged between UE 31 and UPF 32, and further between UE 41 and UPF 42.

[0047] The data network may be an external network different from a mobile network configured by nodes defined in 3GPP. The data network may be a network where server devices and the like managed by a service provider or the like are arranged. For example, the data network may be a network managed by a cloud provider that provides cloud services.

[0048] SMF 50 manages PDU (Protocol Data Unit) sessions. For example, SMF 50 may establish a communication path between UE 31 and UE 41 by selecting UPF 32 and UPF 42 to be used for communication between UE 31 and UE 41 from among a plurality of UPFs.

[0049] AF 60 may be, for example, a node that provides a service for managing a 5G VN (Virtual Network) group. A 5G VN group may be, for example, a group including a plurality of UEs to which communication is provided in a virtual network. In other words, the 5G VN group may be a group including a plurality of UEs to which private communication is provided in a virtual network. Specifically, AF 60 may manage a plurality of UEs related to a specific XR service by including them in one 5G VN group. The virtual network may be a private network constructed on a mobile network. The private network may be, for example, a network in which UEs permitted to access are restricted.

[0050] Figure 6 shows the flow of communication processing related to the coordination of AI / ML (Machine Learning) models. The AI / ML model may be a learning model generated by performing machine learning. The AI / ML model corresponds to a trained learning model. The coordination of AI / ML models may also involve coordinating the respective AI / ML models of different UPFs. Here, we will describe an example in which the AI / ML models of UPF32 and UPF42, which relay U-plane data when UE31 remotely controls UE41, are coordinated.

[0051] First, AF60 sends a request message to UE31 (S31). Then, AF60 sends a request message to UE41 (S32). Steps S31 and S32 may be executed at substantially the same time, or step S31 may be executed after step S32 has been executed.

[0052] The request message may be provided in a service using HTTP (HyperText Transfer Protocol). For example, AF60 may provide it to UPF32 and UPF42 in an Nnef_ParameterProvision_Create / Update / Delete request. Specifically, AF60 may provide data related to the 5G VN group to UPF32 and UPF42 in an Nnef_ParameterProvision_Create / Update / Delete request.

[0053] Data relating to a 5G VN group may include, for example, identification information that identifies 5G VN group communications, the data rate applied to 5G VN group communications, and the security policy applied to 5G VN group communications. Furthermore, data relating to a 5G VN group may include members included in the 5G VN group identified by GPSI (Generic Public Subscription Identifier), and identification information that identifies the 5G VN group. Members included in a 5G VN group may be, for example, UEs. Furthermore, data relating to a 5G VN group may include identification information of UPFs that relay data in 5G VN group communications. Furthermore, data relating to a 5G VN group may include identification information of AI / ML models used in XR services. The XR service is provided to members included in the 5G VN group.

[0054] For example, in step S31, AF60 may notify UPF32 of the identification information of the AI / ML model held by UPF42. Furthermore, in step S32, AF60 may notify UPF42 of the identification information of the AI / ML model held by UPF32.

[0055] UPF32 may manage the identification information of AI / ML models possessed by UPF32 by associating it with the identification information of AI / ML models possessed by UPF42. UPF42 may manage the identification information of AI / ML models possessed by UPF42 by associating it with the identification information of AI / ML models possessed by UPF32.

[0056] Alternatively, in steps S31 and S32, AF60 may notify UPF32 and UPF42 of information indicating a pair of AI / ML model identification information held by UPF32 and AI / ML model identification information held by UPF42.

[0057] By executing the communication processing related to the coordination of AI / ML models shown in Figure 6, UPF32 and UPF42 recognize the AI / ML models that each device possesses and the AI / ML models that it coordinates with.

[0058] Here, we will explain the AI / ML models of UPF32 and UPF42. The AI / ML model of UPF32 is a learning model that, for example, takes the traffic pattern of communication data received from the XR terminal UE31 as input and infers the actions to be performed in UE41, which infers or reproduces the actions of a person wearing the XR terminal.

[0059] UE31 monitors human movement using motion tracking sensors, etc. As a result, UE31 acquires raw data such as three-dimensional position data of hands, feet, and other body parts, movement speed, acceleration, and joint angles. UE31 transmits the raw data to UPF32 using communication data such as IP packets or frames.

