WO2023230818A1 - 一种波束管理方法及装置、用户设备、网络设备 - Google Patents

一种波束管理方法及装置、用户设备、网络设备 Download PDF

Info

Publication number
WO2023230818A1
WO2023230818A1 PCT/CN2022/096170 CN2022096170W WO2023230818A1 WO 2023230818 A1 WO2023230818 A1 WO 2023230818A1 CN 2022096170 W CN2022096170 W CN 2022096170W WO 2023230818 A1 WO2023230818 A1 WO 2023230818A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
sub
beam index
link quality
index
Prior art date
Application number
PCT/CN2022/096170
Other languages
English (en)
French (fr)
Inventor
刘文东
曹建飞
Original Assignee
Oppo广东移动通信有限公司
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 Oppo广东移动通信有限公司 filed Critical Oppo广东移动通信有限公司
Priority to PCT/CN2022/096170 priority Critical patent/WO2023230818A1/zh
Publication of WO2023230818A1 publication Critical patent/WO2023230818A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • G06F30/27Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L27/00Modulated-carrier systems

Definitions

  • the embodiments of the present application relate to the field of mobile communication technology, and specifically relate to a beam management method and device, user equipment, and network equipment.
  • Analog beamforming technology can enhance network coverage while reducing device implementation complexity.
  • Analog beamforming can be used not only for network equipment, but also for user equipment (User Equipment, UE).
  • User Equipment User Equipment
  • the optimal transmit beam and/or the optimal receive beam need to be determined.
  • the UE in order to determine the optimal transmit beam and/or the optimal receive beam, the UE needs to traverse all transmit beams and/or traverse all receive beams to perform corresponding link quality measurements, which will Bringing a lot of overhead and delay.
  • Embodiments of the present application provide a beam management method and device, user equipment, network equipment, chips, computer-readable storage media, computer program products, and computer programs.
  • the UE measures the link quality of multiple beams or beam pairs
  • the UE uses the first sub-model to determine at least one receive beam index and/or uplink feedback information based on the link quality of the multiple beams or beam pairs; wherein the uplink feedback information is used to determine at least one transmit beam index and /or at least one link quality corresponding to the at least one transmit beam index.
  • the network device receives the uplink feedback information sent by the UE;
  • the network device uses the second sub-model to determine at least one transmit beam index and/or at least one link quality corresponding to the at least one transmit beam index based on the uplink feedback information.
  • the beam management device provided by the embodiment of this application is applied to UE, and the device includes:
  • a measurement unit used to measure the link quality of multiple beams or beam pairs
  • a processing unit configured to use the first sub-model to determine at least one receive beam index and/or uplink feedback information based on the link quality of the multiple beams or beam pairs; wherein the uplink feedback information is used to determine at least one transmit beam The index and/or at least one link quality corresponding to the at least one transmit beam index.
  • the beam management device provided by the embodiment of the present application is applied to network equipment.
  • the device includes:
  • a receiving unit configured to receive uplink feedback information sent by the UE
  • a processing unit configured to use a second sub-model to determine at least one transmit beam index and/or at least one link quality corresponding to the at least one transmit beam index based on the uplink feedback information.
  • the user equipment provided by the embodiment of the present application includes a processor and a memory.
  • the memory is used to store computer programs
  • the processor is used to call and run the computer programs stored in the memory to execute the above-mentioned beam management method.
  • the network device provided by the embodiment of the present application includes a processor and a memory.
  • the memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to execute the above-mentioned beam management method.
  • the chip provided by the embodiment of the present application is used to implement the above beam management method.
  • the chip includes: a processor, configured to call and run a computer program from a memory, so that the device installed with the chip executes the above-mentioned beam management method.
  • the computer-readable storage medium provided by the embodiment of the present application is used to store a computer program, and the computer program causes the computer to execute the above-mentioned beam management method.
  • the computer program product provided by the embodiment of the present application includes computer program instructions, which cause the computer to execute the above-mentioned beam management method.
  • the computer program provided by the embodiment of the present application when run on a computer, causes the computer to execute the above-mentioned beam management method.
  • the technical solution of the embodiment of this application deploys the first sub-model on the UE side, so that the UE can measure the link quality of a small number of beams or beam pairs through the first sub-model based on the measured link quality of the beam or beam pair.
  • the path quality is predicted (that is, inferred) to at least one optimal receiving beam index and/or uplink feedback information.
  • the second sub-model is deployed on the network device side, so that the network device can predict at least one optimal transmit beam index and/or corresponding at least one link quality based on the uplink feedback information of the UE through the second sub-model.
