WO2024012319A1 - 用于无线通信的电子设备和方法、计算机可读存储介质 - Google Patents

用于无线通信的电子设备和方法、计算机可读存储介质 Download PDF

Info

Publication number
WO2024012319A1
WO2024012319A1 PCT/CN2023/105811 CN2023105811W WO2024012319A1 WO 2024012319 A1 WO2024012319 A1 WO 2024012319A1 CN 2023105811 W CN2023105811 W CN 2023105811W WO 2024012319 A1 WO2024012319 A1 WO 2024012319A1
Authority
WO
WIPO (PCT)
Prior art keywords
electronic device
user equipment
channel
edge link
processing circuit
Prior art date
Application number
PCT/CN2023/105811
Other languages
English (en)
French (fr)
Inventor
陈巍
吴俊杰
郑策
王晓雪
孙晨
Original Assignee
索尼集团公司
陈巍
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 索尼集团公司, 陈巍 filed Critical 索尼集团公司
Publication of WO2024012319A1 publication Critical patent/WO2024012319A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/10—Scheduling measurement reports ; Arrangements for measurement reports

Definitions

  • the present disclosure relates to the field of wireless communication technology, and in particular to an electronic device and method for wireless communication and a computer-readable storage medium. More specifically, it involves grouping learning models of user equipment related to edge links, and jointly training the learning models within the same group.
  • V2V vehicles
  • sidelinks edge links
  • MDP Markov Decision Process
  • DRP Deep Reinforcement Learning
  • an electronic device for wireless communication which includes a processing circuit.
  • the processing circuit is configured to: based on information about at least one user device reported by at least one user device located within the service range of the electronic device. Channel information of the channel state of at least one edge link, dividing the learning model of the user equipment related to the at least one edge link into at least one group, and for at least a part of the groups in the at least one group, Learning models are jointly trained.
  • electronic devices solve the data heterogeneity problem of edge links due to different environments through grouping, which can improve the efficiency of joint training, and can improve the quality of the learning model and the performance of the system.
  • an electronic device for wireless communication which includes a processing circuit.
  • the processing circuit is configured to: report at least one edge link related to the electronic device to a network side device that provides services for the electronic device.
  • the channel information of the channel status is used for the network side device to use the channel information to create a learning model for the electronic devices related to at least one edge link and other electronic devices related to the at least one edge link that are served by the network side device.
  • the method is divided into at least one group, thereby facilitating the network side device to jointly train the learning models in the same group for at least a part of the at least one group.
  • the electronic device reports channel information about the channel status of the edge link to the network side device, so that the network side device groups learning models of the electronic devices related to the edge link based on the channel information. , which helps network-side devices solve the data heterogeneity problem of edge links due to different environments through grouping, improves the efficiency of joint training, and improves the quality of the learning model and the performance of the system.
  • a method for wireless communication including: based on the channel status of at least one edge link of at least one user equipment reported by at least one user equipment located within the service range of the electronic device. channel information, dividing the learning model of the user equipment related to at least one edge link into at least one group, and for At least some of the groups in one group are missing, and the learning models in the same group are jointly trained.
  • a method for wireless communication including: reporting channel information about the channel status of at least one edge link of the electronic device to a network side device that provides services for the electronic device for the network
  • the side device divides learning models of electronic devices related to at least one edge link and other electronic devices related to the at least one edge link and provided by the network side device into at least one group based on the channel information, thereby facilitating the network side
  • the device jointly trains learning models in the same group for at least a part of the at least one group.
  • computer program codes and computer program products for implementing the above-mentioned method for wireless communication are also provided, as well as computers having the computer program codes for implementing the above-mentioned method for wireless communication recorded thereon.
  • readable storage media are also provided.
  • FIG. 1 shows a functional module block diagram of an electronic device for wireless communication according to one embodiment of the present disclosure
  • Figure 2 is a schematic diagram showing a system structure according to an embodiment of the present disclosure
  • Figure 3 is a schematic diagram illustrating an edge link power rate adaptive control scenario according to an embodiment of the present disclosure
  • 4a and 4b are schematic diagrams illustrating the classification based on the degree of similarity between probability distributions of channel energy gains of edge links according to an embodiment of the present disclosure
  • FIG. 5 is an example diagram illustrating information interaction between an electronic device and a user device according to an embodiment of the present disclosure
  • FIG. 6 shows a functional module block diagram of an electronic device for wireless communication according to another embodiment of the present disclosure
  • FIG. 7 shows a flowchart of a method for wireless communication according to one embodiment of the present disclosure
  • FIG. 8 shows a flowchart of a method for wireless communication according to another embodiment of the present disclosure.
  • FIG. 9 is a block diagram illustrating a first example of a schematic configuration of an eNB or gNB to which the technology of the present disclosure may be applied;
  • FIG. 10 is a block diagram illustrating a second example of a schematic configuration of an eNB or gNB to which the technology of the present disclosure may be applied;
  • FIG. 11 is a block diagram illustrating an example of a schematic configuration of a smartphone to which the technology of the present disclosure may be applied;
  • FIG. 12 is a block diagram showing an example of a schematic configuration of a car navigation device to which the technology of the present disclosure can be applied.
  • FIG. 13 is a block diagram of an exemplary structure of a general-purpose personal computer in which methods and/or apparatuses and/or systems according to embodiments of the present invention may be implemented.
  • FIG. 1 shows a functional module block diagram of an electronic device 100 for wireless communication according to one embodiment of the present disclosure.
  • the electronic device 100 includes: a processing unit 101, which can be based on channel information reported by at least one user device located within the service range of the electronic device 100 regarding the channel status of at least one edge link of the at least one user device. , dividing the learning model of the user equipment related to at least one edge link into at least one group; and the training unit 103, which can jointly train the learning models in the same group for at least a part of the at least one group.
  • the processing unit 101 and the training unit 103 may be implemented by one or more processing circuits, and the processing circuit may be implemented as a chip, for example.
  • the electronic device 100 may serve as a network-side device in a wireless communication system. Specifically, for example, it may be provided on the base station side or communicably connected to the base station.
  • the electronic device 100 may be implemented at a chip level, or may also be implemented at a device level.
  • the electronic device 100 may operate as a base station itself, and may also include external devices such as memory, transceivers (not shown), and the like.
  • the memory can be used to store programs and related data information that the base station needs to execute to implement various functions.
  • the transceiver may include one or more communication interfaces to support communication with different devices (eg, user equipment (UE), other base stations, etc.), and the implementation form of the transceiver is not specifically limited here.
  • the base station may be an eNB or a gNB, for example.
  • electronic device 100 may be connected to a core network.
  • the wireless communication system according to the present disclosure may be a 5G NR (New Radio, New Radio) communication system. Further, the wireless communication system according to the present disclosure may include a non-terrestrial network (Non-terrestrial network, NTN). Optionally, the wireless communication system according to the present disclosure may also include a terrestrial network (Terrestrial network, TN). In addition, those skilled in the art can understand that the wireless communication system according to the present disclosure may also be a 4G or 3G communication system.
  • the user equipment may be a user equipment used for sending on the edge link (SL, Sidelink) (referred to as sending user equipment for short) or may be a user equipment used for receiving on the edge link (referred to as receiving user equipment for short).
  • User equipment is capable of edge link control.
  • federated learning is used for joint training of learning models under multi-user devices.
  • the learning model can be a traditional machine learning model or a deep reinforcement learning model.
  • the learning model is a deep reinforcement learning model as an example. illustrate.
  • FIG. 2 is a schematic diagram showing a system structure according to an embodiment of the present disclosure.
  • the user equipment is shown as a vehicle.
  • the user equipment can be in other forms besides vehicles.
  • the user equipment can be a terminal device such as a mobile phone, iPad, notebook, etc. , as long as there is an edge link between user devices.
  • a single user device performs reinforcement learning on a learning model (which may be called a local model) related to its edge link.
  • the local model is obtained based on the initial global model delivered by the electronic device 100 .
  • the electronic device 100 divides the local model on the user equipment into different groups based on the channel information about the channel status of the edge link of the user equipment (for example, for simplicity, only the ones divided into the same group are shown in FIG.
  • the user equipment uploads the parameters of its local model to the electronic device 100; for the user equipments UE1, UE2 and UE3 in the same group, the electronic device 100 performs joint training of the learning models (aggregation of the learning models) through federated learning to form a global model.
  • the electronic device 100 solves the data heterogeneity problem of edge links due to different environments through grouping, can improve the efficiency of joint training, and can improve the quality of the learning model and the performance of the system.
  • At least one user device is a device in a D2D scenario.
  • at least one user device is a vehicle-mounted device in the Internet of Vehicles.
  • user equipment is not limited to V2X car networking scenarios, and any communication scenario linked by sidelink can be applied.
  • communication between user devices can also be mutual communication between terminal devices (mobile phones, tablet computers, etc.).
  • the communication between user devices can also be the communication between XR devices in the XR (extended reality) scenario; and the communication between user devices can also be the communication between devices in the industrial Internet or smart home scenarios, etc.
  • the user equipment is a vehicle or a vehicle-mounted device in the Internet of Vehicles as an example for description.
  • the user equipment may be in other forms besides vehicle-mounted equipment, as long as there is an edge link between the user equipments.
  • MDP Markov Decision Problem
  • a specific example is power rate adaptive control of edge link information transmission.
  • the data transmission of vehicle equipment requires battery power, and the battery capacity is limited. Therefore, in order to have longer battery life, during the data transmission process of the edge link, the average power constraint of the sending device needs to be given.
  • the wireless channel status between vehicles is random, if we only pursue low energy consumption of information transmission and wait for transmission when the channel condition is good (opportunistic transmission), then the data packet will wait in the queue for a long time, causing serious problems. Data queuing delay.
  • Such a power control problem can be modeled as an MDP problem and solved using deep reinforcement learning.
  • the channel information of the edge link includes the probability distribution of the channel energy gain of the edge link, reference signal received power (RSRP), received signal strength index (RSSI), reference signal quality (RSRQ), signal-to-noise ratio (SNR) ), at least one of information about whether the user equipment as the sender and the user equipment as the receiver of the edge link are within line-of-sight, and statistics of channel interference and noise.
  • RSRP reference signal received power
  • RSSI received signal strength index
  • RSRQ reference signal quality
  • SNR signal-to-noise ratio
  • the channel information of edge links is used to measure the similarity between learning models related to edge links.
  • the electronic device 100 performs processing related to the edge link based on the channel information of the edge link.
  • the learning models of user equipment are grouped, and learning models with high similarity can be jointly trained.
  • the processing unit 101 may be configured to perform partitioning based on a degree of similarity between probability distributions respectively corresponding to at least one edge link.
  • the channel energy gain is divided into a predetermined number of discrete bins, and the probability distribution includes a probability that the channel energy gain is at each bin.
  • the sending vehicle adaptively adjusts the data transmission power and the number of data packets sent in real time according to the current inter-vehicle channel status and data queue status. Since the channel status between vehicles is random, if we only pursue real-time information transmission (that is, instant transmission), the energy consumption cost will be very high when encountering poor channel status. If you only pursue low energy consumption of information transmission and wait for transmission when the channel condition is good (that is, opportunistic transmission), then the data packet will wait in the queue for a long time, resulting in serious data queuing delay. Therefore, it is necessary to achieve the optimal delay-power trade-off relationship through efficient adaptive power control, thereby minimizing the data transmission delay while ensuring that the average power consumption requirements are met.
  • the power rate adaptive control of edge links is used as an example to introduce below.
  • the learning model is used to assist in determining the data transmission rate of the edge link based on the data queue length and the channel energy gain of the edge link.
  • edge links can jointly train the learning model through federated learning.
  • DRL deep reinforcement learning
  • the probability distribution characteristics of the wireless channel are the random environment for training.
  • user equipments with similar wireless channel state probability distribution characteristics are selected to group FL, so that vehicles with similar random environments are selected for joint training of federated learning. , which can improve joint training efficiency.
  • FIG. 3 is a schematic diagram illustrating an edge link power rate adaptive control scenario according to an embodiment of the present disclosure.
  • Tx represents transmission and Rx represents reception.
  • the user equipment needs to be based on the current data queue length q (waiting to be sent in the queue).
  • the number of data packets sent), and the current channel energy gain level h determine the current data sending rate s and the current data sending power P (the sending power P is determined by the sending.
  • the rate s is determined by the channel energy gain h, which can be calculated by the channel capacity formula), thereby ensuring that the average queuing delay of data transmission in the edge link is minimized under the constraints of limited average power consumption.
  • How to select the optimal transmission power and transmission rate in real time based on the real-time dynamically changing queue length and channel energy gain can be established as a Markov decision problem. Markov decision-making problems can be further solved using deep reinforcement learning models.
  • the deep reinforcement learning model fits the value function in the reinforcement learning process through an artificial neural network.
  • the input of the artificial neural network is the queue length q, the channel gain level h, and the transmission rate s; the output is the value function V(q,h,s) corresponding to the state (q,h,s) ).
  • the artificial neural network models trained under different probability distributions of channel energy gain will be different.
  • adding edge links with similar probability distributions of channel energy gains to the same federated learning group for training can improve the accuracy of the aggregated global model.
  • 4a and 4b are diagrams illustrating division based on the degree of similarity between probability distributions of channel energy gains of edge links according to an embodiment of the present disclosure.
  • the degree of similarity between probability distributions includes the KL divergence between probability distributions.
  • the smallest KL divergence is selected from the maximum values taken for each pair of KL divergences. Then use this as a basis to select the user equipment with the smallest KL divergence as a group.
  • the minimum value D 34 is selected from the above six maximum values, which corresponds to D KL (P 3
  • the channel state probability distribution with a high degree of similarity represents a reinforcement learning model with a high degree of similarity.
  • the reinforcement learning models with a high degree of similarity are divided into a group for joint training, which can achieve Better joint training results.
  • the processing unit 101 may be configured to perform partitioning based on the magnitude of the RSRP. For example, the learning models of user equipment whose RSRP amplitude is greater than a predetermined threshold can be divided into the same group, and the learning models of user equipment whose RSRP amplitude is less than or equal to the predetermined threshold can be divided into another group. Or, based on the amplitude of RSRP, the learning model of the user equipment The type is divided into multiple groups.
  • the processing unit 101 may be configured to perform partitioning based on the magnitude of the RSSI. For example, the learning models of user equipment whose RSSI amplitude is greater than a predetermined threshold can be divided into the same group, and the learning models of user equipment whose RSSI amplitude is less than or equal to the predetermined threshold can be divided into another group. Or, based on the amplitude of RSSI, the learning model of the user equipment is divided into multiple groups.
  • the processing unit 101 may be configured to perform partitioning based on the magnitude of the RSRQ. For example, the learning models of user equipment whose RSRQ amplitude is greater than a predetermined threshold can be divided into the same group, and the learning models of user equipment whose RSRQ amplitude is less than or equal to the predetermined threshold can be divided into another group. Or, based on the amplitude of RSRQ, the learning model of the user equipment is divided into multiple groups.
  • the processing unit 101 may be configured to perform partitioning based on the magnitude of the SNR. For example, the learning models of user equipment whose SNR amplitude is greater than a predetermined threshold can be divided into the same group, and the learning models of user equipment whose SNR amplitude is less than or equal to the predetermined threshold can be divided into another group. Alternatively, the learning model of the user equipment is divided into multiple groups based on the magnitude of the SNR.
  • the processing unit 101 may be configured to perform division according to whether the user equipment as the sender and the user equipment as the receiver of the edge link are within a line of sight (LOS) range. For example, if several sending user equipments are to be jointly trained, then the sending user equipments that are in line-of-sight range with the corresponding receiving user equipments are divided into the same group, and those that are in line-of-sight range with the corresponding receiving user equipments are divided into the same group.
  • the sending user equipment of the range (NLOS) is divided into another grouping.
  • the processing unit 101 may be configured to perform partitioning based on the size of statistics of interference and noise of the channel.
  • statistics of interference and noise of the channel include mean and/or variance.
  • the channel information of the edge link includes the probability distribution of the channel energy gain of the edge link, RSRP, RSSI, RSRQ, SNR, whether the user equipment as the sender and the user equipment as the receiver of the edge link are related. How to group learning models given an indicator of information within line-of-sight, channel interference and noise statistics.
  • the channel information on the edge link described below includes the probability distribution of the channel energy gain of the edge link, RSRP, RSSI, RSRQ, SNR, and whether the user equipment as the sender and the user equipment as the receiver of the edge link are in view. How to group learning models based on at least two indicators from the statistics of information within the range, interference of the channel, and noise.
  • the learning models may be grouped based on the priorities of the at least two indicators.
  • the priority of indicators can be set based on experience or application scenarios.
  • the learning model is first grouped based on the first priority indicator to obtain the first grouping result; and then based on the first grouping result, the learning model is grouped based on the first priority indicator. The indicators with the second priority are grouped again to obtain the final grouping result.
  • the two indicators include the first indicator RSRP and the second indicator the probability distribution of the channel energy gain of the edge link, and the priority of the first indicator is higher than the priority of the second indicator
  • the learning models for example, the learning models of user equipment whose RSRP amplitude is greater than a predetermined threshold can be divided into the first group, and the learning models of user equipment whose RSRP amplitude is less than or equal to the predetermined threshold can be divided into the second group
  • the first grouping result for example, it includes the first grouping and the second grouping
  • the first grouping and the second grouping are respectively Grouping (for example, divide the learning models with high similarity in the first group into the first sub-group, and divide other learning models in the first sub-group into the second sub-group; divide the highly similar learning models in the second group into
  • the learning model is divided into
  • the learning models are first grouped based on the indicator with the highest priority to obtain the first grouping result, and then based on the first grouping result, the first grouping result is obtained.
  • the second priority indicators are grouped again to obtain the second grouping result; finally, based on the second grouping result, the third priority indicator is grouped again to obtain the final grouping result.
  • grouping can be performed when the at least two indicators include more than four indicators, which will not be described again here.
  • the processing unit 101 may be configured to receive channel information of the edge link via Radio Resource Control (RRC) signaling.
  • RRC Radio Resource Control
  • the RRC signaling may be MeasResultsSL signaling.
  • the signaling "measResultListOther-NR” can be added to the MeasResultsSL signaling to transmit other channel information of the edge link, such as RSRP, RSRQ, RSSI, SNR, LOS, channel interference and noise statistics, etc.
  • both signaling "measResultListPDCS-NR" and “measResultListOther-NR” may be added to the MeasResultsSL signaling.
  • the processing unit 101 may be configured to send partitioning-related information to at least some of the user equipments related to edge links in each group through a physical downlink control channel (PDCCH).
  • PDCH physical downlink control channel
  • the processing unit 101 may be configured to send parameters related to the initial global learning model to at least a portion of the user devices in the first round of joint training. In this way, the user equipment can perform local training based on the initial global model and obtain a local model.
  • the processing unit 101 may be configured to receive auxiliary status information for uplink resource allocation from at least a part of user equipment through an uplink.