[0060] For example, the more actively a person moves their hands or feet, the more raw data is obtained from the motion tracking sensor, resulting in a higher frequency of IP packet generation. Furthermore, during specific motion phases, such as when hand or foot movements reach their peak speed, or during static phases like holding an object, raw data containing important motion information is generated frequently. As a result, both the frequency and size of IP packets may increase.

[0061] Furthermore, the size of the IP packets may increase when the hand or foot starts or ends an action, as UE31 provides initial or termination information. This initial or termination information may include information indicating the initial or stopping position of the hand or foot. If there is a waiting phase between hand or foot movements, the amount of data transmitted by UE31 will also decrease, which may increase the interval between IP packets. In addition, as the amount of data transmitted by UE31 decreases, the size of the IP packets may also decrease.

[0062] Furthermore, when comparing hand movements to foot movements, finger movements are more complex than toe movements. Therefore, raw data representing hand movements is likely to be more numerous and occur more frequently than raw data representing foot movements. Consequently, IP packets generated based on raw data representing hand movements will occur more frequently and may also be larger in size than IP packets generated based on raw data representing foot movements. In this way, the IP packet traffic pattern can change depending on the body part.

[0063] The AI / ML model in UPF32 is a learning model generated by training on traffic patterns of communication data and human actions that caused the communication data. Individual human actions are distinguished, for example, by a label name that represents the name of the action. For example, in the learning process, the AI / ML model in UPF32 is trained using a specific label name Y for a given traffic pattern X. Also, if there is a traffic pattern X' that is slightly different from traffic pattern X, but the human actions are the same, a specific label name Y is associated with it and the model is trained accordingly. By training the AI / ML model in UPF32 to associate traffic pattern X and similar patterns with label name Y, the AI / ML model in UPF32 will be able to infer from the traffic patterns from UE31 after deployment whether there were human actions associated with label name Y. The training data, which consists of traffic patterns of communication data and human actions, may be generated by performing simulations. Alternatively, training data including traffic patterns and the actions indicated by those traffic patterns may be generated by actually operating the UE31 and analyzing the communication data received from the UE31.

[0064] The AI / ML model in UPF42 is a learning model that takes the actions of the XR terminal UE31 as input and infers the action parameters that enable the actions of the robot UE41.

[0065] UE41 is a robot that simulates the movements of UE31. Therefore, UE41 may operate by combining the movements of various joints. In order to realize the movements of UE31 in UE41, UE41 operates according to motion parameters such as which joints of UE41 to move, by how much, and for how long to move them.

[0066] The AI / ML model of UPF42 is a learning model generated by training data such as the operation details of the XR terminal UE31 and the operation parameters required to realize the operation of the XR terminal in a robot. The training data, which consists of the operation details of the XR terminal and the operation parameters of the robot, may be generated by performing a simulation, or they may be generated according to the raw data generated by actually operating UE31 and UE41.

[0067] Figure 7 shows the flow of communication processing between UE31 and UE41. First, UE31 generates raw data by monitoring human movements (S41). Next, UE31 generates communication data according to the generation of raw data (S42). The communication data uses the raw data as data in the application layer, which is a higher layer than the IP layer. In other words, the raw data may be set as application data in the payload of the IP packet. The packet size of the IP packet is assumed to be predetermined. Therefore, the number of IP packets, the communication interval between IP packets, the IP packet size, etc., are determined according to the amount of raw data.

[0068] By shortening the cycle for generating communication data from raw data in UE31, the operation of UE31 can be classified into shorter cycles. In other words, the shorter the cycle for generating communication data from raw data, the more finely the operation of UE31 can be classified.

[0069] Next, UE31 transmits the communication data to UPF32 (S43). After receiving the communication data, UPF32 may analyze the communication data and generate a traffic pattern.

[0070] Next, UPF32 inputs the traffic pattern of the received communication data into the AI / ML model to obtain an inference result of the operation of UE31 (S44). In other words, UPF32 holds the inference result output from the AI / ML model.

[0071] Next, UPF32 sends data containing the inference results to UPF42 (S45). UPF32 sends the data to UPF42, which holds the AI / ML model associated with the AI / ML model held by UPF32.

[0072] Next, UPF42 inputs the operation details of UE31, which are shown as inference results, into the AI / ML model to obtain the operation parameters of UE41 as inference results. In other words, UPF42 holds the inference results output from the AI / ML model.