  • Figure 1 is a schematic diagram of an application scenario according to the embodiment of the present application.
  • Figure 2 is a schematic diagram of a neuronal structure
  • Figure 3 is a schematic diagram of a deep neural network
  • Figure 4 is a schematic diagram of a convolutional neural network
  • Figure 5 is a schematic diagram of an LSTM unit structure
  • Figure 6 is a schematic diagram of a downlink beam scanning process
  • Figure 7 is a schematic flow chart of a beam management method provided by an embodiment of the present application.
  • Figure 8 is a schematic diagram of the model provided by the embodiment of the present application being deployed separately on the UE side and the NW side;
  • Figure 9-1 is a schematic structural diagram of the UE side sub-model provided by the embodiment of this application.
  • Figure 9-2 is a schematic structural diagram 2 of the UE side sub-model provided by the embodiment of this application.
  • Figure 9-3 is a schematic structural diagram three of the UE side sub-model provided by the embodiment of this application.
  • Figure 9-4 is a schematic structural diagram 4 of the UE side sub-model provided by the embodiment of this application.
  • Figure 10 is a schematic diagram of the model training process provided by the embodiment of the present application.
  • Figure 11 is a schematic diagram of the beam indication process provided by the embodiment of the present application.
  • Figure 12 is a schematic structural diagram of the beam management device provided by the embodiment of the present application.
  • Figure 13 is a schematic diagram 2 of the structure of the beam management device provided by the embodiment of the present application.
  • Figure 14 is a schematic structural diagram of a communication device provided by an embodiment of the present application.
  • Figure 15 is a schematic structural diagram of a chip according to an embodiment of the present application.
  • Figure 16 is a schematic block diagram of a communication system provided by an embodiment of the present application.
  • Figure 1 is a schematic diagram of an application scenario according to the embodiment of the present application.
  • the communication system may include user equipment 110 and network equipment 120.
  • Network device 120 may communicate with user device 110 over the air interface.
  • the network device 120 may be an access network device that communicates with the user equipment 110 .
  • Access network equipment may provide communication coverage for a specific geographic area and may communicate with user equipment 110 located within the coverage area.
  • the access network device may be a base station (gNB) in the NR system.
  • gNB base station
  • the user equipment 110 can be any user equipment, such as a mobile phone, a handheld device with a wireless communication function, a vehicle-mounted device, a wearable device, etc.
  • the wireless communication system shown in Figure 1 may also include a core network device 130 that communicates with the access network device.
  • the name of the core network device 130 may be different.
  • A indicates B, which can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also mean that there is an association between A and B. relation.
  • the "correspondence" mentioned in the embodiments of this application can mean that there is a direct correspondence or indirect correspondence between the two, it can also mean that there is an associated relationship between the two, or it can mean indicating and being instructed. , the relationship between configuring and being configured.
  • predefined can refer to what is defined in the protocol.
  • protocol may refer to a standard protocol in the communication field, which may include, for example, the NR protocol and related protocols applied in future communication systems. This application does not limit this.
  • Neural network is a computing model composed of multiple neurons (referred to as nodes) connected to each other.
  • the connections between nodes represent weighted values, called weights.
  • a node can be connected to multiple upper-level nodes, and the output of multiple upper-level nodes is used as the input of the node.
  • the node uses weights to perform a weighted sum of the inputs, and the result is processed through a specific activation function.
  • the neuron structure is shown in Figure 2.
  • the neuron uses n weights (denoted as w1, w2..., wn) to perform weighted summation of n inputs (denoted as a1, a2,..., an), and The obtained result is processed through the activation function (denoted as f), and the final output is t; optionally, the neuron can also add a bias b to the obtained result after performing a weighted summation.
  • DNN Deep Neural Network
  • CNN convolutional Neural Network
  • RNN Recurrent Neural Network
  • Figure 3 is a typical deep neural network.
  • the basic structure of a deep neural network includes: input layer, hidden layer and output layer.
  • the input of the input layer is the input of the deep neural network, and the output of the output layer That is the output of the deep neural network.
  • the deep neural network can produce different outputs, and then fit the mapping relationship from input to output.
  • the deep neural network shown in Figure 3 is a fully connected model, and each node is connected to all its next-level nodes.
  • Figure 4 is a typical convolutional neural network.
  • the basic structure of a convolutional neural network includes: input layer, multiple convolutional layers, multiple pooling layers, fully connected layers and output layers.