  • the user equipment uploads the auxiliary status information to the electronic device 100 for uploading Perform resource allocation.
  • the auxiliary status information includes at least one of the number of samples used by the user equipment to train the learning model, location information of the user equipment, movement speed of the user equipment, computing power of the user equipment, and CPU occupancy of the user equipment.
  • the processing unit 101 may be configured to allocate uplink resources to at least a part of the user equipment based on the assistance status information. That is, the electronic device 100 obtains available wireless resource block information to prepare for local model upload of federated learning.
  • the electronic device 100 performs corresponding uplink resource allocation according to the auxiliary status information uploaded by the user equipment in the group, which can solve or alleviate the Straggler problem, thereby speeding up the FL process and improving system performance.
  • the main manifestations of Straggler are that the user equipment has: 1) a large amount of data: a larger weight in the aggregation process; 2) a higher priority; 3) poor computing power or high CPU usage; and 4 ), is far away from the electronic device 100 or has poor channel quality, resulting in longer transmission time or lower transmission rate.
  • the electronic device 100 can reduce transmission delays and speed up the convergence process of the learning model.
  • the number of samples used by the user device to train the learning model is the number of samples used by the user device to train the local model in each iteration of the federated learning iterative training process. According to the information on the number of samples used by the user equipment to train the local model, the importance of the local model trained by the user equipment is judged, which is used to determine the allocation of uplink wireless resources in the federated learning process.
  • the computing capability of the user equipment used for local model training is the computing speed of the CPU
  • the CPU occupancy rate of the user equipment is the CPU occupancy rate of the local model training process.
  • the CPU usage information used for local model training is used to estimate the computing power of the user device during the federated learning process to judge Straggler.
  • the processing unit 101 may be configured to send information about uplink resource allocation to at least a part of the user equipment through a downlink.
  • the training unit 103 may be configured to receive parameters related to the local learning model uploaded by at least part of the user equipment based on information about uplink resource allocation, where the local learning model is based on the initial global learning model issued by the electronic device 100 And it comes from training.
  • joint training includes aggregating local learning models related to edge links within the same group as an updated global learning model, resulting in an aggregated learning model.
  • aggregation is a weighted average of the parameters of locally learned models associated with edge links within the same group.
  • the electronic device 100 base station determines the importance of the uploaded reinforcement learning model based on the information on the number of training samples used when training the local reinforcement learning model for each edge link, and uses it to determine the weighting coefficient of the local model during aggregation, so that in this round A more accurate global model is aggregated during the iterative process. That is, the electronic device 100 can determine the weight of the local model based on the quantity information of the samples used by the user equipment to train the local model, thereby minimizing the error of the global model trained by federated learning.
  • the training unit 103 may be configured to broadcast parameters related to the aggregated learning model (which may also be referred to as the updated global model) of each group to the user equipment in the group.
  • the aggregated learning model which may also be referred to as the updated global model
  • the training unit 103 may be configured to repeatedly perform partitioning and joint training until a predetermined condition is met.
  • the predetermined condition is that a predetermined number of iterations is reached, or the error of the aggregated learning model is less than the predetermined error, etc.
  • FIG. 5 is an example diagram illustrating information interaction between the electronic device 100 and the user equipment UE according to an embodiment of the present disclosure.
  • the channel information of the edge link is the probability distribution of the channel energy gain of the edge link as an example for explanation.
  • the UE uploads the collected probability distribution to the electronic device 100.
  • the electronic device 100 determines the similarity of the learning model of the user equipment based on the probability distribution of the channel energy gain of the edge link, and divides the learning models with higher similarity into the same group. In S52, the electronic device 100 delivers the grouping information and the initial global model to the UE.
  • the UE uploads the assistance status information to the electronic device 100.
  • the electronic device 100 allocates uplink resources to the user equipment based on the auxiliary status information. For example, the electronic device 100 finds user devices with Straggler problems and allocates more resources to such user devices.
  • the electronic device 100 delivers the uplink resource allocation information to the UE.
  • the UE performs local training based on the initial global model based on local sample data to obtain a local model.
  • the UE uploads the parameters of the local model to the electronic device 100 through the allocated uplink resources.
  • the electronic device 100 aggregates local models of user devices in the same group to obtain an updated global model.
  • the updated global model represents the final model of federated learning in this round of training, and its model error represents the effect of this round of federated learning training.
  • the electronic device 100 broadcasts the updated global model to all user devices participating in federated learning.
  • the training of federated learning requires the UE and the electronic device 100 to perform several rounds of iteration and aggregation of the learning model, that is, repeatedly perform division and joint training, that is, repeatedly perform the processing of S51-S55 until the predetermined conditions are met.
  • FIG. 6 shows a functional module block diagram of an electronic device 600 for wireless communication according to yet another embodiment of the present disclosure.
  • the electronic device 600 includes: a communication unit 601 .
  • the communication unit 601 can report channel information about the channel status of at least one edge link of the electronic device 600 to a network side device that provides services for the electronic device 600 .
  • the network side device divides the learning model of the electronic device 600 related to at least one edge link and other electronic devices related to the at least one edge link and provided by the network side device into at least one group based on the channel information, thereby facilitating
  • the network-side device jointly trains learning models in the same group for at least a part of the at least one group.
  • the communication unit 601 may be implemented by one or more processing circuits, which may be implemented as a chip, for example.
  • the electronic device 600 may, for example, be provided on a user equipment (UE) side or be communicatively connected to the user equipment.
  • the device related to the electronic device 600 may be the user equipment.
  • the electronic device 600 may be implemented at a chip level, or may also be implemented at a device level.
  • the electronic device 600 may operate as a user device itself, and may also include external devices such as memory, transceivers (not shown in the figure), and the like.
  • the memory can be used to store programs and related data information that the user equipment needs to execute to implement various functions.
  • the transceiver may include one or more communication interfaces to support communication with different devices (eg, base stations, other user equipment, etc.), and the implementation form of the transceiver is not specifically limited here.
  • the network side device may be the electronic device 100 mentioned above.
  • the electronic device 600 may be the user device referred to in the above embodiment of the electronic device 100 .
  • the wireless communication system according to the present disclosure may be a 5G NR communication system. Further, the wireless communication system according to the present disclosure may include non-terrestrial networks. Optionally, the wireless communication system according to the present disclosure may also include a terrestrial network. In addition, those skilled in the art can understand that the wireless communication system according to the present disclosure may also be a 4G or 3G communication system.
  • the electronic device 600 reports channel information about the channel status of the edge link to the network side device, so that the network side device can learn a learning model of the electronic device 600 related to the edge link based on the channel information.
  • Grouping helps network-side devices solve data heterogeneity problems caused by different environments on edge links through grouping, improves the efficiency of joint training, and improves the quality of the learning model and the performance of the system.
  • the channel information of the edge link includes the probability distribution of the channel energy gain of the edge link, reference signal received power (RSRP), received signal strength index (RSSI), reference signal quality (RSRQ), signal-to-noise ratio (SNR) ), at least one of information about whether the user equipment as the sender and the user equipment as the receiver of the edge link are within line-of-sight, and statistics of channel interference and noise.
  • RSRP reference signal received power
  • RSSI received signal strength index
  • RSRQ reference signal quality
  • SNR signal-to-noise ratio
  • the network side device performs division based on the similarity between probability distributions respectively corresponding to at least one edge link.
  • classification based on the degree of similarity between probability distributions of channel energy gains of edge links, please refer to the description in conjunction with FIG. 3 in the embodiment of the electronic device 100, which will not be repeated here.
  • the degree of similarity between probability distributions includes the KL divergence between probability distributions.
  • the channel energy gain is divided into a predetermined number of discrete bins, and the probability distribution includes a probability that the channel energy gain is at each bin.
  • the probability that the channel energy gain is at each level please refer to Pi in the embodiment of the electronic device 100, which will not be described again here.
  • the network side device performs division based on the amplitude of RSRP.
  • the network side device performs division based on the amplitude of RSSI.
  • the network side device performs division based on the amplitude of RSRQ.
  • the network side device performs division based on the amplitude of SNR.
  • the network side device performs classification according to whether the electronic device as the sender and the electronic device as the receiver of the edge link are within line-of-sight range.
  • the network side device performs division based on the size of the interference and noise statistics of the channel.
  • statistics of interference and noise of the channel include mean and/or variance.
  • the communication unit 601 may be configured to report channel information via Radio Resource Control RRC signaling.
  • RRC signaling may be MeasResultsSL signaling.
  • MeasResultsSL please refer to the description in the embodiment of the electronic device 100, which will not be repeated here.
  • the communication unit 601 may be configured to receive information about partitioning from the network side device through a physical downlink control channel (PDCCH).
  • PDCH physical downlink control channel
  • the communication unit 601 may be configured to receive parameters related to the initial global learning model in the first round of joint training.
  • the communication unit 601 may be configured to send auxiliary status information for uplink resource allocation to the network side device through an uplink.
  • the auxiliary status information includes at least one of the number of samples used by the electronic device 600 to train the learning model, the location information of the electronic device 600 , the moving speed of the electronic device 600 , the computing power of the electronic device 600 , and the CPU occupancy rate of the electronic device 600 one.
  • auxiliary status information please refer to the description in the embodiment of the electronic device 100, which will not be described again here.
  • the communication unit 601 may be configured to receive information about uplink resource allocation from the network side device through a downlink.
  • the communication unit 601 may be configured to send parameters related to the local learning model to the network side device based on information about uplink resource allocation, where the local learning model is trained based on the initial global learning model issued by the network side device. of.
  • joint training includes aggregating local learning models related to edge links within the same group as an updated global learning model, thereby obtaining an aggregated learning model
  • the communication unit 601 may be configured to receive data from the network side device. Parameters about the aggregated learning model.
  • the network side device repeatedly performs partitioning and joint training until predetermined conditions are met.
  • the above learning model is used to assist in determining the data transmission rate of the edge link based on the data queue length and the channel energy gain of the edge link.
  • the learning model may be the deep reinforcement learning model involved in Figure 3.
  • the electronic device 600 is a device in a D2D scenario.
  • the electronic device 600 is a vehicle-mounted device in the Internet of Vehicles.
  • FIG. 7 shows a flowchart of a method S700 for wireless communication according to one embodiment of the present disclosure.
  • Method S700 begins at step S702.
  • step S704 based on the channel information about the channel status of at least one edge link of at least one user equipment reported by at least one user equipment located within the service range of the electronic device, the user equipment related to the at least one edge link is The learning model is divided into at least one group.
  • step S706 joint training is performed on the learning models in the same group for at least a part of the at least one group.
  • Method S700 ends at step S708.
  • This method can be performed, for example, by the electronic device 100 described above.
  • the electronic device 100 described above.
  • specific details please refer to the above description of related processing of the electronic device 100, which will not be repeated here.
  • FIG. 8 shows a flowchart of a method S800 for wireless communication according to one embodiment of the present disclosure.
  • Method S800 begins at step S802.
  • step S804 report the channel status of at least one edge link of the electronic device to the network side device that provides services for the electronic device.
  • Channel information for the network side device to divide the learning model of the electronic device related to at least one edge link into at least one group based on the channel information, thereby facilitating the network side device to target at least a part of the group in the same group.
  • the learning models within the group are jointly trained.
  • Method S800 ends at step S806.
  • This method can be performed, for example, by the electronic device 600 described above.
  • the electronic device 600 described above.
  • the technology of the present disclosure can be applied to a variety of products.
  • the electronic device 100 may be implemented as various network side devices such as a base station.
  • the base station may be implemented as any type of evolved Node B (eNB) or gNB (5G base station).
  • eNBs include, for example, macro eNBs and small eNBs.
  • a small eNB may be an eNB covering a smaller cell than a macro cell, such as a pico eNB, a micro eNB, and a home (femto) eNB.
  • gNB evolved Node B
  • the base station may be implemented as any other type of base station, such as NodeB and Base Transceiver Station (BTS).
  • BTS Base Transceiver Station
  • the base station may include: a main body (also referred to as a base station device) configured to control wireless communications; and one or more remote wireless heads (RRHs) disposed at a different place from the main body.
  • a main body also referred to as a base station device
  • RRHs remote wireless heads
  • various types of electronic devices may operate as base stations by temporarily or semi-persistently performing base station functions.
  • the Electronic device 600 may be implemented as various user devices.
  • the user equipment may be implemented as a mobile terminal such as a smartphone, a tablet personal computer (PC), a notebook PC, a portable game terminal, a portable/dongle-type mobile router, and a digital camera, or a vehicle-mounted terminal such as a car navigation device.
  • the user equipment may also be implemented as a terminal performing machine-to-machine (M2M) communication (also known as a machine type communication (MTC) terminal).
  • M2M machine-to-machine
  • MTC machine type communication
  • the user equipment may be a wireless communication module (such as an integrated circuit module including a single die) installed on each of the above-mentioned terminals.
  • eNB 800 includes one or more antennas 810 and base station equipment 820.
  • the base station device 820 and each antenna 810 may be connected to each other via an RF cable.
  • Each of the antennas 810 includes a single or multiple antenna elements (such as those included in a multi-input Multiple antenna elements in a multiple output (MIMO) antenna) and is used by the base station device 820 to transmit and receive wireless signals.
  • eNB 800 may include multiple antennas 810.
  • multiple antennas 810 may be compatible with multiple frequency bands used by eNB 800.
  • FIG. 9 shows an example in which eNB 800 includes multiple antennas 810, eNB 800 may also include a single antenna 810.
  • the base station device 820 includes a controller 821, a memory 822, a network interface 823, and a wireless communication interface 825.
  • the controller 821 may be, for example, a CPU or a DSP, and operates various functions of higher layers of the base station device 820 . For example, the controller 821 generates data packets based on the data in the signal processed by the wireless communication interface 825 and delivers the generated packets via the network interface 823 . The controller 821 may bundle data from multiple baseband processors to generate bundled packets, and deliver the generated bundled packets. The controller 821 may have logical functions to perform controls such as radio resource control, radio bearer control, mobility management, admission control, and scheduling. This control can be performed in conjunction with nearby eNBs or core network nodes.
  • the memory 822 includes RAM and ROM, and stores programs executed by the controller 821 and various types of control data such as terminal lists, transmission power data, and scheduling data.
  • the network interface 823 is a communication interface used to connect the base station device 820 to the core network 824. Controller 821 may communicate with core network nodes or additional eNBs via network interface 823. In this case, the eNB 800 and the core network node or other eNBs may be connected to each other through logical interfaces such as the S1 interface and the X2 interface.
  • the network interface 823 may also be a wired communication interface or a wireless communication interface for a wireless backhaul line. If the network interface 823 is a wireless communication interface, the network interface 823 may use a higher frequency band for wireless communication than the frequency band used by the wireless communication interface 825 .
  • the wireless communication interface 825 supports any cellular communication scheme, such as Long Term Evolution (LTE) and LTE-Advanced, and provides wireless connectivity to terminals located in the cell of the eNB 800 via the antenna 810 .
  • Wireless communication interface 825 may generally include, for example, a baseband (BB) processor 826 and RF circuitry 87.
  • the BB processor 826 may perform, for example, encoding/decoding, modulation/demodulation, and multiplexing/demultiplexing, and perform layer 1, medium access control (MAC), radio link control (RLC), and packet data aggregation protocols. (PDCP)) various types of signal processing.
  • the BB processor 826 may have some or all of the above-mentioned logical functions.
  • the BB processor 826 may be a memory that stores a communication control program, or a module including a processor and related circuitry configured to execute the program. Updater can enable BB processor 826 function changes.
  • the module may be a card or blade that plugs into a slot of the base station device 820. Alternatively, the module may be a chip mounted on a card or blade.
  • the RF circuit 87 may include, for example, a mixer, filter, and amplifier, and transmit and receive wireless signals via the antenna 810 .
  • the wireless communication interface 825 may include multiple BB processors 826 .
  • multiple BB processors 826 may be compatible with multiple frequency bands used by eNB 800.
  • wireless communication interface 825 may include a plurality of RF circuits 87 .
  • multiple RF circuits 87 may be compatible with multiple antenna elements.
  • FIG. 9 shows an example in which the wireless communication interface 825 includes multiple BB processors 826 and multiple RF circuits 87 , the wireless communication interface 825 may also include a single BB processor 826 or a single RF circuit 87 .
  • the electronic device 100 when the electronic device 100 is implemented as a base station, its transceiver can be implemented by the wireless communication interface 825. At least part of the functionality may also be implemented by controller 821.
  • the controller 821 may perform grouping and joint training by performing functions of units in the electronic device 100 .
  • eNB 830 includes one or more antennas 840, base station equipment 850, and RRH 860.
  • the RRH 860 and each antenna 840 may be connected to each other via RF cables.
  • the base station equipment 850 and the RRH 860 may be connected to each other via high-speed lines such as fiber optic cables.
  • Antennas 840 each include a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna) and are used by RRH 860 to transmit and receive wireless signals.
  • eNB 830 may include multiple antennas 840.
  • multiple antennas 840 may be compatible with multiple frequency bands used by eNB 830.
  • FIG. 10 shows an example in which eNB 830 includes multiple antennas 840, eNB 830 may also include a single antenna 840.
  • the base station device 850 includes a controller 851, a memory 852, a network interface 853, a wireless communication interface 855 and a connection interface 857.
  • the controller 851, the memory 852, and the network interface 853 are the same as the controller 821, the memory 822, and the network interface 823 described with reference to FIG. 9 .
  • the wireless communication interface 855 supports any cellular communication scheme such as LTE and LTE-Advanced and provides wireless communication to terminals located in the sector corresponding to the RRH 860 via the RRH 860 and the antenna 840 .
  • the wireless communication interface 855 may generally include a BB processor 856, for example.
  • the BB processor 856 is the same as the BB processor 826 described with reference to FIG. 9 , except that the BB processor 856 is connected to the RF circuit 864 of the RRH 860 via the connection interface 857 .
  • the wireless communication interface 855 may include multiple BB processors 856 .
  • multiple BB processors 856 may be compatible with multiple frequency bands used by eNB 830.
  • FIG. 10 shows an example in which the wireless communication interface 855 includes multiple BB processors 856, the wireless communication interface 855 may also include a single BB processor 856.
  • connection interface 857 is an interface for connecting the base station device 850 (wireless communication interface 855) to the RRH 860.
  • the connection interface 857 may also be a communication module for communication in the above-mentioned high-speed line that connects the base station device 850 (wireless communication interface 855) to the RRH 860.
  • RRH 860 includes a connection interface 861 and a wireless communication interface 863.
  • connection interface 861 is an interface for connecting the RRH 860 (wireless communication interface 863) to the base station device 850.
  • the connection interface 861 may also be a communication module used for communication in the above-mentioned high-speed line.
  • Wireless communication interface 863 transmits and receives wireless signals via antenna 840.
  • Wireless communication interface 863 may generally include RF circuitry 864, for example.
  • RF circuitry 864 may include, for example, mixers, filters, and amplifiers, and transmits and receives wireless signals via antenna 840 .