[0073] Next, UPF42 transmits data including the inference results to UE41 (S47). UE41 operates as a robot by receiving the operation parameters indicated as the inference results.

[0074] As explained above, the actions performed in UE31 are notified to UE41 by performing inference using the AI / ML models held in UPF32 and UPF42. Furthermore, UPF32 and UPF42 perform inference using the AI / ML models by using the traffic patterns of IP packets without performing any processing on application data. As a result, in order to perform application processing on the data transmitted from UE31, it is not necessary, for example, for an application server located in the data network to identify the actions of UE31. In other words, the actions and parameters are identified at a location closer to UE31 and UE42 than for an application server located in the data network, which is an external network of the mobile network including UPF32 and UPF42. Moreover, UPF32 and UPF42 perform processing using IP packets, which are data at a lower layer of the application layer. Therefore, the delay that occurs between the time UE31 operates using the XR terminal and the time UE41 operates can be shortened.

[0075] Embodiment 3 Figure 8 shows an example of the configuration of a communication system. In Figure 8, the anchor point UPF is one of the UPF 70, which is different from the communication system in Figure 6. In the communication system in Figure 6, UPF 32 and UPF 42 are the anchor points.

[0076] In the communication system shown in Figure 8, UPF70 holds the AI / ML model. The AI / ML model held by UPF70 may be two models: the AI / ML model held by UPF32 and the AI / ML model held by UPF42. Alternatively, the AI / ML model held by UPF70 may be a single AI / ML model formed by combining the AI / ML model held by UPF32 and the AI / ML model held by UPF42.

[0077] When UPF70 holds a single AI / ML model, the AI / ML model takes the traffic pattern or communication data as input and outputs the operating parameters as the inference result.

[0078] Figure 9 shows the flow of communication processing between UE31 and UE41. Steps S51 to S53 are the same as steps S41 to S43 in Figure 7, so a detailed explanation is omitted.

[0079] UPF70 inputs the traffic pattern of the received communication data into the AI / ML model to obtain the operating parameters of UE31 as an inference result (S54). If UPF70 has two AI / ML models, it may also obtain the operating content as an inference result by taking the traffic pattern as input. Furthermore, UPF70 may also obtain the operating parameters as an inference result by taking the operating content as input.

[0080] Step S55 is the same as step S47 in Figure 7, so a detailed explanation is omitted.

[0081] As explained above, the communication system in Figure 8, like the communication system in Figure 6, can reduce the delay between the operation performed in UE41 and the operation performed in UE31.

[0082] Embodiment 4 Figure 10 shows an example of the configuration of a communication system. In the communication system of Figure 10, a Computation node 80 (which may also be referred to as a Computation Node 80 or Network Node 80) is placed between UPF 32 and UPF 42. The Computation node 80 may be a computer device that operates by having a processor execute a program stored in memory. The Computation node 80 is assumed to have an AI / ML model (trained model) that infers the content of actions from traffic patterns and an AI / ML model that infers action parameters from the content of actions. Alternatively, the Computation node 80 is assumed to have an AI / ML model that estimates the actions of a person (operator) from traffic patterns. For example, the Computation node 80 may infer action parameters based on the trained model and traffic pattern data relating to the operator's actions.

[0083] Computation node 80 may be deployed on a mobile network, for example, as a node that performs inference using an AI / ML model. The mobile network may be a network composed of nodes or functions defined in 3GPP. On the other hand, the data network may be an external network to the mobile network and may be a network not defined in 3GPP.

[0084] Figure 10 shows one Computation node 80, but multiple Computation nodes 80 may be deployed in the mobile network. For example, each Computation node 80 may have a different AI / ML model. For example, each Computation node 80 may be used to provide different services. XRM may be one of the services provided by the Computation node 80. For example, each Computation node 80 may have multiple different AI / ML models.

[0085] The interfaces between UPF32 and Computation node 80, and between UPF42 and Computation node 80, may use interfaces defined in 3GPP.

[0086] When UPF32 receives communication data from UE31, it may select a Computation node 80 from among several Computation nodes 80 that has (holds) an AI / ML model related to the service used by UE31. UPF32 then transmits the communication data to the selected Computation node 80. Alternatively, SMF50 may select UPF32 and UPF42, as well as a Computation node 80.