  • the input of the layer is the input of the convolutional neural network
  • the output of the output layer is the output of the convolutional neural network.
  • each neuron of the convolution kernel in the convolution layer is locally connected to its input, and the local maximum or average feature of a certain layer is extracted by introducing a pooling layer, which effectively reduces the network's Parameters and local features are mined, allowing the convolutional neural network to converge quickly and obtain excellent performance.
  • Recurrent neural network is a neural network that models sequence data. It has achieved remarkable results in the field of natural language processing, such as machine translation, speech recognition and other applications. The specific performance is that the recurrent neural network remembers the information of the past moment and uses it in the calculation of the current output, that is, the nodes between the hidden layers are no longer unconnected but connected, and the input of the hidden layer not only includes the above The input of a layer also includes the output of the hidden layer at the previous moment.
  • Recurrent neural networks generally include long-short-term memory artificial neural network (Long-Short Term Memory, LSTM) unit structure.
  • Figure 5 is a typical LSTM unit structure, as shown in Figure 5.
  • the LSTM unit structure includes: tanh node and ⁇ node .
  • the cell state of LSTM determines which states should be kept and which states should be forgotten, solving the shortcomings of traditional convolutional neural networks in long-term memory.
  • the neural network in the technical solution of the embodiment of the present application may be a deep neural network, but is not limited thereto.
  • the deep neural network may also be replaced by a convolutional neural network or a recurrent neural network as the implementation of the present application.
  • the beam management mechanism is divided into uplink beam management mechanism and downlink beam management mechanism.
  • the downlink beam management mechanism includes downlink beam scanning, UE optimal beam reporting, network equipment instructions for downlink beams, and other processes. It should be noted that, for a network device, the downlink beam refers to the transmit beam of the network device; for a UE, the downlink beam refers to the receive beam of the UE.
  • P1 process the network device traverses all transmit beams and the UE traverses all receive beams, so that all beam pair combinations can be traversed, and the UE measures the corresponding link quality for each beam pair.
  • P2 process the UE's receive beam is specific, and the network device traverses all transmit beams, so that all transmit beams can be traversed for a specific receive beam, and the UE measures the corresponding link quality for each transmit beam.
  • the transmit beam of the network device is specific, and the UE traverses all receive beams, so that it can traverse all receive beams for a specific transmit beam, and the UE measures the corresponding link quality for each receive beam.
  • the UE After the UE performs link quality measurement, it selects the K beams or beam pairs with the highest link quality (that is, the K optimal beams or beam pairs) by comparing the measured link quality of all beams or beam pairs, and then passes The uplink control information is reported to the network device, and K is a positive integer.
  • the network device After the network device obtains the K optimal beams or beam pairs, it selects the transmit beam used for downlink transmission from the K optimal beams or beam pairs, and intervenes through the media control element (Media Access Control-Control Element, MAC CE ) or the transmission configuration indication (TCI) status carried in the downlink control information (Downlink Control Information, DCI) to complete the instructions for the transmit beam.
  • the UE uses the receive beam corresponding to the transmit beam to receive downlink transmission.
  • all transmit beams and/or receive beams need to be traversed, which will bring a lot of overhead and delay.
  • the network device has 64 transmit beams and the UE has 4 receive beams
  • the UE needs to measure 256 (ie, 64 ⁇ 4) beam pairs, thus requiring 256 downlink resource overhead. From a time perspective, the measurement of each beam pair takes a certain amount of time to complete, and the measurement of 256 beam pairs will take a very long time. It can be seen that the current beam scanning process will bring a lot of overhead and delay.
  • the technical solution of the embodiment of this application uses artificial intelligence/machine learning (AI/ML) technology to train the neural network by constructing a data set and deploying it on the UE side and/or the network device side, so that the UE At least one optimal receiving beam index and/or uplink feedback information can be predicted (that is, inferred) through the neural network when the measurement results of some beams or beam pairs are used as inputs to the neural network.
  • the network device predicts at least one optimal transmit beam index and/or corresponding at least one link quality based on the uplink feedback information reported by the UE through the neural network. In this way, the network device and the UE can quickly match the optimal transmit beam and/or the optimal receive beam, and significantly reduce the resource overhead and delay in the beam scanning process.
  • the beam described in the embodiment of this application refers to a transmitting beam or a receiving beam
  • a beam pair refers to a pair of transmitting beam and receiving beam
  • model described in the embodiments of this application can also be called a "neural network”, which can be, but is not limited to, a deep neural network, a convolutional neural network, or a recurrent neural network.