  • wireless communication interface 863 may include a plurality of RF circuits 864.
  • multiple RF circuits 864 may support multiple antenna elements.
  • FIG. 10 shows an example in which the wireless communication interface 863 includes a plurality of RF circuits 864, the wireless communication interface 863 may also include a single RF circuit 864.
  • the electronic device 100 when the electronic device 100 is implemented as a base station, its transceiver can be implemented by the wireless communication interface 855. At least part of the functionality may also be implemented by controller 851.
  • the controller 851 may perform grouping and joint training by performing functions of units in the electronic device 100 .
  • the smart phone 900 includes a processor 901, a memory 902, a storage device 903, an external connection interface 904, a camera 906, a sensor 907, a microphone 908, an input device 909, a display device 910, a speaker 911, a wireless communication interface 912, one or more Antenna switch 915, one or more antennas 916, bus 917, battery 918, and auxiliary controller 919.
  • the processor 901 may be, for example, a CPU or a system on a chip (SoC), and controls functions of the application layer and other layers of the smartphone 900 .
  • the memory 902 includes RAM and ROM, and stores data and programs executed by the processor 901 .
  • the storage device 903 may include storage media such as semiconductor memory and hard disk.
  • the external connection interface 904 is an interface for connecting external devices, such as memory cards and Universal Serial Bus (USB) devices, to the smartphone 900 .
  • the camera 906 includes an image sensor such as a charge coupled device (CCD) and a complementary metal oxide semiconductor (CMOS) and generates a captured image.
  • Sensors 907 may include a group of sensors such as measurement sensors, gyroscope sensors, geomagnetic sensors, and acceleration sensors.
  • the microphone 908 converts the sound input to the smartphone 900 into an audio signal.
  • the input device 909 includes, for example, a touch sensor, a keypad, a keyboard, a button, or a switch configured to detect a touch on the screen of the display device 910, and receives an operation or information input from a user.
  • the display device 910 includes a screen such as a liquid crystal display (LCD) and an organic light emitting diode (OLED) display, and displays an output image of the smartphone 900 .
  • the speaker 911 converts the audio signal output from the smartphone 900 into sound.
  • the wireless communication interface 912 supports any cellular communication scheme such as LTE and LTE-Advanced, and performs wireless communication.
  • the wireless communication interface 912 may generally include a BB processor 913 and an RF circuit 914, for example.
  • the BB processor 913 can perform, for example, encoding/decoding, modulation/demodulation, and multiplexing/demultiplexing, and perform various types of signal processing for wireless communication.
  • RF circuitry 914 may include, for example, mixers, filters, and amplifiers, and transmit and receive wireless signals via antenna 916 . Note that although the figure shows a situation where one RF link is connected to one antenna, this is only schematic, and it also includes a situation where one RF link is connected to multiple antennas through multiple phase shifters.
  • the wireless communication interface 912 may be a chip module on which the BB processor 913 and the RF circuit 914 are integrated. As shown in FIG. 11 , the wireless communication interface 912 may include multiple BB processors 913 and multiple RF circuits 914 . Although FIG. 11 shows an example in which the wireless communication interface 912 includes a plurality of BB processors 913 and a plurality of RF circuits 914, the wireless communication interface 912 may also include a single BB processor 913 or a single RF circuit 914.
  • the wireless communication interface 912 may support other types of wireless communication schemes, such as short-range wireless communication schemes, near field communication schemes, and wireless local area network (LAN) schemes.
  • the wireless communication interface 912 may include a BB processor 913 and an RF circuit 914 for each wireless communication scheme.
  • Each of the antenna switches 915 switches the connection destination of the antenna 916 between a plurality of circuits included in the wireless communication interface 912 (for example, circuits for different wireless communication schemes).
  • Antennas 916 each include a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna) and are used by wireless communication interface 912 to transmit and receive wireless signals.
  • smartphone 900 may include multiple antennas 916 .
  • FIG. 11 shows an example in which smartphone 900 includes multiple antennas 916
  • smartphone 900 may also include a single antenna 916 .
  • smartphone 900 may include an antenna 916 for each wireless communication scheme.
  • the antenna switch 915 may be omitted from the configuration of the smartphone 900 .
  • the bus 917 connects the processor 901, the memory 902, the storage device 903, the external connection interface 904, the camera 906, the sensor 907, the microphone 908, the input device 909, the display device 910, the speaker 911, the wireless communication interface 912 and the auxiliary controller 919 to each other. connect.
  • the battery 918 provides power to the various blocks of the smartphone 900 shown in Figure 11 via feeders, which are partially shown in the figure as dotted lines.
  • the auxiliary controller 919 operates the minimum necessary functions of the smartphone 900 in the sleep mode, for example.
  • the transceiver of the electronic device 600 may be implemented by the wireless communication interface 912 .
  • At least part of the functionality may also be implemented by the processor 901 or the auxiliary controller 919.
  • the processor 901 or the auxiliary controller 919 can report the channel information of the edge link by executing the functions of the above-mentioned units in the electronic device 600 .
  • FIG. 12 is a block diagram showing an example of a schematic configuration of a car navigation device 920 to which the technology of the present disclosure can be applied.
  • the car navigation device 920 includes a processor 921, a memory 922, a global positioning system (GPS) module 924, a sensor 925, a data interface 926, a content player 97, a storage media interface 928, an input device 99, a display device 930, a speaker 931, a wireless Communication interface 913, one or more antenna switches 936, one or more antennas 937, and battery 938.
  • GPS global positioning system
  • the processor 921 may be, for example, a CPU or an SoC, and controls the navigation function and other functions of the car navigation device 920 .
  • the memory 922 includes RAM and ROM, and stores data and programs executed by the processor 921 .
  • the GPS module 924 measures the location (such as latitude, longitude, and altitude) of the car navigation device 920 using GPS signals received from GPS satellites.
  • Sensors 925 may include a group of sensors such as gyroscope sensors, geomagnetic sensors, and air pressure sensors.
  • the data interface 926 is connected to, for example, the vehicle-mounted network 941 via a terminal not shown, and acquires data generated by the vehicle (such as vehicle speed data).
  • the content player 97 reproduces content stored in storage media such as CDs and DVDs, which are inserted into the storage media interface 928 .
  • the input device 99 includes, for example, a touch sensor, a button, or a switch configured to detect a touch on the screen of the display device 930, and receives an operation or information input from a user.
  • the display device 930 includes a screen such as an LCD or an OLED display, and displays an image of a navigation function or reproduced content.
  • the speaker 931 outputs the sound of the navigation function or the reproduced content.
  • the wireless communication interface 913 supports any cellular communication scheme such as LTE and LTE-Advanced, and performs wireless communication.
  • Wireless communication interface 913 may generally include, for example, BB processor 934 and RF circuitry 935.
  • the BB processor 934 can perform, for example, encoding/decoding, modulation/demodulation, and multiplexing/demultiplexing, and perform various types of signal processing for wireless communications.
  • the RF circuit 935 may include, for example, a mixer, filter, and amplifier, and transmit and receive wireless signals via the antenna 937 .
  • the wireless communication interface 913 may also be a chip module on which the BB processor 934 and the RF circuit 935 are integrated. As shown in FIG.
  • the wireless communication interface 913 may include multiple BB processors 934 and multiple RF circuits 935 .
  • FIG. 12 shows an example in which the wireless communication interface 913 includes a plurality of BB processors 934 and a plurality of RF circuits 935, the wireless communication interface 913 may also include a single BB processor 934 or a single RF circuit 935.
  • the wireless communication interface 913 may support other types of wireless communication schemes, such as short-range wireless communication schemes, near field communication schemes, and wireless LAN schemes.
  • the wireless communication interface 913 may include a BB processor 934 and an RF circuit 935 for each wireless communication scheme.
  • Each of the antenna switches 936 switches the connection destination of the antenna 937 between a plurality of circuits included in the wireless communication interface 913, such as circuits for different wireless communication schemes.
  • Antennas 937 each include a single or multiple antenna elements (such as multiple antenna elements included in a MIMO antenna) and are used by wireless communication interface 913 to transmit and receive wireless signals.
  • the car navigation device 920 may include a plurality of antennas 937 .
  • Figure 12 An example is shown in which the car navigation device 920 includes multiple antennas 937 , but the car navigation device 920 may also include a single antenna 937 .
  • the car navigation device 920 may include an antenna 937 for each wireless communication scheme.
  • the antenna switch 936 may be omitted from the configuration of the car navigation device 920.
  • the battery 938 provides power to the various blocks of the car navigation device 920 shown in FIG. 12 via feeders, which are partially shown as dashed lines in the figure. Battery 938 accumulates power provided from the vehicle.
  • the transceiver of the electronic device 600 may be implemented by the wireless communication interface 933 .
  • At least part of the functionality may also be implemented by processor 921.
  • the processor 921 may report the channel information of the edge link by executing the functions of the above-mentioned units in the electronic device 600 .
  • the technology of the present disclosure may also be implemented as an in-vehicle system (or vehicle) 940 including a car navigation device 920 , an in-vehicle network 941 , and one or more blocks of a vehicle module 942 .
  • vehicle module 942 generates vehicle data such as vehicle speed, engine speed, and fault information, and outputs the generated data to the in-vehicle network 941 .
  • the present invention also proposes a program product storing machine-readable instruction codes.
  • the instruction code is read and executed by the machine, the above method according to the embodiment of the present invention can be executed.
  • Storage media include but are not limited to floppy disks, optical disks, magneto-optical disks, memory cards, memory sticks, etc.
  • a program constituting the software is installed from a storage medium or a network to a computer having a dedicated hardware structure (such as the general-purpose computer 1300 shown in FIG. 13) in which various programs are installed. , can perform various functions, etc.
  • a central processing unit (CPU) 1301 performs various processes according to a program stored in a read-only memory (ROM) 1302 or a program loaded from a storage section 1308 into a random access memory (RAM) 1303 .
  • ROM read-only memory
  • RAM random access memory
  • data required when the CPU 1301 performs various processes and the like is also stored as necessary.
  • the CPU 1301, ROM 1302 and RAM 1303 are connected to each other via a bus 1304.
  • Input/output interface 1305 is also connected to bus 1304.
  • input section 1306 including keyboard, mouse, etc.
  • output section 1307 including display, such as cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.
  • Storage part 1308 including hard disk, etc.
  • communication part 1309 including network interface card such as LAN card, modem, etc.
  • the communication section 1309 performs communication processing via a network such as the Internet.
  • Driver 1310 may also be connected to input/output interface 1305 as needed.
  • Removable media 1311 such as magnetic disks, optical disks, magneto-optical disks, semiconductor memories, etc. are installed on the drive 1310 as needed, so that computer programs read therefrom are installed into the storage portion 1308 as needed.
  • the program constituting the software is installed from a network such as the Internet or a storage medium such as the removable medium 1311.
  • storage media are not limited to the removable media 1311 shown in FIG. 13 in which the program is stored and distributed separately from the device to provide the program to users.
  • the removable media 1311 include magnetic disks (including floppy disks (registered trademark)), optical disks (including compact disk read-only memory (CD-ROM) and digital versatile disks (DVD)), magneto-optical disks (including minidiscs (MD) (registered trademark)). Trademark)) and semiconductor memory.
  • the storage medium may be a ROM 1302, a hard disk contained in the storage section 1308, or the like, in which programs are stored and distributed to users together with the device containing them.
  • each component or each step can be decomposed and/or recombined.
  • These decompositions and/or recombinations should be regarded as equivalent versions of the present invention.
  • the steps for executing the above series of processes can naturally be executed in chronological order in the order described, but do not necessarily need to be executed in chronological order. Certain steps can be performed in parallel or independently of each other.
  • This technology can also be implemented as follows.
  • An electronic device for wireless communication including:
  • processing circuit configured as:
  • the user equipment related to the at least one edge link learning model is divided into at least one grouping, and
  • joint training is performed on learning models in the same group.
  • the channel information of the edge link includes the probability distribution of the channel energy gain of the edge link, reference signal received power RSRP, received signal strength index RSSI, reference signal quality RSRQ, signal-to-noise ratio SNR, information about the edge link At least one of the statistics of whether the user equipment as the sender and the user equipment as the receiver are within line-of-sight range, channel interference and noise.
  • the processing circuit is configured to perform the partitioning based on a degree of similarity between probability distributions respectively corresponding to the at least one edge link.
  • the degree of similarity includes the KL divergence between probability distributions.
  • Option 5 The electronic device according to any one of Options 2 to 4, wherein,
  • the channel energy gain is divided into a predetermined number of discrete bins, and the probability
  • the distribution includes the probability that the channel energy gain is at each level.
  • the processing circuit is configured to perform the partitioning based on the magnitude of the RSRP.
  • the processing circuit is configured to perform the partitioning based on the magnitude of the RSSI.
  • the processing circuit is configured to perform the partitioning based on the magnitude of the RSRQ.
  • the processing circuit is configured to perform the partitioning based on the magnitude of the SNR.
  • the processing circuit is configured to perform the division according to whether the user equipment as the sender and the user equipment as the receiver of the edge link are within line-of-sight range.
  • the processing circuit is configured to perform the partitioning based on magnitudes of statistics of interference and noise of the channel.
  • the statistics of interference and noise of the channel include mean and/or variance.
  • Item 13 The electronic device according to any one of Items 1 to 12, wherein,
  • the processing circuitry is configured to receive the channel information via Radio Resource Control RRC signaling.
  • Item 14 The electronic device according to any one of Items 1 to 13, wherein,
  • the processing circuit is configured to send the information about the partitioning to at least some of the user equipments related to the edge links in each group through a physical downlink control channel PDCCH.
  • Item 15 The electronic device according to Item 14, wherein,
  • the processing circuit is configured to send parameters related to the initial global learning model to the at least a portion of the user equipment in the first round of the joint training.
  • Item 16 The electronic device according to Item 15, wherein,
  • the processing circuit is configured to receive auxiliary status information for uplink resource allocation from the at least part of the user equipment through an uplink.
  • Item 17 The electronic device according to Item 16, wherein,
  • the auxiliary status information includes at least one of the number of samples used by the user equipment to train the learning model, location information of the user equipment, moving speed of the user equipment, computing power of the user equipment, and CPU occupancy of the user equipment.
  • Option 18 The electronic device according to Solution 16 or 17, wherein the processing circuit is configured to allocate uplink resources to the at least part of the user equipment based on the auxiliary status information.
  • Option 19 The electronic device according to Solution 18, wherein the processing circuit is configured to send information about uplink resource allocation to the at least part of the user equipment through a downlink.
  • Option 20 The electronic device according to Solution 18 or 19, wherein the processing circuit is configured to receive parameters related to the local learning model uploaded by the at least part of the user equipment based on the information about uplink resource allocation, Wherein, the local learning model is trained based on the initial global learning model issued by the electronic device.
  • Item 21 The electronic device according to item 20, wherein,
  • the joint training includes aggregating local learning models related to edge links in the same group, thereby obtaining the aggregated learning model as an updated global learning model, and
  • the processing circuit is configured to broadcast parameters related to the aggregated learning model of each group to the user equipments in the group.
  • Embodiment 22 The electronic device according to any one of Embodiments 1 to 21, wherein the processing circuit is configured to repeatedly perform the division and the joint training until a predetermined condition is met.
  • Item 23 The electronic device according to any one of Items 1 to 22, wherein,
  • the learning model is used to assist in determining the data transmission rate of the edge link based on the data queue length and the channel energy gain of the edge link.
  • Item 24 The electronic device according to any one of Items 1 to 23, wherein,
  • the at least one user equipment is a device in a D2D scenario.
  • An electronic device for wireless communication including:
  • processing circuit configured as:
  • Item 26 The electronic device according to item 25, wherein,
  • the channel information of the edge link includes the probability distribution of the channel energy gain of the edge link, reference signal received power RSRP, received signal strength index RSSI, reference signal quality RSRQ, signal-to-noise ratio SNR, information about the edge link At least one of the statistics of whether the user equipment as the sender and the user equipment as the receiver are within line-of-sight range, channel interference and noise.
  • Item 27 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the degree of similarity between probability distributions respectively corresponding to the at least one edge link.
  • Item 28 The electronic device according to item 27, wherein,
  • the degree of similarity includes the KL divergence between probability distributions.
  • Item 29 The electronic device according to any one of Items 26 to 28, wherein,
  • the channel energy gain is divided into a predetermined number of discrete bins, and the probability distribution includes a probability that the channel energy gain is at each bin.
  • Item 30 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the amplitude of the RSRP.
  • Item 31 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the amplitude of the RSSI.
  • Item 32 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the amplitude of the RSRQ.
  • Item 33 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the amplitude of the SNR.
  • Item 34 The electronic device according to Item 26, wherein,
  • the network side device performs the division according to whether the electronic device as the sender and the electronic device as the receiver of the edge link are within line-of-sight range.
  • Item 35 The electronic device according to Item 26, wherein,
  • the network side device performs the division based on the size of statistics of interference and noise of the channel.
  • Item 36 The electronic device according to Item 35, wherein,
  • the statistics of interference and noise of the channel include mean and/or variance.
  • Item 37 The electronic device according to any one of Items 25 to 36, wherein,
  • the processing circuit is configured to report the channel information via Radio Resource Control (RRC) signaling.
  • RRC Radio Resource Control
  • Item 38 The electronic device according to any one of Items 25 to 37, wherein,
  • the processing circuit is configured to receive information about the division from the network side device through a physical downlink control channel (PDCCH).
  • PDCH physical downlink control channel
  • Item 39 The electronic device according to Item 38, wherein,
  • the processing circuit is configured to receive parameters regarding an initial global learning model in a first round of joint training.
  • Item 40 The electronic device according to Item 39, wherein,
  • the processing circuit is configured to send auxiliary status information for uplink resource allocation to the network side device through an uplink.
  • the auxiliary status information includes the number of samples used by the electronic device to train the learning model. At least one of the quantity, the location information of the electronic device, the moving speed of the electronic device, the computing power of the electronic device, and the CPU occupancy rate of the electronic device.
  • Option 42 The electronic device according to Solution 40 or 41, wherein the processing circuit is configured to receive information about uplink resource allocation from the network side device through a downlink.
  • Option 43 The electronic device according to Solution 42, wherein the processing circuit is configured to send parameters related to the local learning model to the network side device based on the information about uplink resource allocation, wherein the local The learning model is trained based on the initial global learning model delivered by the network side device.
  • Item 44 The electronic device according to Item 43, wherein,
  • the joint training includes aggregating local learning models related to edge links within the same group as an updated global learning model, thereby obtaining an aggregated learning model, and
  • the processing circuit is configured to receive parameters related to the aggregated learning model from the network side device.
  • Solution 45 The electronic device according to any one of solutions 25 to 44, wherein the network side device repeatedly performs the division and the joint training until a predetermined condition is met.
  • Item 46 The electronic device according to any one of Items 25 to 45, wherein,
  • the learning model is used to assist in determining the data transmission rate of the edge link based on the data queue length and the channel energy gain of the edge link.
  • Item 47 The electronic device according to any one of Items 25 to 46, wherein,
  • the electronic device is a device in a D2D scenario.
  • a method for wireless communication comprising:
  • the learning of the user equipment related to the at least one edge link is the model is divided into at least one grouping, and
  • a method for wireless communication comprising:
  • Learning models of related electronic devices and other electronic devices related to the at least one edge link and provided by the network side device are divided into at least one group, thereby facilitating the network side device to target at least one of the at least one group. Part of the group is used to jointly train the learning models in the same group.
  • Option 50 A computer-readable storage medium having computer-executable instructions stored thereon. When the computer-executable instructions are executed, the method for wireless communication according to Item 48 or 49 is performed.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请涉及用于无线通信的电子设备和方法、计算机可读存储介质。其中，用于无线通信的电子设备包括处理电路，处理电路被配置为：基于位于电子设备服务范围内的至少一个用户设备上报的、有关至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。 (图1)