[0087] As explained above, a dedicated node with an AI / ML model may be deployed in the mobile network. By deploying a dedicated node with an AI / ML model, the processing load on UPF32 and UPF42, which perform data communication processing, is reduced.

[0088] Figure 11 is a block diagram showing an example configuration of communication device 10, communication device 20, UPF 32, UPF 42, UPF 70, and Computation node 80 (hereinafter referred to as "communication device 10, etc."). Referring to Figure 11, communication device 10, etc. includes a network interface 1201, a processor 1202, and memory 1203. The network interface 1201 may be used to communicate with network nodes. The network interface 1201 may include, for example, a network interface card (NIC) compliant with the IEEE 802.3 series. IEEE stands for Institute of Electrical and Electronics Engineers.

[0089] The processor 1202 reads and executes software (computer programs) from the memory 1203 to perform the processing of the communication device 10, etc., as explained using the flowchart. The processor 1202 may be, for example, an MPU (Micro Processor Unit) or a CPU (Central Processing Unit). The processor 1202 may include multiple processors.

[0090] Memory 1203 is composed of a combination of volatile and non-volatile memory. Memory 1203 may include storage located away from the processor 1202. In this case, the processor 1202 may access memory 1203 via an I / O (Input / Output) interface, which is not shown.

[0091] In the example shown in Figure 11, memory 1203 is used to store a group of software modules. The processor 1202 can perform processing on the communication device 10, etc., by reading and executing these software modules from memory 1203.

[0092] As explained using Figure 11, each processor in the communication device 10, etc., executes one or more programs that include a set of instructions for causing the computer to perform the algorithm described in the diagram.

[0093] In the examples described above, the program includes a set of instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more of the functions described in the embodiments. The program may be stored in a non-temporary computer-readable medium or a physical storage medium. Examples, but not limited to, include RAM (random-access memory), ROM (read-only memory), flash memory, SSD (solid-state drive), or other memory technologies, CD-ROM, DVD (digital versatile disc), Blu-ray® disc, or other optical disc storage, magnetic cassette, magnetic tape, magnetic disk storage, or other magnetic storage devices. The program may be transmitted over a temporary computer-readable medium or a communication medium. Examples, but not limited to, include, a temporary computer-readable medium or a communication medium that includes an electrical, optical, acoustic, or other form of propagating signal.

[0094] Although the present disclosure has been described above with reference to embodiments, the present disclosure is not limited to the embodiments described above. Various modifications to the structure and details of the present disclosure can be made as can be understood by those skilled in the art within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0095] Each drawing is merely illustrative to illustrate one or more embodiments. Each drawing may be associated with one or more other embodiments, rather than being associated with only one specific embodiment. As those skilled in the art will understand, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings, for example, to create embodiments not explicitly shown or described. Not all features or steps shown in any one drawing to illustrate an exemplary embodiment are necessarily required, and some features or steps may be omitted. The order of steps described in any of the drawings may be changed as appropriate.