Description

用于无线通信的电子设备和方法、计算机可读存储介质
本申请要求于2022年7月11日提交中国专利局、申请号为202210809772.4、发明名称为“用于无线通信的电子设备和方法、计算机可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本公开涉及无线通信技术领域，具体地涉及一种用于无线通信的电子设备和方法以及计算机可读存储介质。更具体地，涉及对与边缘链路相关的用户设备的学习模型进行分组，并且对处于同一分组内的学习模型进行联合训练。
背景技术
随着无线网络和人工智能的发展，网络逐渐趋于智能化。特别是对于未来的6G，无线网络智能化将是其发展的重要方向。更具体地看，联邦学习(Federated Learning,FL)作为当前最重要的分布式人工智能框架，其与无线网络的结合将是未来无线网络智能化应用的主要内容之一。因此，如何将FL与当前的5G NR进行有效的联合设计，将会对未来的人工智能应用产生重要影响。尤其是，在高度智能化的无线网络中，如何根据无线通信的特点，高效地使用FL对智能无线网络中的机器学习模型进行联合训练将受到越来越广泛的关注。
在无线网络演进过程中，诸多机器学习模型可被用于优化网络的决策与运行。例如，车联网中车与车(V2V)之间的通信由边缘链路(sidelink)实现，多数情况下可建模为马尔科夫决策过程(Markov Decision Process,MDP)，使用深度强化学习(Deep Reinforcement Learning,DRP)来解决。
如何高效地使用FL针对边缘链路进行联合训练，是目前研究的热点。
发明内容
在下文中给出了关于本发明的简要概述，以便提供关于本发明的某些方面的基本理解。应当理解，这个概述并不是关于本发明的穷举性概述。它并不是意图确定本发明的关键或重要部分，也不是意图限定本发明的范围。其目的仅仅是以简化的形式给出某些概念，以此作为稍后论述的更详细描述的前序。
根据本公开的一个方面，提供了一种用于无线通信的电子设备，其包括处理电路，处理电路被配置为：基于位于电子设备服务范围内的至少一个用户设备上报的、有关至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
在根据本公开的实施例中，电子设备通过分组解决边缘链路因环境不同导致的数据异构性问题，能够提高联合训练的效率，并且能够提高学习模型的质量和系统的性能。
根据本公开的一个方面，提供了一种用于无线通信的电子设备，其包括处理电路，处理电路被配置为：向为电子设备提供服务的网络侧设备上报有关电子设备的至少一个边缘链路的信道状态的信道信息，以供网络侧设备基于信道信息将与至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于网络侧设备针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
在根据本公开的实施例中，电子设备向网络侧设备上报有关边缘链路的信道状态的信道信息，以供网络侧设备基于信道信息来对与边缘链路相关的电子设备的学习模型进行分组，有助于网络侧设备通过分组解决边缘链路因环境不同导致的数据异构性问题，能够提高联合训练的效率，并且能够提高学习模型的质量和系统的性能。
根据本公开的一个方面，提供了一种用于无线通信的方法，包括：基于位于电子设备服务范围内的至少一个用户设备上报的、有关至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及针对至 少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
根据本公开的一个方面，提供了一种用于无线通信的方法，包括：向为电子设备提供服务的网络侧设备上报有关电子设备的至少一个边缘链路的信道状态的信道信息，以供网络侧设备基于信道信息将与至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于网络侧设备针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
依据本发明的其它方面，还提供了用于实现上述用于无线通信的方法的计算机程序代码和计算机程序产品以及其上记录有该用于实现上述用于无线通信的方法的计算机程序代码的计算机可读存储介质。
附图说明
为了进一步阐述本发明的以上和其它优点和特征，下面结合附图对本发明的具体实施方式作进一步详细的说明。所述附图连同下面的详细说明一起包含在本说明书中并且形成本说明书的一部分。具有相同的功能和结构的元件用相同的参考标号表示。应当理解，这些附图仅描述本发明的典型示例，而不应看作是对本发明的范围的限定。在附图中：
图1示出了根据本公开的一个实施例的用于无线通信的电子设备的功能模块框图；
图2是示出根据本公开实施例的系统结构的示意图；
图3是示出根据本公开实施例的边缘链路功率速率自适应控制场景的示意图；
图4a和4b是示出根据本公开实施例的基于边缘链路的信道能量增益的概率分布之间的相似程度进行划分的示意图；
图5是示出根据本公开实施例的电子设备与用户设备之间的信息交互的示例图；
图6示出了根据本公开另一实施例的用于无线通信的电子设备的功能模块框图；
图7示出了根据本公开的一个实施例的用于无线通信的方法的流程图；
图8示出了根据本公开的另一实施例的用于无线通信的方法的流程图；
图9是示出可以应用本公开内容的技术的eNB或gNB的示意性配置的第一示例的框图；
图10是示出可以应用本公开内容的技术的eNB或gNB的示意性配置的第二示例的框图；
图11是示出可以应用本公开内容的技术的智能电话的示意性配置的示例的框图；
图12是示出可以应用本公开内容的技术的汽车导航设备的示意性配置的示例的框图；以及
图13是其中可以实现根据本发明的实施例的方法和/或装置和/或系统的通用个人计算机的示例性结构的框图。
具体实施方式
在下文中将结合附图对本发明的示范性实施例进行描述。为了清楚和简明起见，在说明书中并未描述实际实施方式的所有特征。然而，应该了解，在开发任何这种实际实施例的过程中必须做出很多特定于实施方式的决定，以便实现开发人员的具体目标，例如，符合与系统及业务相关的那些限制条件，并且这些限制条件可能会随着实施方式的不同而有所改变。此外，还应该了解，虽然开发工作有可能是非常复杂和费时的，但对得益于本公开内容的本领域技术人员来说，这种开发工作仅仅是例行的任务。
在此，还需要说明的一点是，为了避免因不必要的细节而模糊了本发明，在附图中仅仅示出了与根据本发明的方案密切相关的设备结构和/或处理步骤，而省略了与本发明关系不大的其他细节。
图1示出了根据本公开的一个实施例的用于无线通信的电子设备100的功能模块框图。
如图1所示，电子设备100包括：处理单元101，其可以基于位于电子设备100服务范围内的至少一个用户设备上报的、有关至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组；以及训练单元103，其可以针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
其中，处理单元101和训练单元103可以由一个或多个处理电路实现，该处理电路例如可以实现为芯片。
电子设备100可以作为无线通信系统中的网络侧设备，具体地例如可以设置在基站侧或者可通信地连接到基站。这里，还应指出，电子设备100可以以芯片级来实现，或者也可以以设备级来实现。例如，电子设备100可以工作为基站本身，并且还可以包括诸如存储器、收发器(未示出)等外部设备。存储器可以用于存储基站实现各种功能需要执行的程序和相关数据信息。收发器可以包括一个或多个通信接口以支持与不同设备(例如，用户设备(UE)、其他基站等等)间的通信，这里不具体限制收发器的实现形式。
作为示例，基站例如可以是eNB或gNB。
例如，电子设备100可以与核心网相连。
根据本公开的无线通信系统可以是5G NR(New Radio，新空口)通信系统。进一步，根据本公开的无线通信系统可以包括非地面网络(Non-terrestrial network，NTN)。可选地，根据本公开的无线通信系统还可以包括地面网络(Terrestrial network，TN)。另外，本领域技术人员可以理解，根据本公开的无线通信系统还可以是4G或3G通信系统。
例如，用户设备可以是边缘链路(SL,Sidelink)上用于发送的用户设备(简称为发送用户设备)也可以是边缘链路上用于接收的用户设备(简称为接收用户设备)。用户设备能够进行边缘链路控制。
在根据本公开实施例的电子设备100中，联邦学习用于多用户设备下的学习模型的联合训练。
作为示例，学习模型可以是传统机器学习模型或深度强化学习模型。在下文中，为了方便，有时以学习模型是深度强化学习模型为例来进行 说明。
图2是示出根据本公开实施例的系统结构的示意图。
如图2所示，为了简便，将用户设备示出为车辆，本领域技术人员可以理解，用户设备可以是除了车辆之外的其他形式，例如，用户设备可以是手机、iPad、笔记本等终端设备，只要用户设备之间存在边缘链路即可。单个用户设备针对与其边缘链路相关的学习模型(可以称为本地模型)进行强化学习，例如，本地模型是基于电子设备100下发的初始全局模型而得到的。电子设备100基于有关用户设备的边缘链路的信道状态的信道信息，将用户设备上的本地模型划分成不同的分组(例如，为了简单，图2中仅示出了被划分为同一个分组的三个用户设备UE1、UE2和UE3)。用户设备将其本地模型的参数上传至电子设备100；对于处于同一分组内的用户设备UE1、UE2和UE3，电子设备100通过联邦学习进行对学习模型的联合训练(对学习模型进行聚合)，以形成全局模型。
对于现有技术中的不使用联邦学习进行学习模型的联合训练而言，用于训练学习的样本数量通常不足，使得学习模型的训练困难。相比不使用联邦学习进行学习模型的联合训练，考虑使用联邦学习对多个学习模型进行联合训练，能够克服单个强化学习训练样本不足，收敛速度慢的问题。
在现有技术的联邦学习中，不同用户设备由于所处环境不同而导致数据的异构性问题。由于数据的异构性，聚合过程中增加了不相关的样本，使得训练过程中学习模型发散或者收敛速度变慢，进而降低了学习模型的质量和系统的性能。也就是说，随机信道环境的差异使得不同用户设备之间收集的数据具有异构性，导致不同边缘链路训练的学习模型具有不同程度的差异性。该情况在深度强化学习中表现的尤为严重。这种差异性将导致现有技术中的基于联邦学习的整体训练性能下降。例如，在车联网(V2X)中，该问题更加严重。主要由以下两个原因：(1)由于车辆的移动行，V2X中的设备面临更加复杂多变多样的环境，数据的异构性加剧；(2)V2X中的训练多处需要用到强化学习模型，强化学习的应用需要设备不断的从环境中获得回报，这加剧了多样的环境对系统性能的影响。
而在根据本公开的实施例中，电子设备100通过分组解决边缘链路因环境不同导致的数据异构性问题，能够提高联合训练的效率，并且能够提高学习模型的质量和系统的性能。
作为示例，至少一个用户设备是D2D场景下的设备。例如，至少一个用户设备是车联网中的车载设备。然而，用户设备不只局限于V2X车联网场景，任何由sidelink链接的通信场景均可适用。例如，在D2D场景下，用户设备之间的通信还可以是终端设备(手机，平版电脑等)之间的相互通信。用户设备之间的通信也可是XR(扩展现实)场景下，XR设备之间的通信；以及用户设备之间的通信还可以是工业互联网或智能家居等场景下设备之间的通信等。
在下文中，为了方便，以用户设备是车联网中的车辆或车载设备为例进行描述。本领域技术人员可以理解，用户设备可以是除了车载设备之外的其他形式，只要用户设备之间存在边缘链路即可。
在V2X中，多数问题可建模为马尔可夫决策(Markov Decision Problem,MDP)问题。因此可以使用深度强化学习来解决。一个具体例子是边缘链路信息发送的功率速率自适应控制。车辆设备的数据发送需要消耗电池的电量，而电池容量是有限的。因此为了有更长的续航时间，在边缘链路的数据发送过程中，需要给定发送设备的平均功率约束。然而，由于车辆之间的无线信道状态是随机的，如果只追求信息传输的低能耗，等待信道状况好的时候进行传输(机会传输)，此时数据包会在队列里等待很久，导致严重的数据排队时延。因此，需要通过高效的自适应功率控制达到最优的时延功率的折中关系，从而在保证满足平均功耗要求的情况下最小化数据的传输时延。这样一个功率控制问题可以建模为MDP问题，使用深度强化学习来解决。
作为示例，边缘链路的信道信息包括边缘链路的信道能量增益的概率分布、参考信号接收功率(RSRP)、接收信号的强度指标(RSSI)、参考信号质量(RSRQ)、信噪比(SNR)、有关边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
边缘链路的信道信息用于度量与边缘链路相关的学习模型之间的相似程度。电子设备100基于边缘链路的信道信息来对与边缘链路相关的 用户设备的学习模型进行分组，能够将相似程度高的学习模型进行联合训练。
作为示例，处理单元101可以被配置为基于与至少一个边缘链路分别对应的概率分布之间的相似程度，进行划分。
作为示例，信道能量增益被划分为预定数量的离散档位，以及概率分布包括信道能量增益处于每个档位的概率。
在无线网络演进过程中，诸多机器学习模型可被用于优化网络的决策与运行。如上所述，一个示例是数据发送过程中的功率控制问题：发送端车辆根据当前的车辆间信道状态以及数据队列状态去实时地自适应地调节数据的发送功率与发送的数据包数目。由于车辆之间的信道状态是随机的，如果只追求信息传输的实时性(也就是即时传输)，则遇到恶劣信道状态时能耗成本会非常高。如果只追求信息传输的低能耗，等待信道状况好的时候进行传输(也就是机会传输)，此时数据包会在队列里等待很久，导致严重的数据排队时延。因此，需要通过高效的自适应功率控制达到最优的时延功率的折中关系，从而在保证满足平均功耗要求的情况下最小化数据的传输时延。
为了说明联邦学习分组的依据，下文中以边缘链路的功率速率自适应控制为例做介绍。例如，学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定边缘链路的数据发送速率。
以用户设备为车辆为例，如上所述，多个边缘链路间可以通过联邦学习来联合训练学习模型。但是，由于不同车辆之间的异构性型，以及学习模型(特别是深度强化学习(DRL)模型)对环境的敏感性，不同随机环境下训练的DRL模型具有一定差异性。在边缘链路的通信问题中，无线信道的概率分布特征就是训练的随机环境。在根据本公开的实施例中，通过选择无线信道状态概率分布特征(例如，信道能量增益的概率分布)相似的用户设备进行FL的分组，使得选择了随机环境相似的车辆进行联邦学习的联合训练，能够提高联合训练效率。
图3是示出根据本公开实施例的边缘链路功率速率自适应控制场景的示意图。在图3中，Tx表示发送，Rx表示接收。
结合图3，用户设备需要根据当前的数据队列长度q(队列中等待发 送的数据包数量)，以及当前的信道能量增益的档位h(信道能量增益被划分为w个离散档位)确定当前的数据发送速率s以及当前的数据发送功率P(发送功率P由发送速率s与信道能量增益h决定，可以通过信道容量公式计算)，从而保证在有限的平均功耗的限制下最小化边缘链路中数据发送的平均排队延时。其中，边缘链路的信道能量增益在每个发送时隙是独立同分布的，令Pi＝[pi(h1)，pi(h2)，...，pi(hw)]表示第i个边缘链路上的信道能量增益服从的概率分布，pi(hk)表示第i个边缘链路上的信道能量增益处于第k档的概率，其中，1≤k≤w。如何根据实时动态变化的队列长度以及信道能量增益，实时地选择最优的发送功率与发送速率可以建立为马尔可夫决策问题。对于马尔可夫决策问题可以进一步使用深度强化学习模型来解决。具体来说，深度强化学习模型通过人工神经网络拟合强化学习过程中的值函数。在图3中，人工神经网络的输入为队列长度q，信道增益所处的档位h，以及发送速率s；输出为状态(q,h,s)对应的值函数V(q,h,s)。根据人工神经网络所提供的值函数V(q,h,s)，用户设备可以得到队列长度为q、信道增益所处档位为h情况下，最优的发送速率S*,其中S*＝argminsV(q，h，s)。进一步来说，在图3中，不同的信道能量增益的概率分布下训练得出的人工神经网络模型会不同。
根据联邦学习的特点可知，将信道能量增益的概率分布相似的边缘链路加入同一个联邦学习的分组进行训练能够提高聚合得到的全局模型的准确度。
图4a和4b是示出根据本公开实施例的基于边缘链路的信道能量增益的概率分布之间的相似程度进行划分的示意。
在图4a中，假设边缘链路1上的信道能量增益的概率分布P1与边缘链路2上的信道能量增益的概率分布P2相似，将与边缘链路1相关的学习模型和与边缘链路1相关的学习模型划分成同一分组，能够提升联合训练的效果。
在图4b中，假设边缘链路1上的信道能量增益的概率分布P1与边缘链路3上的信道能量增益的概率分布P3不相似，如果将与边缘链路1相关的学习模型和与边缘链路3相关的学习模型划分成同一分组，则不利于提升联合训练的效果。
作为示例，概率分布之间的相似程度包括概率分布之间的KL散度。
对于一个离散随机变量的两个概率分布Pi和Pj来讲,即Pi＝[pi(h1)，pi(h2)，...，pi(hw)]和Pj＝[pj(h1)，pj(h2)，...，pj(hw)],KL散度定义为：

由于KL散度的不对称性，首先针对每对KL散度，取每对KL散度中的最大值Dij＝max{DKL(Pi||Pj)，DKL(Pj||Pi)}。即，针对式1和式2所表示的每对KL散度，假设取最大值表示为Dij。
随后，从针对每对KL散度取出的最大值当中，选出最小的KL散度。然后以此为基准选取KL散度最小的用户设备为一组。
比如有4个用户设备6条边缘链路，其中选取3个用户设备分为一组。对应的KL散度假设为：
DKL(P1||P2)＝0.5；DKL(P2||P1)＝0.6；
DKL(P1||P3)＝0.7；DKL(P3||P1)＝0.6；
DKL(P1||P4)＝0.5；DKL(P4||P1)＝0.4；
DKL(P2||P3)＝0.4；DKL(P2||P3)＝0.3；
DKL(P2||P4)＝1.0；DKL(P4||P2)＝0.8；
DKL(P3||P4)＝0.3；DKL(P4||P3)＝0.2；
首先，针对每对KL散度取出的最大值分别为：D12＝0.6；D13＝0.7；D14＝0.5；D23＝0.4；D24＝1.0；D34＝0.3。
随后，从上面6个最大值当中选取出最小值D34，其对应DKL(P3||P4)。
最后，选取DKL(P1||P4)＝0.5，DKL(P2||P4)＝0.4和DKL(P3||P4)中最小的两个值对应的用户设备，即用户设备2和用户设备3与用户4设备为一组。
除了上述示例外，本领域技术人员还可以想到基于KL散度进行分组的其他方式，这里不再累述。
其中，KL散度越小，表明两个概率分布越相近，相似程度高的信道状态概率分布表示相似程度高的强化学习模型，将相似程度高的强化学习模型划分成一组进行联合训练，能达到更好的联合训练效果。
除了KL散度之外，本领域技术人员还可以想到概率分布之间的相似程度的其他示例，这里不再累述。
作为示例，处理单元101可以被配置为基于RSRP的幅值来进行划分。例如，可以将RSRP的幅值大于预定阈值的用户设备的学习模型划分为同一分组，而将RSRP的幅值小于等于预定阈值的用户设备的学习模型划分为另一分组。或者，基于RSRP的幅值，将用户设备的学习模 型划分为多个分组。
作为示例，处理单元101可以被配置为基于RSSI的幅值来进行划分。例如，可以将RSSI的幅值大于预定阈值的用户设备的学习模型划分为同一分组，而将RSSI的幅值小于等于预定阈值的用户设备的学习模型划分为另一分组。或者，基于RSSI的幅值，将用户设备的学习模型划分为多个分组。
作为示例，处理单元101可以被配置为基于RSRQ的幅值来进行划分。例如，可以将RSRQ的幅值大于预定阈值的用户设备的学习模型划分为同一分组，而将RSRQ的幅值小于等于预定阈值的用户设备的学习模型划分为另一分组。或者，基于RSRQ的幅值，将用户设备的学习模型划分为多个分组。
作为示例，处理单元101可以被配置为基于SNR的幅值来进行划分。例如，可以将SNR的幅值大于预定阈值的用户设备的学习模型划分为同一分组，而将SNR的幅值小于等于预定阈值的用户设备的学习模型划分为另一分组。或者，基于SNR的幅值，将用户设备的学习模型划分为多个分组。
作为示例，处理单元101可以被配置为根据边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距(LOS)范围内，进行划分。例如，如果要对若干个发送用户设备进行联合训练，则将满足其与对应的接收用户设备处于视距范围的发送用户设备划分成同一分组，而将其与对应的接收用户设备处于非视距范围(NLOS)的发送用户设备划分成另一分组。
作为示例，处理单元101可以被配置为基于信道的干扰和噪声的统计量的大小来进行划分。
作为示例，信道的干扰和噪声的统计量包括均值和/或方差。
上文中描述了在边缘链路的信道信息包括边缘链路的信道能量增益的概率分布、RSRP、RSSI、RSRQ、SNR、有关边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中的一个指标的情况下，如何对学习模型进行分组。
下面描述在边缘链路的信道信息包括边缘链路的信道能量增益的概率分布、RSRP、RSSI、RSRQ、SNR、有关边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中的至少两个指标的情况下，如何对学习模型进行分组。
例如，可以基于所述至少两个指标的优先级，来对学习模型进行分组。例如，可以根据经验或者应用场景等设置指标的优先级。
例如，在所述至少两个指标包括两个指标的情况下，先基于第一优先级的指标来对学习模型进行分组得到第一次分组结果；然后在第一次分组结果的基础上，基于第二优先级的指标，再次进行分组，从而得到最终分组结果。例如，假设两个指标包括第一指标RSRP和第二指标边缘链路的信道能量增益的概率分布，并且第一指标的优先级高于第二指标的优先级，则可以先基于RSRP的幅值来对学习模型进行分组(例如，可以将RSRP的幅值大于预定阈值的用户设备的学习模型划分为第一分组，而将RSRP的幅值小于等于预定阈值的用户设备的学习模型划分为第二分组)得到第一次分组结果(例如，其包括第一分组和第二分组)；然后基于边缘链路的信道能量增益的概率分布之间的相似程度，分别对第一分组和第二分组进行分组(例如，将第一分组中的相似程度高的学习模型划分成第一子分组，而将第一子分组中其他学习模型划分成第二子分组；将第二分组中的相似程度高的学习模型划分成第三子分组，而将第二分组中其他学习模型划分成第四子分组)，从而得到最终分组(例如，其包括第一子分组、第二子分组、第三子分组、以及第四子分组)。
例如，在所述至少两个指标包括三个指标的情况下，先基于优先级最高的指标来对学习模型进行分组得到第一次分组结果，然后在第一次分组结果的基础上，基于第二优先级的指标，再次进行分组，从而得到第二次分组结果；最后，在第二次分组结果的基础上，基于第三优先级的指标，再进行分组，从而得到最终分组结果。以此类推，可以对于在所述至少两个指标包括四个指标之上的情况下进行分组，这里不再累述。
作为示例，处理单元101可以被配置为经由无线资源控制(RRC)信令接收边缘链路的信道信息。
例如，RRC信令可以是MeasResultsSL信令。
例如，如上所示，可以在MeasResultsSL信令中增加信令“measResultListPDCS-NR”来传输信道状态的概率分布Pi＝[pi(h1)，pi(h2)，...，pi(hw)]，其中，pi(hk)表示第i个边缘链路上的信道能量增益处于第k档的概率。或者，可以在MeasResultsSL信令中增加信令“measResultListOther-NR”来传输边缘链路的其他信道信息如RSRP、RSRQ、RSSI、SNR、LOS、信道的干扰和噪声的统计量等。或者，可以在MeasResultsSL信令中增加信令“measResultListPDCS-NR”和“measResultListOther-NR”两者。
MeasResultsSL的详细内容在标准请参见“TS 38.131 Radio Resource Control(RRC)protocol specification”，这里不再累述。
作为示例，处理单元101可以被配置为通过物理下行控制信道(PDCCH)将有关划分的信息发送给与每个分组中的边缘链路有关的用户设备中的至少一部分用户设备。
作为示例，处理单元101可以被配置为在联合训练的第一轮中，将有关初始全局学习模型的参数发送给至少一部分用户设备。这样，用户设备可以基于初始全局模型进行本地训练，得到本地模型。
作为示例，处理单元101可以被配置为通过上行链路从至少一部分用户设备接收用于进行上行资源分配的辅助状态信息。
在同一个联邦学习的分组内，用户设备本身的状态存在较大的差异。因此，分组完成后用户设备向电子设备100上传辅助状态信息，用于上 行资源分配。
作为示例，辅助状态信息包括用户设备训练学习模型所用的样本数量、用户设备的位置信息、用户设备的移动速度、用户设备的计算能力、用户设备的CPU占有率中的至少之一。
作为示例，处理单元101可以被配置为基于辅助状态信息，对至少一部分用户设备进行上行资源分配。即，电子设备100获取可用的无线资源块信息，为联邦学习的本地模型上传做准备。
电子设备100根据分组内用户设备上传的辅助状态信息进行相应的上行资源分配，能够解决或缓解Straggler问题，从而加快FL的进程，提升系统性能。
Straggler的主要表现为用户设备具有：1)大量数据：在聚合过程中具有较大的权值；2)较高的优先级；3)较差的计算能力或者较高的CPU占用率；以及4)、距离电子设备100较远或者信道质量较差，导致较长的传输时间或者较低的传输速率。
电子设备100通过分配更多的频率资源给具有Straggler问题的用户设备，可以减少传输时延，加快学习模型收敛过程。
作为示例，用户设备训练学习模型所用的样本数量为联邦学习迭代训练过程中，每一轮迭代中用户设备用于训练本地模型的样本的数量。根据用户设备用于训练本地模型的样本的数量信息，判断用户设备训练的本地模型的重要程度，用于确定联邦学习过程中上行无线资源分配。
作为示例，用于本地模型训练的用户设备的计算能力为CPU的计算速度，以及用户设备的CPU占有率为本地模型训练过程的CPU占用率。例如，用于本地模型训练的CPU占用率信息用于估计联邦学习过程中用户设备的计算能力，从而判断Straggler。
相比不考虑用户设备的位置信息(即，不使用用户设备与电子设备100之间的距离信息)以及不考虑训练本地模型的CPU使用的情况，在本公开中考虑此类信息从而判断联邦学习过程中的Straggler，能够减少一轮迭代所需的时间。
作为示例，处理单元101可以被配置为通过下行链路将有关上行资源分配的信息发送给至少一部分用户设备。
作为示例，训练单元103可以被配置为接收至少一部分用户设备基于有关上行资源分配的信息所上传的、有关本地学习模型的参数，其中，本地学习模型是基于电子设备100下发的初始全局学习模型而训练得到的。
作为示例，联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合作为更新的全局学习模型，从而得到聚合后的学习模型。例如，聚合是对与处于同一分组内的边缘链路相关的本地学习模型的参数进行加权平均。例如，电子设备100基站根据每个边缘链路训练本地强化学习模型时使用的训练样本数量信息来判断上传的强化学习模型的重要程度，用于确定聚合时本地模型的加权系数，从而在本轮迭代过程中聚合出更准确的全局模型。即，电子设备100可以根据用户设备用于训练本地模型的样本的数量信息，确定本地模型的权重，从而最小化联邦学习训练出的全局模型的误差。
作为示例，训练单元103可以被配置为将每个分组的有关聚合后的学习模型(也可称为更新后的全局模型)的参数广播给该分组中的用户设备。
作为示例，训练单元103可以被配置为重复地进行划分和联合训练，直到满足预定条件为止。例如，预定条件是达到预定的迭代次数，或者聚合后的学习模型的误差小于预定误差，等等。
图5是示出根据本公开实施例的电子设备100与用户设备UE之间的信息交互的示例图。在图5中，以边缘链路的信道信息为边缘链路的信道能量增益的概率分布为例来进行说明。
UE收集信道能量增益的概率分布Pi＝[pi(h1)，pi(h2)，...，pi(hw)]。在S51中，UE将所收集的概率分布上传给电子设备100。
电子设备100基于边缘链路的信道能量增益的概率分布来确定用户设备的学习模型的相似度，并将具有较高相似度的学习模型划分成同一分组。在S52中，电子设备100将分组信息和初始全局模型下发给UE。
在S53中，UE将辅助状态信息上传到电子设备100。
电子设备100基于辅助状态信息，对用户设备进行上行资源分配。例如，电子设备100找到具有Straggler问题的用户设备并且为这样的用户设备分配更多的资源。
在S54中，电子设备100将上行资源分配信息下发给UE。
UE根据本地的样本数据，基于初始全局模型进行本地训练，得到本地模型。
在S55中，UE通过被分配的上行资源，将本地模型的参数上传到电子设备100。
电子设备100对同一分组内的用户设备的本地模型进行聚合，以得到更新后的全局模型。更新后的全局模型表示联邦学习在本轮训练下的最终模型，其模型误差代表着本轮联邦学习训练的效果。
电子设备100将更新后的全局模型广播给所有参与联邦学习的用户设备。
联邦学习的训练需要UE和电子设备100进行学习模型的若干轮次的迭代与聚合，即，重复地进行划分和联合训练，即，重复地执行S51-S55的处理，直到满足预定条件为止。
本公开还提供了一种根据另一实施例的用于无线通信的电子设备。图6示出了根据本公开又一个实施例的用于无线通信的电子设备600的功能模块框图。
如图6所示，电子设备600包括：通信单元601，通信单元601可以向为电子设备600提供服务的网络侧设备上报有关电子设备600的至少一个边缘链路的信道状态的信道信息，以供网络侧设备基于信道信息将与至少一个边缘链路相关的电子设备600以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于网络侧设备针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
其中，通信单元601可以由一个或多个处理电路实现，该处理电路例如可以实现为芯片。
电子设备600例如可以设置在用户设备(UE)侧或者可通信地连接到用户设备。在电子设备600设置在用户设备侧或者可通信地连接到用 户设备的情况下，与电子设备600有关的装置可以是用户设备。这里，还应指出，电子设备600可以以芯片级来实现，或者也可以以设备级来实现。例如，电子设备600可以工作为用户设备本身，并且还可以包括诸如存储器、收发器(图中未示出)等外部设备。存储器可以用于存储用户设备实现各种功能需要执行的程序和相关数据信息。收发器可以包括一个或多个通信接口以支持与不同设备(例如，基站、其他用户设备等等)间的通信，这里不具体限制收发器的实现形式。
作为示例，网络侧设备可以是上文中提到的电子设备100。作为示例，电子设备600可以是上文电子设备100实施例中涉及的用户设备。
根据本公开的无线通信系统可以是5G NR通信系统。进一步，根据本公开的无线通信系统可以包括非地面网络。可选地，根据本公开的无线通信系统还可以包括地面网络。另外，本领域技术人员可以理解，根据本公开的无线通信系统还可以是4G或3G通信系统。
在根据本公开的实施例中，电子设备600向网络侧设备上报有关边缘链路的信道状态的信道信息，以供网络侧设备基于信道信息来对与边缘链路相关的电子设备600的学习模型进行分组，有助于网络侧设备通过分组解决边缘链路因环境不同导致的数据异构性问题，能够提高联合训练的效率，并且能够提高学习模型的质量和系统的性能。
作为示例，边缘链路的信道信息包括边缘链路的信道能量增益的概率分布、参考信号接收功率(RSRP)、接收信号的强度指标(RSSI)、参考信号质量(RSRQ)、信噪比(SNR)、有关边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
作为示例，网络侧设备基于与至少一个边缘链路分别对应的概率分布之间的相似程度，进行划分。有关基于边缘链路的信道能量增益的概率分布之间的相似程度进行划分的示例，可参见电子设备100实施例中结合图3进行的描述，这里不再累述。
作为示例，概率分布之间的相似程度包括概率分布之间的KL散度。
作为示例，信道能量增益被划分为预定数量的离散档位，以及概率分布包括信道能量增益处于每个档位的概率。有关信道能量增益处于每个档位的概率，可参见电子设备100实施例中的Pi，这里不再累述。
作为示例，网络侧设备基于RSRP的幅值来进行划分。
作为示例，网络侧设备基于RSSI的幅值来进行划分。
作为示例，网络侧设备基于RSRQ的幅值来进行划分。
作为示例，网络侧设备基于SNR的幅值来进行划分。
作为示例，网络侧设备根据边缘链路的作为发送方的电子设备和作为接收方的电子设备是否处于视距范围内，进行划分。
作为示例，网络侧设备基于信道的干扰和噪声的统计量的大小来进行划分。
作为示例，信道的干扰和噪声的统计量包括均值和/或方差。
作为示例，通信单元601可以被配置为经由无线资源控制RRC信令上报信道信息。例如，RRC信令可以是MeasResultsSL信令。有关MeasResultsSL的示例，可参见电子设备100实施例中的描述，这里不再累述。
作为示例，通信单元601可以被配置为通过物理下行控制信道(PDCCH)从网络侧设备接收有关划分的信息。
作为示例，通信单元601可以被配置为在联合训练的第一轮中，接收有关初始全局学习模型的参数。
作为示例，通信单元601可以被配置为通过上行链路向网络侧设备发送用于进行上行资源分配的辅助状态信息。
作为示例，辅助状态信息包括电子设备600训练学习模型所用的样本数量、电子设备600的位置信息、电子设备600的移动速度、电子设备600的计算能力、电子设备600的CPU占有率中的至少之一。有关辅助状态信息的示例，可参见电子设备100实施例中的描述，这里不再累述。
作为示例，通信单元601可以被配置为通过下行链路从网络侧设备接收有关上行资源分配的信息。
作为示例，通信单元601可以被配置为基于有关上行资源分配的信息，向网络侧设备发送有关本地学习模型的参数，其中，本地学习模型是基于网络侧设备下发的初始全局学习模型而训练得到的。
作为示例，联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合作为更新的全局学习模型，从而得到聚合后的学习模型，以及通信单元601可以被配置从网络侧设备接收有关聚合后的学习模型的参数。
作为示例，网络侧设备重复地进行划分和联合训练，直到满足预定条件为止。
作为示例，上述学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定边缘链路的数据发送速率。例如，学习模型可以是图3中涉及的深度强化学习的模型。
作为示例，电子设备600是D2D场景下的设备。例如，电子设备600是车联网中的车载设备。
在上文的实施方式中描述用于无线通信的电子设备的过程中，显然还公开了一些处理或方法。下文中，在不重复上文中已经讨论的一些细节的情况下给出这些方法的概要，但是应当注意，虽然这些方法在描述用于无线通信的电子设备的过程中公开，但是这些方法不一定采用所描述的那些部件或不一定由那些部件执行。例如，用于无线通信的电子设备的实施方式可以部分地或完全地使用硬件和/或固件来实现，而下面讨论的用于无线通信的方法可以完全由计算机可执行的程序来实现，尽管这些方法也可以采用用于无线通信的电子设备的硬件和/或固件。
图7示出了根据本公开的一个实施例的用于无线通信的方法S700的流程图。方法S700在步骤S702开始。在步骤S704中，基于位于电子设备服务范围内的至少一个用户设备上报的、有关至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组。在步骤S706中，针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。方法S700在步骤S708结束。
该方法例如可以通过上文所描述的电子设备100来执行，其具体细节可参见上述有关电子设备100的相关处理的描述，在此不再重复。
图8示出了根据本公开的一个实施例的用于无线通信的方法S800的流程图。方法S800在步骤S802开始。在步骤S804中，向为电子设备提供服务的网络侧设备上报有关电子设备的至少一个边缘链路的信道状态 的信道信息，以供网络侧设备基于信道信息将与至少一个边缘链路相关的电子设备的学习模型划分成至少一个分组，从而便于网络侧设备针对至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。方法S800在步骤S806结束。
该方法例如可以通过上文所描述的电子设备600来执行，其具体细节可参见上述有关电子设备600的相关处理的描述，在此不再重复。
本公开内容的技术能够应用于各种产品。
电子设备100可以被实现为各种网络侧设备例如基站。基站可以被实现为任何类型的演进型节点B(eNB)或gNB(5G基站)。eNB例如包括宏eNB和小eNB。小eNB可以为覆盖比宏小区小的小区的eNB，诸如微微eNB、微eNB和家庭(毫微微)eNB。对于gNB也可以由类似的情形。代替地，基站可以被实现为任何其他类型的基站，诸如NodeB和基站收发台(BTS)。基站可以包括：被配置为控制无线通信的主体(也称为基站设备)；以及设置在与主体不同的地方的一个或多个远程无线头端(RRH)。另外，各种类型的电子设备均可以通过暂时地或半持久性地执行基站功能而作为基站工作。
电子设备600可以被实现为各种用户设备。用户设备可以被实现为移动终端(诸如智能电话、平板个人计算机(PC)、笔记本式PC、便携式游戏终端、便携式/加密狗型移动路由器和数字摄像装置)或者车载终端(诸如汽车导航设备)。用户设备还可以被实现为执行机器对机器(M2M)通信的终端(也称为机器类型通信(MTC)终端)。此外，用户设备可以为安装在上述终端中的每个终端上的无线通信模块(诸如包括单个晶片的集成电路模块)。
[关于基站的应用示例]
(第一应用示例)
图9是示出可以应用本公开内容的技术的eNB或gNB的示意性配置的第一示例的框图。注意，以下的描述以eNB作为示例，但是同样可以应用于gNB。eNB 800包括一个或多个天线810以及基站设备820。基站设备820和每个天线810可以经由RF线缆彼此连接。
天线810中的每一个均包括单个或多个天线元件(诸如包括在多输入 多输出(MIMO)天线中的多个天线元件)，并且用于基站设备820发送和接收无线信号。如图9所示，eNB 800可以包括多个天线810。例如，多个天线810可以与eNB 800使用的多个频带兼容。虽然图9示出其中eNB 800包括多个天线810的示例，但是eNB 800也可以包括单个天线810。
基站设备820包括控制器821、存储器822、网络接口823以及无线通信接口825。
控制器821可以为例如CPU或DSP，并且操作基站设备820的较高层的各种功能。例如，控制器821根据由无线通信接口825处理的信号中的数据来生成数据分组，并经由网络接口823来传递所生成的分组。控制器821可以对来自多个基带处理器的数据进行捆绑以生成捆绑分组，并传递所生成的捆绑分组。控制器821可以具有执行如下控制的逻辑功能：该控制诸如为无线资源控制、无线承载控制、移动性管理、接纳控制和调度。该控制可以结合附近的eNB或核心网节点来执行。存储器822包括RAM和ROM，并且存储由控制器821执行的程序和各种类型的控制数据(诸如终端列表、传输功率数据以及调度数据)。
网络接口823为用于将基站设备820连接至核心网824的通信接口。控制器821可以经由网络接口823而与核心网节点或另外的eNB进行通信。在此情况下，eNB 800与核心网节点或其他eNB可以通过逻辑接口(诸如S1接口和X2接口)而彼此连接。网络接口823还可以为有线通信接口或用于无线回程线路的无线通信接口。如果网络接口823为无线通信接口，则与由无线通信接口825使用的频带相比，网络接口823可以使用较高频带用于无线通信。
无线通信接口825支持任何蜂窝通信方案(诸如长期演进(LTE)和LTE-先进)，并且经由天线810来提供到位于eNB 800的小区中的终端的无线连接。无线通信接口825通常可以包括例如基带(BB)处理器826和RF电路87。BB处理器826可以执行例如编码/解码、调制/解调以及复用/解复用，并且执行层(例如层1、介质访问控制(MAC)、无线链路控制(RLC)和分组数据汇聚协议(PDCP))的各种类型的信号处理。代替控制器821，BB处理器826可以具有上述逻辑功能的一部分或全部。BB处理器826可以为存储通信控制程序的存储器，或者为包括被配置为执行程序的处理器和相关电路的模块。更新程序可以使BB处理器826 的功能改变。该模块可以为插入到基站设备820的槽中的卡或刀片。可替代地，该模块也可以为安装在卡或刀片上的芯片。同时，RF电路87可以包括例如混频器、滤波器和放大器，并且经由天线810来传送和接收无线信号。
如图9所示，无线通信接口825可以包括多个BB处理器826。例如，多个BB处理器826可以与eNB 800使用的多个频带兼容。如图9所示，无线通信接口825可以包括多个RF电路87。例如，多个RF电路87可以与多个天线元件兼容。虽然图9示出其中无线通信接口825包括多个BB处理器826和多个RF电路87的示例，但是无线通信接口825也可以包括单个BB处理器826或单个RF电路87。
在图9所示的eNB 800中，电子设备100当实施为基站时，其收发器可以由无线通信接口825实现。功能的至少一部分也可以由控制器821实现。例如，控制器821可以通过执行电子设备100中的单元的功能来进行分组和联合训练。
(第二应用示例)