[0096] Some or all of the above embodiments may also be described as follows, but are not limited to the following: (Note 1) A first communication device comprising: a management unit that manages a learning model that has been machine-trained using the content of actions performed by a remote operator on a remotely located device and the traffic pattern of communication data generated based on the content of the actions as learning data; a communication unit that receives the first communication data from a remote operation terminal that generates the first communication data based on the first content of actions indicating remote operation on a first remotely controlled device; and an estimation unit that estimates the first content of actions from the first traffic pattern of the first communication data using the learning model, wherein the communication unit transmits the first content of actions to a second communication device that controls the operation of the first remotely controlled device based on the first content of actions. (Note 2) The first communication device according to Note 1, wherein the communication data and the first communication data are packet data transmitted over a network. (Note 3) The first communication device according to Note 1 or 2, wherein the content of actions is an action of at least one of the remote operator's hands and feet. (Note 4) The communication unit is the first communication device according to Note 1 or 2, which receives the first communication data including sensor data indicating the first operation content. (Note 5) The traffic pattern is the generation pattern of the communication data that occurs together with the generation of sensor data indicating the operation content, according to Note 1 or 2, for the first communication device. (Note 6) The second communication device comprises: a management unit that manages a learning model that has been machine-trained using operation content performed by a remote operator on a remotely located device and operation parameters that realize the operation in accordance with the operation content on the device as learning data; a communication unit that receives the first operation content from a first communication device that has estimated the first operation content indicating remote operation on a first remotely controlled device; and an estimation unit that uses the learning model to estimate first operation parameters that realize the operation in accordance with the first operation content from the first operation content, wherein the communication unit transmits the first operation parameters to the first remotely controlled device.(Note 7) A third communication device comprising: a management unit that manages a machine learning model that uses as learning data the actions performed by a remote operator on a remote device, the traffic pattern of communication data generated based on the actions, and the operation parameters that enable the device to perform actions according to the actions; a communication unit that receives first communication data from a remote operation terminal that generates first communication data based on first actions indicating remote operation on a first remotely controlled device; and an estimation unit that uses the learning model to estimate first operation parameters that enable actions according to the first actions from the first traffic pattern of the first communication data, wherein the communication unit transmits the first operation parameters to the first remotely controlled device. (Note 8) A communication method performed in a first communication device, comprising: receiving first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; using a machine learning model that has been trained with operation content performed by a remote operator on a remotely located device and traffic patterns of communication data generated based on said operation content as training data, estimating the first operation content from the first traffic pattern of the first communication data; and transmitting the first operation content to a second communication device that controls the operation of the first remotely controlled device based on the first operation content. (Note 9) A communication method performed in a second communication device, which receives a first operation content indicating remote operation of a first remotely controlled device from a first communication device that has estimated the first operation content, estimates a first operation parameter that realizes an operation according to the first operation content from the first operation content using a machine learning model that has been trained with the operation content performed by a remote operator on a remotely located device and operation parameters that realize an operation according to the operation content on the device as training data, and transmits the first operation parameter to the first remotely controlled device.(Note 10) A communication method performed in a third communication device, comprising: receiving first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; using a machine learning model that has been trained with operation content performed by a remote operator on a remotely located device, traffic patterns of communication data generated based on the operation content, and operation parameters that realize operation in accordance with the operation content on the device as training data, estimating first operation parameters that realize operation in accordance with the first operation content from the first traffic pattern of the first communication data; and transmitting the first operation parameters to the first remotely controlled device. (Note 11) A program that causes a computer to perform the following actions: receive first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; estimate the first operation content from the first traffic pattern of the first communication data using a machine learning model that has been trained with the operation content performed by a remote operator on a remotely located device and the traffic pattern of the communication data generated based on the operation content as training data; and transmit the first operation content to a second communication device that controls the operation of the first remotely controlled device based on the first operation content. (Note 12) A program that causes a computer to receive a first operation from a first communication device that has estimated a first operation indicating remote operation of a first remotely controlled device, to estimate a first operation from the first operation to realize an operation according to the first operation using a machine learning model that has been trained with the operation to be performed by a remote operator on a remote device and the operation parameters that realize an operation according to the operation to be performed on the device as training data, and to transmit the first operation to the first remotely controlled device.(Note 13) A program that causes a computer to perform the following actions: receive first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; estimate first operation parameters that realize operation according to the first operation content from the first traffic pattern of the first communication data using a machine learning model that has been trained on operation content performed by a remote operator on a remotely located device, traffic pattern of communication data generated based on said operation content, and operation parameters that realize operation according to said operation content on the device as training data; and transmit the first operation parameters to the first remotely controlled device.

[0097] Some or all of the elements (e.g., configuration and function) described in Appendices 2 to 5 that are subordinate to Appendice 1 may also be subordinate to Appendices 6 to 13 in the same manner as those described in Appendices 2 to 5. Some or all of the elements described in any appendice may be applied to various hardware, software, recording means, systems, and methods for recording software.

[0098] This application claims priority based on Japanese Patent Application No. 2025-051133, filed on 26 March 2025, and incorporates all of its disclosures herein.

[0099] 10 Communication device 11 Management unit 12 Communication unit 13 Estimation unit 20 Communication device 21 Management unit 22 Communication unit 23 Estimation unit 31 UE 32 UPF 41 UE 42 UPF 50 SMF 60 AF 70 UPF 80 Computation node

Claims

1. A first communication device comprising: management means for managing a machine learning model that uses the actions performed by a remote operator on a remotely located device and the traffic patterns of communication data generated based on the actions as training data; communication means for receiving first communication data from a remote operation terminal that generates first communication data based on first actions indicating remote operation on a first remotely controlled device; and estimation means for estimating the first actions from the first traffic patterns of the first communication data using the learning model, wherein the communication means transmits the first actions to a second communication device that controls the operation of the first remotely controlled device based on the first actions.

2. The first communication device according to claim 1, wherein the communication data and the first communication data are packet data transmitted over a network.

3. The first communication device according to claim 1 or 2, wherein the operation is an operation of at least one of the remote operator's hands and feet.

4. The first communication device according to claim 1 or 2, wherein the communication means receives the first communication data including sensor data indicating the first operation content.

5. The first communication device according to claim 1 or 2, wherein the traffic pattern is a pattern of communication data generated together with the generation of sensor data indicating the operation content.

6. A second communication device comprising: management means for managing a machine learning model that uses as training data the actions performed by a remote operator on a remotely located device and the action parameters that enable the device to perform actions according to the actions; communication means for receiving the first actions from a first communication device that has estimated the first actions indicating remote operation on a first remotely controlled device; and estimation means for estimating first action parameters that enable actions according to the first actions from the first actions using the machine learning model, wherein the communication means is a second communication device that transmits the first action parameters to the first remotely controlled device.

7. A third communication device comprising: management means for managing a machine learning model that uses as training data the actions performed by a remote operator on a remotely located device, the traffic pattern of communication data generated based on the actions, and the operation parameters that enable the device to perform actions according to the actions; communication means for receiving first communication data from a remote operation terminal that generates first communication data based on first actions indicating remote operation on a first remotely controlled device; and estimation means for estimating first operation parameters that enable actions according to the first actions from the first traffic pattern of the first communication data using the learning model, wherein the communication means transmits the first operation parameters to the first remotely controlled device.

8. A communication method performed in a first communication device, comprising: receiving first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; using a machine learning model trained with operation content performed by a remote operator on a remotely located device and the traffic pattern of communication data generated based on said operation content as training data, estimating the first operation content from the first traffic pattern of the first communication data; and transmitting the first operation content to a second communication device that controls the operation of the first remotely controlled device based on the first operation content.

9. A communication method performed in a second communication device, which receives a first operation content indicating remote operation of a first remotely controlled device from a first communication device that has estimated the first operation content, estimates a first operation parameter that realizes an operation according to the first operation content from the first operation content using a machine learning model that has been trained with the operation content performed by a remote operator on a remotely located device and operation parameters that realize an operation according to the operation content on the device as training data, and transmits the first operation parameter to the first remotely controlled device.

10. A communication method performed in a third communication device, comprising: receiving first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; using a machine learning model that has been trained with training data consisting of operation content performed by a remote operator on a remotely located device, a traffic pattern of communication data generated based on the operation content, and operation parameters that realize operation in accordance with the operation content on the device, estimating first operation parameters that realize operation in accordance with the first operation content from the first traffic pattern of the first communication data; and transmitting the first operation parameters to the first remotely controlled device.

11. A program that causes a computer to perform the following actions: receive first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; estimate the first operation content from the first traffic pattern of the first communication data using a machine learning model that has been trained with the operation content performed by a remote operator on a remote device and the traffic pattern of the communication data generated based on the operation content as training data; and transmit the first operation content to a second communication device that controls the operation of the first remotely controlled device based on the first operation content.

12. A program that causes a computer to receive a first operation from a first communication device that has estimated a first operation indicating remote operation of a first remotely controlled device, estimate a first operation from the first operation to realize an operation according to the first operation using a machine learning model that has been trained with the operation performed by a remote operator on a remote device and the operation parameters that realize the operation according to the operation as training data, and transmit the first operation to the first remotely controlled device.

13. A program that causes a computer to perform the following actions: receive first communication data from a remote operation terminal that generates first communication data based on first operation content indicating remote operation of a first remotely controlled device; estimate first operation parameters that realize operation according to the first operation content from the first traffic pattern of the first communication data using a machine learning model that has been trained with training data consisting of operation content performed by a remote operator on a remotely located device, traffic patterns of communication data generated based on the operation content, and operation parameters that realize operation according to the operation content on the device; and transmit the first operation parameters to the first remotely controlled device.

Images