图10是示出可以应用本公开内容的技术的eNB或gNB的示意性配置的第二示例的框图。注意，类似地，以下的描述以eNB作为示例，但是同样可以应用于gNB。eNB 830包括一个或多个天线840、基站设备850和RRH 860。RRH 860和每个天线840可以经由RF线缆而彼此连接。基站设备850和RRH 860可以经由诸如光纤线缆的高速线路而彼此连接。
天线840中的每一个均包括单个或多个天线元件(诸如包括在MIMO天线中的多个天线元件)并且用于RRH 860发送和接收无线信号。如图10所示，eNB 830可以包括多个天线840。例如，多个天线840可以与eNB 830使用的多个频带兼容。虽然图10示出其中eNB 830包括多个天线840的示例，但是eNB 830也可以包括单个天线840。
基站设备850包括控制器851、存储器852、网络接口853、无线通信接口855以及连接接口857。控制器851、存储器852和网络接口853与参照图9描述的控制器821、存储器822和网络接口823相同。
无线通信接口855支持任何蜂窝通信方案(诸如LTE和LTE-先进)，并且经由RRH 860和天线840来提供到位于与RRH 860对应的扇区中的终端的无线通信。无线通信接口855通常可以包括例如BB处理器856。 除了BB处理器856经由连接接口857连接到RRH 860的RF电路864之外，BB处理器856与参照图9描述的BB处理器826相同。如图10所示，无线通信接口855可以包括多个BB处理器856。例如，多个BB处理器856可以与eNB 830使用的多个频带兼容。虽然图10示出其中无线通信接口855包括多个BB处理器856的示例，但是无线通信接口855也可以包括单个BB处理器856。
连接接口857为用于将基站设备850(无线通信接口855)连接至RRH 860的接口。连接接口857还可以为用于将基站设备850(无线通信接口855)连接至RRH 860的上述高速线路中的通信的通信模块。
RRH 860包括连接接口861和无线通信接口863。
连接接口861为用于将RRH 860(无线通信接口863)连接至基站设备850的接口。连接接口861还可以为用于上述高速线路中的通信的通信模块。
无线通信接口863经由天线840来传送和接收无线信号。无线通信接口863通常可以包括例如RF电路864。RF电路864可以包括例如混频器、滤波器和放大器，并且经由天线840来传送和接收无线信号。如图10所示，无线通信接口863可以包括多个RF电路864。例如，多个RF电路864可以支持多个天线元件。虽然图10示出其中无线通信接口863包括多个RF电路864的示例，但是无线通信接口863也可以包括单个RF电路864。
在图10所示的eNB 830中，电子设备100当实施为基站时，其收发器可以由无线通信接口855实现。功能的至少一部分也可以由控制器851实现。例如，控制器851可以通过执行电子设备100中的单元的功能来进行分组和联合训练。
[关于用户设备的应用示例]
(第一应用示例)
图11是示出可以应用本公开内容的技术的智能电话900的示意性配置的示例的框图。智能电话900包括处理器901、存储器902、存储装置903、外部连接接口904、摄像装置906、传感器907、麦克风908、输入装置909、显示装置910、扬声器911、无线通信接口912、一个或多个 天线开关915、一个或多个天线916、总线917、电池918以及辅助控制器919。
处理器901可以为例如CPU或片上系统(SoC)，并且控制智能电话900的应用层和另外层的功能。存储器902包括RAM和ROM，并且存储数据和由处理器901执行的程序。存储装置903可以包括存储介质，诸如半导体存储器和硬盘。外部连接接口904为用于将外部装置(诸如存储卡和通用串行总线(USB)装置)连接至智能电话900的接口。
摄像装置906包括图像传感器(诸如电荷耦合器件(CCD)和互补金属氧化物半导体(CMOS))，并且生成捕获图像。传感器907可以包括一组传感器，诸如测量传感器、陀螺仪传感器、地磁传感器和加速度传感器。麦克风908将输入到智能电话900的声音转换为音频信号。输入装置909包括例如被配置为检测显示装置910的屏幕上的触摸的触摸传感器、小键盘、键盘、按钮或开关，并且接收从用户输入的操作或信息。显示装置910包括屏幕(诸如液晶显示器(LCD)和有机发光二极管(OLED)显示器)，并且显示智能电话900的输出图像。扬声器911将从智能电话900输出的音频信号转换为声音。
无线通信接口912支持任何蜂窝通信方案(诸如LTE和LTE-先进)，并且执行无线通信。无线通信接口912通常可以包括例如BB处理器913和RF电路914。BB处理器913可以执行例如编码/解码、调制/解调以及复用/解复用，并且执行用于无线通信的各种类型的信号处理。同时，RF电路914可以包括例如混频器、滤波器和放大器，并且经由天线916来传送和接收无线信号。注意，图中虽然示出了一个RF链路与一个天线连接的情形，但是这仅是示意性的，还包括一个RF链路通过多个移相器与多个天线连接的情形。无线通信接口912可以为其上集成有BB处理器913和RF电路914的一个芯片模块。如图11所示，无线通信接口912可以包括多个BB处理器913和多个RF电路914。虽然图11示出其中无线通信接口912包括多个BB处理器913和多个RF电路914的示例，但是无线通信接口912也可以包括单个BB处理器913或单个RF电路914。
此外，除了蜂窝通信方案之外，无线通信接口912可以支持另外类型的无线通信方案，诸如短距离无线通信方案、近场通信方案和无线局域网(LAN)方案。在此情况下，无线通信接口912可以包括针对每种无线通信方案的BB处理器913和RF电路914。
天线开关915中的每一个在包括在无线通信接口912中的多个电路(例如用于不同的无线通信方案的电路)之间切换天线916的连接目的地。
天线916中的每一个均包括单个或多个天线元件(诸如包括在MIMO天线中的多个天线元件)，并且用于无线通信接口912传送和接收无线信号。如图11所示，智能电话900可以包括多个天线916。虽然图11示出其中智能电话900包括多个天线916的示例，但是智能电话900也可以包括单个天线916。
此外，智能电话900可以包括针对每种无线通信方案的天线916。在此情况下，天线开关915可以从智能电话900的配置中省略。
总线917将处理器901、存储器902、存储装置903、外部连接接口904、摄像装置906、传感器907、麦克风908、输入装置909、显示装置910、扬声器911、无线通信接口912以及辅助控制器919彼此连接。电池918经由馈线向图11所示的智能电话900的各个块提供电力，馈线在图中被部分地示为虚线。辅助控制器919例如在睡眠模式下操作智能电话900的最小必需功能。
在图11所示的智能电话900中，当电子设备600例如被实施为作为用户设备侧的智能电话的情况下、电子设备600的收发器可以由无线通信接口912实现。功能的至少一部分也可以由处理器901或辅助控制器919实现。例如，处理器901或辅助控制器919可以通过执行上述电子设备600中的单元的功能来上报边缘链路的信道信息。
(第二应用示例)
图12是示出可以应用本公开内容的技术的汽车导航设备920的示意性配置的示例的框图。汽车导航设备920包括处理器921、存储器922、全球定位系统(GPS)模块924、传感器925、数据接口926、内容播放器97、存储介质接口928、输入装置99、显示装置930、扬声器931、无线通信接口913、一个或多个天线开关936、一个或多个天线937以及电池938。
处理器921可以为例如CPU或SoC，并且控制汽车导航设备920的导航功能和另外的功能。存储器922包括RAM和ROM，并且存储数据和由处理器921执行的程序。
GPS模块924使用从GPS卫星接收的GPS信号来测量汽车导航设备920的位置(诸如纬度、经度和高度)。传感器925可以包括一组传感器，诸如陀螺仪传感器、地磁传感器和空气压力传感器。数据接口926经由未示出的终端而连接到例如车载网络941，并且获取由车辆生成的数据(诸如车速数据)。
内容播放器97再现存储在存储介质(诸如CD和DVD)中的内容，该存储介质被插入到存储介质接口928中。输入装置99包括例如被配置为检测显示装置930的屏幕上的触摸的触摸传感器、按钮或开关，并且接收从用户输入的操作或信息。显示装置930包括诸如LCD或OLED显示器的屏幕，并且显示导航功能的图像或再现的内容。扬声器931输出导航功能的声音或再现的内容。
无线通信接口913支持任何蜂窝通信方案(诸如LTE和LTE-先进)，并且执行无线通信。无线通信接口913通常可以包括例如BB处理器934和RF电路935。BB处理器934可以执行例如编码/解码、调制/解调以及复用/解复用，并且执行用于无线通信的各种类型的信号处理。同时，RF电路935可以包括例如混频器、滤波器和放大器，并且经由天线937来传送和接收无线信号。无线通信接口913还可以为其上集成有BB处理器934和RF电路935的一个芯片模块。如图12所示，无线通信接口913可以包括多个BB处理器934和多个RF电路935。虽然图12示出其中无线通信接口913包括多个BB处理器934和多个RF电路935的示例，但是无线通信接口913也可以包括单个BB处理器934或单个RF电路935。
此外，除了蜂窝通信方案之外，无线通信接口913可以支持另外类型的无线通信方案，诸如短距离无线通信方案、近场通信方案和无线LAN方案。在此情况下，针对每种无线通信方案，无线通信接口913可以包括BB处理器934和RF电路935。
天线开关936中的每一个在包括在无线通信接口913中的多个电路(诸如用于不同的无线通信方案的电路)之间切换天线937的连接目的地。
天线937中的每一个均包括单个或多个天线元件(诸如包括在MIMO天线中的多个天线元件)，并且用于无线通信接口913传送和接收无线信号。如图12所示，汽车导航设备920可以包括多个天线937。虽然图12 示出其中汽车导航设备920包括多个天线937的示例，但是汽车导航设备920也可以包括单个天线937。
此外，汽车导航设备920可以包括针对每种无线通信方案的天线937。在此情况下，天线开关936可以从汽车导航设备920的配置中省略。
电池938经由馈线向图12所示的汽车导航设备920的各个块提供电力，馈线在图中被部分地示为虚线。电池938累积从车辆提供的电力。
在图12示出的汽车导航设备920中，当电子设备600例如被实施为作为用户设备侧的汽车导航设备的情况下、电子设备600的收发器可以由无线通信接口933实现。功能的至少一部分也可以由处理器921实现。例如，处理器921可以通过执行上述电子设备600中的单元的功能来上报边缘链路的信道信息。
本公开内容的技术也可以被实现为包括汽车导航设备920、车载网络941以及车辆模块942中的一个或多个块的车载系统(或车辆)940。车辆模块942生成车辆数据(诸如车速、发动机速度和故障信息)，并且将所生成的数据输出至车载网络941。
以上结合具体实施例描述了本发明的基本原理，但是，需要指出的是，对本领域的技术人员而言，能够理解本发明的方法和装置的全部或者任何步骤或部件，可以在任何计算装置(包括处理器、存储介质等)或者计算装置的网络中，以硬件、固件、软件或者其组合的形式实现，这是本领域的技术人员在阅读了本发明的描述的情况下利用其基本电路设计知识或者基本编程技能就能实现的。
而且，本发明还提出了一种存储有机器可读取的指令代码的程序产品。指令代码由机器读取并执行时，可执行上述根据本发明实施例的方法。
相应地，用于承载上述存储有机器可读取的指令代码的程序产品的存储介质也包括在本发明的公开中。存储介质包括但不限于软盘、光盘、磁光盘、存储卡、存储棒等等。
在通过软件或固件实现本发明的情况下，从存储介质或网络向具有专用硬件结构的计算机(例如图13所示的通用计算机1300)安装构成该软件的程序，该计算机在安装有各种程序时，能够执行各种功能等。
在图13中，中央处理单元(CPU)1301根据只读存储器(ROM)1302中存储的程序或从存储部分1308加载到随机存取存储器(RAM)1303的程序执行各种处理。在RAM 1303中，也根据需要存储当CPU 1301执行各种处理等等时所需的数据。CPU 1301、ROM 1302和RAM 1303经由总线1304彼此连接。输入/输出接口1305也连接到总线1304。
下述部件连接到输入/输出接口1305：输入部分1306(包括键盘、鼠标等等)、输出部分1307(包括显示器，比如阴极射线管(CRT)、液晶显示器(LCD)等，和扬声器等)、存储部分1308(包括硬盘等)、通信部分1309(包括网络接口卡比如LAN卡、调制解调器等)。通信部分1309经由网络比如因特网执行通信处理。根据需要，驱动器1310也可连接到输入/输出接口1305。可移除介质1311比如磁盘、光盘、磁光盘、半导体存储器等等根据需要被安装在驱动器1310上，使得从中读出的计算机程序根据需要被安装到存储部分1308中。
在通过软件实现上述系列处理的情况下，从网络比如因特网或存储介质比如可移除介质1311安装构成软件的程序。
本领域的技术人员应当理解，这种存储介质不局限于图13所示的其中存储有程序、与设备相分离地分发以向用户提供程序的可移除介质1311。可移除介质1311的例子包含磁盘(包含软盘(注册商标))、光盘(包含光盘只读存储器(CD-ROM)和数字通用盘(DVD))、磁光盘(包含迷你盘(MD)(注册商标))和半导体存储器。或者，存储介质可以是ROM 1302、存储部分1308中包含的硬盘等等，其中存有程序，并且与包含它们的设备一起被分发给用户。
还需要指出的是，在本发明的装置、方法和系统中，各部件或各步骤是可以分解和/或重新组合的。这些分解和/或重新组合应该视为本发明的等效方案。并且，执行上述系列处理的步骤可以自然地按照说明的顺序按时间顺序执行，但是并不需要一定按时间顺序执行。某些步骤可以并行或彼此独立地执行。
最后，还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。此外， 在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上虽然结合附图详细描述了本发明的实施例，但是应当明白，上面所描述的实施方式只是用于说明本发明，而并不构成对本发明的限制。对于本领域的技术人员来说，可以对上述实施方式作出各种修改和变更而没有背离本发明的实质和范围。因此，本发明的范围仅由所附的权利要求及其等效含义来限定。
本技术还可以如下实现。
方案1.一种用于无线通信的电子设备，包括：
处理电路，被配置为：
基于位于所述电子设备服务范围内的至少一个用户设备上报的、有关所述至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与所述至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及
针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
方案2.根据方案1所述的电子设备，其中，
边缘链路的信道信息包括所述边缘链路的信道能量增益的概率分布、参考信号接收功率RSRP、接收信号的强度指标RSSI、参考信号质量RSRQ、信噪比SNR、有关所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
方案3.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于与所述至少一个边缘链路分别对应的概率分布之间的相似程度，进行所述划分。
方案4.根据方案3所述的电子设备，其中，
所述相似程度包括概率分布之间的KL散度。
方案5.根据方案2至4中任一项所述的电子设备，其中，
所述信道能量增益被划分为预定数量的离散档位，以及所述概率分 布包括所述信道能量增益处于每个档位的概率。
方案6.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于所述RSRP的幅值来进行所述划分。
方案7.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于所述RSSI的幅值来进行所述划分。
方案8.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于所述RSRQ的幅值来进行所述划分。
方案9.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于所述SNR的幅值来进行所述划分。
方案10.根据方案2所述的电子设备，其中，
所述处理电路被配置为根据所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内，进行所述划分。
方案11.根据方案2所述的电子设备，其中，
所述处理电路被配置为基于所述信道的干扰和噪声的统计量的大小来进行所述划分。
方案12.根据方案11所述的电子设备，其中，
所述信道的干扰和噪声的统计量包括均值和/或方差。
方案13.根据方案1至12中任一项所述的电子设备，其中，
所述处理电路被配置为经由无线资源控制RRC信令接收所述信道信息。
方案14.根据方案1至13中任一项所述的电子设备，其中，
所述处理电路被配置为通过物理下行控制信道PDCCH将有关所述划分的信息发送给与每个分组中的边缘链路有关的用户设备中的至少一部分用户设备。
方案15.根据方案14所述的电子设备，其中，
所述处理电路被配置为在所述联合训练的第一轮中，将有关初始全局学习模型的参数发送给所述至少一部分用户设备。
方案16.根据方案15所述的电子设备，其中，
所述处理电路被配置为通过上行链路从所述至少一部分用户设备接收用于进行上行资源分配的辅助状态信息。
方案17.根据方案16所述的电子设备，其中，
所述辅助状态信息包括用户设备训练学习模型所用的样本数量、用户设备的位置信息、用户设备的移动速度、用户设备的计算能力、用户设备的CPU占有率中的至少之一。
方案18.根据方案16或17所述的电子设备，其中，所述处理电路被配置为基于所述辅助状态信息，对所述至少一部分用户设备进行上行资源分配。
方案19.根据方案18所述的电子设备，其中，所述处理电路被配置为通过下行链路将有关上行资源分配的信息发送给所述至少一部分用户设备。
方案20.根据方案18或19所述的电子设备，其中，所述处理电路被配置为接收所述至少一部分用户设备基于所述有关上行资源分配的信息所上传的、有关本地学习模型的参数，其中，所述本地学习模型是基于所述电子设备下发的初始全局学习模型而训练得到的。
方案21.根据方案20所述的电子设备，其中，
所述联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合，从而得到聚合后的学习模型作为更新的全局学习模型，以及
所述处理电路被配置为将每个分组的有关所述聚合后的学习模型的参数广播给该分组中的用户设备。
方案22.根据方案1至21中任一项所述的电子设备，其中，所述处理电路被配置为重复地进行所述划分和所述联合训练，直到满足预定条件为止。
方案23.根据方案1至22中任一项所述的电子设备，其中，
所述学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定所述边缘链路的数据发送速率。
方案24.根据方案1至23中任一项所述的电子设备，其中，
所述至少一个用户设备是D2D场景下的设备。
方案25.一种用于无线通信的电子设备，包括：
处理电路，被配置为：
向为所述电子设备提供服务的网络侧设备上报有关所述电子设备的至少一个边缘链路的信道状态的信道信息，以供所述网络侧设备基于所述信道信息将与所述至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于所述网络侧设备针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
方案26.根据方案25所述的电子设备，其中，
边缘链路的信道信息包括所述边缘链路的信道能量增益的概率分布、参考信号接收功率RSRP、接收信号的强度指标RSSI、参考信号质量RSRQ、信噪比SNR、有关所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
方案27.根据方案26所述的电子设备，其中，
所述网络侧设备基于与所述至少一个边缘链路分别对应的概率分布之间的相似程度，进行所述划分。
方案28.根据方案27所述的电子设备，其中，
所述相似程度包括概率分布之间的KL散度。
方案29.根据方案26至28中任一项所述的电子设备，其中，
所述信道能量增益被划分为预定数量的离散档位，以及所述概率分布包括所述信道能量增益处于每个档位的概率。
方案30.根据方案26所述的电子设备，其中，
所述网络侧设备基于所述RSRP的幅值来进行所述划分。
方案31.根据方案26所述的电子设备，其中，
所述网络侧设备基于所述RSSI的幅值来进行所述划分。
方案32.根据方案26所述的电子设备，其中，
所述网络侧设备基于所述RSRQ的幅值来进行所述划分。
方案33.根据方案26所述的电子设备，其中，
所述网络侧设备基于所述SNR的幅值来进行所述划分。
方案34.根据方案26所述的电子设备，其中，
所述网络侧设备根据所述边缘链路的作为发送方的电子设备和作为接收方的电子设备是否处于视距范围内，进行所述划分。
方案35.根据方案26所述的电子设备，其中，
所述网络侧设备基于所述信道的干扰和噪声的统计量的大小来进行所述划分。
方案36.根据方案35所述的电子设备，其中，
所述信道的干扰和噪声的统计量包括均值和/或方差。
方案37.根据方案25至36中任一项所述的电子设备，其中，
所述处理电路被配置为经由无线资源控制RRC信令上报所述信道信息。
方案38.根据方案25至37中任一项所述的电子设备，其中，
所述处理电路被配置为通过物理下行控制信道PDCCH从所述网络侧设备接收有关所述划分的信息。
方案39.根据方案38所述的电子设备，其中，
所述处理电路被配置为在联合训练的第一轮中，接收有关初始全局学习模型的参数。
方案40.根据方案39所述的电子设备，其中，
所述处理电路被配置为通过上行链路向所述网络侧设备发送用于进行上行资源分配的辅助状态信息。
方案41.根据方案40所述的电子设备，其中，
所述辅助状态信息包括所述电子设备训练学习模型所用的样本数 量、所述电子设备的位置信息、所述电子设备的移动速度、所述电子设备的计算能力、所述电子设备的CPU占有率中的至少之一。
方案42.根据方案40或41所述的电子设备，其中，所述处理电路被配置为通过下行链路从所述网络侧设备接收有关上行资源分配的信息。
方案43.根据方案42所述的电子设备，其中，所述处理电路被配置为基于所述有关上行资源分配的信息，向所述网络侧设备发送有关本地学习模型的参数，其中，所述本地学习模型是基于所述网络侧设备下发的初始全局学习模型而训练得到的。
方案44.根据方案43所述的电子设备，其中，
所述联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合作为更新的全局学习模型，从而得到聚合后的学习模型，以及
所述处理电路被配置从所述网络侧设备接收有关所述聚合后的学习模型的参数。
方案45.根据方案25至44中任一项所述的电子设备，其中，所述网络侧设备重复地进行所述划分和所述联合训练，直到满足预定条件为止。
方案46.根据方案25至45中任一项所述的电子设备，其中，
所述学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定所述边缘链路的数据发送速率。
方案47.根据方案25至46中任一项所述的电子设备，其中，
所述电子设备是D2D场景下的设备。
方案48.一种用于无线通信的方法，包括：
基于位于电子设备服务范围内的至少一个用户设备上报的、有关所述至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与所述至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及
针对所述至少一个分组中的至少一部分分组，对处于同一分组内的 学习模型进行联合训练。
方案49.一种用于无线通信的方法，包括：
向为电子设备提供服务的网络侧设备上报有关所述电子设备的至少一个边缘链路的信道状态的信道信息，以供所述网络侧设备基于所述信道信息将与所述至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于所述网络侧设备针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
方案50.一种计算机可读存储介质，其上存储有计算机可执行指令，当所述计算机可执行指令被执行时，执行根据方案48或49所述的用于无线通信的方法。

Claims (50)

  1. 一种用于无线通信的电子设备，包括：
    处理电路，被配置为：
    基于位于所述电子设备服务范围内的至少一个用户设备上报的、有关所述至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与所述至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及
    针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
  2. 根据权利要求1所述的电子设备，其中，
    边缘链路的信道信息包括所述边缘链路的信道能量增益的概率分布、参考信号接收功率RSRP、接收信号的强度指标RSSI、参考信号质量RSRQ、信噪比SNR、有关所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
  3. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于与所述至少一个边缘链路分别对应的概率分布之间的相似程度，进行所述划分。
  4. 根据权利要求3所述的电子设备，其中，
    所述相似程度包括概率分布之间的KL散度。
  5. 根据权利要求2至4中任一项所述的电子设备，其中，
    所述信道能量增益被划分为预定数量的离散档位，以及所述概率分布包括所述信道能量增益处于每个档位的概率。
  6. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于所述RSRP的幅值来进行所述划分。
  7. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于所述RSSI的幅值来进行所述划分。
  8. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于所述RSRQ的幅值来进行所述划分。
  9. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于所述SNR的幅值来进行所述划分。
  10. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为根据所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内，进行所述划分。
  11. 根据权利要求2所述的电子设备，其中，
    所述处理电路被配置为基于所述信道的干扰和噪声的统计量的大小来进行所述划分。
  12. 根据权利要求11所述的电子设备，其中，
    所述信道的干扰和噪声的统计量包括均值和/或方差。
  13. 根据权利要求1至12中任一项所述的电子设备，其中，
    所述处理电路被配置为经由无线资源控制RRC信令接收所述信道信息。
  14. 根据权利要求1至13中任一项所述的电子设备，其中，
    所述处理电路被配置为通过物理下行控制信道PDCCH将有关所述划分的信息发送给与每个分组中的边缘链路有关的用户设备中的至少一部分用户设备。
  15. 根据权利要求14所述的电子设备，其中，
    所述处理电路被配置为在所述联合训练的第一轮中，将有关初始全局学习模型的参数发送给所述至少一部分用户设备。
  16. 根据权利要求15所述的电子设备，其中，
    所述处理电路被配置为通过上行链路从所述至少一部分用户设备接收用于进行上行资源分配的辅助状态信息。
  17. 根据权利要求16所述的电子设备，其中，
    所述辅助状态信息包括用户设备训练学习模型所用的样本数量、用 户设备的位置信息、用户设备的移动速度、用户设备的计算能力、用户设备的CPU占有率中的至少之一。
  18. 根据权利要求16或17所述的电子设备，其中，所述处理电路被配置为基于所述辅助状态信息，对所述至少一部分用户设备进行上行资源分配。
  19. 根据权利要求18所述的电子设备，其中，所述处理电路被配置为通过下行链路将有关上行资源分配的信息发送给所述至少一部分用户设备。
  20. 根据权利要求18或19所述的电子设备，其中，所述处理电路被配置为接收所述至少一部分用户设备基于所述有关上行资源分配的信息所上传的、有关本地学习模型的参数，其中，所述本地学习模型是基于所述电子设备下发的初始全局学习模型而训练得到的。
  21. 根据权利要求20所述的电子设备，其中，
    所述联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合，从而得到聚合后的学习模型作为更新的全局学习模型，以及
    所述处理电路被配置为将每个分组的有关所述聚合后的学习模型的参数广播给该分组中的用户设备。
  22. 根据权利要求1至21中任一项所述的电子设备，其中，所述处理电路被配置为重复地进行所述划分和所述联合训练，直到满足预定条件为止。
  23. 根据权利要求1至22中任一项所述的电子设备，其中，
    所述学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定所述边缘链路的数据发送速率。
  24. 根据权利要求1至23中任一项所述的电子设备，其中，
    所述至少一个用户设备是D2D场景下的设备。
  25. 一种用于无线通信的电子设备，包括：
    处理电路，被配置为：
    向为所述电子设备提供服务的网络侧设备上报有关所述电子设备 的至少一个边缘链路的信道状态的信道信息，以供所述网络侧设备基于所述信道信息将与所述至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于所述网络侧设备针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
  26. 根据权利要求25所述的电子设备，其中，
    边缘链路的信道信息包括所述边缘链路的信道能量增益的概率分布、参考信号接收功率RSRP、接收信号的强度指标RSSI、参考信号质量RSRQ、信噪比SNR、有关所述边缘链路的作为发送方的用户设备和作为接收方的用户设备是否处于视距范围内的信息、信道的干扰和噪声的统计量中至少之一。
  27. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于与所述至少一个边缘链路分别对应的概率分布之间的相似程度，进行所述划分。
  28. 根据权利要求27所述的电子设备，其中，
    所述相似程度包括概率分布之间的KL散度。
  29. 根据权利要求26至28中任一项所述的电子设备，其中，
    所述信道能量增益被划分为预定数量的离散档位，以及所述概率分布包括所述信道能量增益处于每个档位的概率。
  30. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于所述RSRP的幅值来进行所述划分。
  31. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于所述RSSI的幅值来进行所述划分。
  32. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于所述RSRQ的幅值来进行所述划分。
  33. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于所述SNR的幅值来进行所述划分。
  34. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备根据所述边缘链路的作为发送方的电子设备和作为接收方的电子设备是否处于视距范围内，进行所述划分。
  35. 根据权利要求26所述的电子设备，其中，
    所述网络侧设备基于所述信道的干扰和噪声的统计量的大小来进行所述划分。
  36. 根据权利要求35所述的电子设备，其中，
    所述信道的干扰和噪声的统计量包括均值和/或方差。
  37. 根据权利要求25至36中任一项所述的电子设备，其中，
    所述处理电路被配置为经由无线资源控制RRC信令上报所述信道信息。
  38. 根据权利要求25至37中任一项所述的电子设备，其中，
    所述处理电路被配置为通过物理下行控制信道PDCCH从所述网络侧设备接收有关所述划分的信息。
  39. 根据权利要求38所述的电子设备，其中，
    所述处理电路被配置为在联合训练的第一轮中，接收有关初始全局学习模型的参数。
  40. 根据权利要求39所述的电子设备，其中，
    所述处理电路被配置为通过上行链路向所述网络侧设备发送用于进行上行资源分配的辅助状态信息。
  41. 根据权利要求40所述的电子设备，其中，
    所述辅助状态信息包括所述电子设备训练学习模型所用的样本数量、所述电子设备的位置信息、所述电子设备的移动速度、所述电子设备的计算能力、所述电子设备的CPU占有率中的至少之一。
  42. 根据权利要求40或41所述的电子设备，其中，所述处理电路被配置为通过下行链路从所述网络侧设备接收有关上行资源分配的信息。
  43. 根据权利要求42所述的电子设备，其中，所述处理电路被配置为基于所述有关上行资源分配的信息，向所述网络侧设备发送有关本地 学习模型的参数，其中，所述本地学习模型是基于所述网络侧设备下发的初始全局学习模型而训练得到的。
  44. 根据权利要求43所述的电子设备，其中，
    所述联合训练包括对与处于同一分组内的边缘链路相关的本地学习模型进行聚合作为更新的全局学习模型，从而得到聚合后的学习模型，以及
    所述处理电路被配置从所述网络侧设备接收有关所述聚合后的学习模型的参数。
  45. 根据权利要求25至44中任一项所述的电子设备，其中，所述网络侧设备重复地进行所述划分和所述联合训练，直到满足预定条件为止。
  46. 根据权利要求25至45中任一项所述的电子设备，其中，
    所述学习模型用于根据数据队列长度和边缘链路的信道能量增益，来辅助确定所述边缘链路的数据发送速率。
  47. 根据权利要求25至46中任一项所述的电子设备，其中，
    所述电子设备是D2D场景下的设备。
  48. 一种用于无线通信的方法，包括：
    基于位于电子设备服务范围内的至少一个用户设备上报的、有关所述至少一个用户设备的至少一个边缘链路的信道状态的信道信息，将与所述至少一个边缘链路相关的用户设备的学习模型划分成至少一个分组，以及
    针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
  49. 一种用于无线通信的方法，包括：
    向为电子设备提供服务的网络侧设备上报有关所述电子设备的至少一个边缘链路的信道状态的信道信息，以供所述网络侧设备基于所述信道信息将与所述至少一个边缘链路相关的电子设备以及与所述至少一个边缘链路相关的由网络侧设备提供服务的其他电子设备的学习模型划分成至少一个分组，从而便于所述网络侧设备针对所述至少一个分组中的至少一部分分组，对处于同一分组内的学习模型进行联合训练。
  50. 一种计算机可读存储介质，其上存储有计算机可执行指令，当所述计算机可执行指令被执行时，执行根据权利要求48或49所述的用于无线通信的方法。
PCT/CN2023/105811 2022-07-11 2023-07-05 用于无线通信的电子设备和方法、计算机可读存储介质 WO2024012319A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210809772.4 2022-07-11
CN202210809772.4A CN117454952A (zh) 2022-07-11 2022-07-11 用于无线通信的电子设备和方法、计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2024012319A1 true WO2024012319A1 (zh) 2024-01-18

Family

ID=89535543

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/105811 WO2024012319A1 (zh) 2022-07-11 2023-07-05 用于无线通信的电子设备和方法、计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN117454952A (zh)
WO (1) WO2024012319A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113496291A (zh) * 2020-03-18 2021-10-12 索尼公司 用于联邦学习的装置、方法和存储介质
CN114051222A (zh) * 2021-11-08 2022-02-15 北京工业大学 一种车联网环境下基于联邦学习的无线资源分配和通信优化方法
CN114418131A (zh) * 2020-10-28 2022-04-29 索尼公司 用于联邦学习的电子设备以及方法
WO2022089751A1 (en) * 2020-10-29 2022-05-05 Nokia Technologies Oy Sampling user equipments for federated learning model collection

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113496291A (zh) * 2020-03-18 2021-10-12 索尼公司 用于联邦学习的装置、方法和存储介质
CN114418131A (zh) * 2020-10-28 2022-04-29 索尼公司 用于联邦学习的电子设备以及方法
WO2022089751A1 (en) * 2020-10-29 2022-05-05 Nokia Technologies Oy Sampling user equipments for federated learning model collection
CN114051222A (zh) * 2021-11-08 2022-02-15 北京工业大学 一种车联网环境下基于联邦学习的无线资源分配和通信优化方法

Also Published As

Publication number Publication date
CN117454952A (zh) 2024-01-26

Similar Documents

Publication Publication Date Title
US11991542B2 (en) Electronic device and method for wireless communication, and computer-readable storage medium
US10687217B2 (en) Spectrum management apparatus and method, apparatus and method for base station side and user device side
US11711133B2 (en) Electronic device, wireless communication method and computer readable storage medium related to transmission configuration indication (TCI) state
US20210385029A1 (en) User devices, wireless communication methods, and computer readable storage medium
US10849129B2 (en) Communication control apparatus, communication control method and terminal apparatus
KR102460473B1 (ko) 전자 디바이스, 정보 처리 장치, 및 정보 처리 방법
US20210136845A1 (en) Electronic device and method for wireless communications
WO2021036910A1 (zh) 数据传输方法及装置
US20230189317A1 (en) User scheduling using a graph neural network
US10674515B2 (en) User equipment and base station in wireless communications system, and wireless communications method
WO2024012319A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
WO2021254211A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
US20230362687A1 (en) Electronic device, wireless communication method, and non-transitory computer-readable storage medium
WO2021190435A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
WO2020156472A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
US11095405B2 (en) Electronic device and method for wireless communication
CN115589609A (zh) 用于无线通信系统的电子设备、方法和存储介质
WO2023185562A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
WO2023134570A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
WO2024027676A1 (zh) 用于分层联邦学习网络中的切换的装置、方法和介质
WO2021185129A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
WO2022253127A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
US20220400431A1 (en) Electronic device and method for wireless communication, and computer-readable storage medium
WO2024083034A1 (zh) 用于无线通信的电子设备和方法、计算机可读存储介质
CN113330758B (zh) 无线通信系统中的用户设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23838811

Country of ref document: EP

Kind code of ref document: A1