  • FIG. 7 is a schematic flow chart of a beam management method provided by an embodiment of the present application. As shown in Figure 7, the beam management method includes the following steps:
  • Step 701 The UE measures the link quality of multiple beams or beam pairs.
  • Step 702 The UE uses the first sub-model to determine at least one receiving beam index and/or uplink feedback information based on the link quality of the multiple beams or beam pairs; wherein the uplink feedback information is used to determine at least one transmit The beam index and/or at least one link quality corresponding to the at least one transmit beam index.
  • the multiple beams measured by the UE are a subset of all beams.
  • all beams refer to all transmit beams or all receive beams, that is, all transmit beams of the network device or all receive beams of the UE.
  • the beam pairs measured by the UE are a subset of all beam pairs.
  • all beam pairs refer to all transmit beam and receive beam pairs, that is, transmit beams and receive beam pairs formed by all possible combinations of all receive beams of the UE and all transmit beams of the network device.
  • the link quality of the beam refers to the link quality corresponding to the transmission beam.
  • the link quality of a beam pair refers to the link quality corresponding to the transmit beam and receive beam pairs.
  • measurement of a beam or a pair of beams refers to measurement of a reference signal associated with a beam or a pair of beams.
  • the reference signal includes at least one of the following: synchronization signal block (Synchronization Signal and PBCH Block, SSB), channel state information-reference signal (Channel State Information-Reference Signal, CSI-RS).
  • synchronization signal block Synchronization Signal and PBCH Block, SSB
  • channel state information-reference signal Channel State Information-Reference Signal, CSI-RS
  • the link quality includes at least one of the following: Layer 1-Reference Signal Receiving Power (L1-RSRP), Layer 1-signal to interference plus noise ratio (Layer 1-Reference Signal Receiving Power, L1-RSRP) -Signal to Interference plus Noise Ratio, L1-SINR), Layer 1-Reference Signal Receiving Quality (Layer1-Reference Signal Receiving Quality, L1-RSRQ).
  • L1-RSRP Layer 1-Reference Signal Receiving Power
  • L1-RSRP Layer 1-signal to interference plus noise ratio
  • L1-RSRP Layer 1-Reference Signal Receiving Power -RSRP
  • L1-RSRP Layer 1-signal to interference plus noise ratio
  • L1-SINR Layer 1-Reference Signal Receiving Quality
  • Layer1-RSRQ Layer 1-Reference Signal Receiving Quality
  • the UE uses the first sub-model to determine at least one receiving beam index and/or uplink feedback information based on the link quality of the multiple beams or beam pairs.
  • the plurality of beams are multiple receiving beams or multiple transmitting beams
  • the plurality of beam pairs are multiple pairs of transmitting beams and receiving beams.
  • the UE uses the first sub-model to determine the uplink feedback information based on the link quality of the multiple beams.
  • the UE uses the first sub-model to determine at least one receiving beam index based on the link quality of the multiple beams.
  • the UE uses the first sub-model to determine at least one receiving link based on the link quality of the multiple beam pairs. Beam index and uplink feedback information.
  • the UE when the UE determines the uplink feedback information, the UE sends the uplink feedback information to a network device.
  • the network device receives the uplink feedback information sent by the UE, and the network The device uses the second sub-model to determine at least one transmit beam index and/or at least one link quality corresponding to the at least one transmit beam index based on the uplink feedback information.
  • the uplink feedback information is used by the network device to determine at least one transmit beam index and/or at least one link quality corresponding to the at least one transmit beam index using the second sub-model.
  • the at least one receiving beam index is the index of the K receiving beams with the best link quality among the multiple beams or beam pairs, and K is a positive integer.
  • the at least one receiving beam index may be recorded as K optimal receiving beam indexes.
  • the at least one transmit beam index is the index of the K transmit beams with the best link quality among the plurality of beams or beam pairs, and K is a positive integer.
  • the at least one transmit beam index may be recorded as K optimal transmit beam indexes.
  • the value of K may be configured by the network device or predefined or determined based on the UE itself.
  • the link quality is the link quality of the beam or beam pair
  • there is a corresponding relationship between the transmitting beam and the link quality or there is a corresponding relationship between the receiving beam and the link quality, or there is a corresponding relationship between the transmitting beam and the link quality.
  • the corresponding relationship is determined by the ordering of the at least one receive beam index, the order of the at least one transmit beam index, and the order of the at least one link quality.
  • the first sub-model on the UE side outputs K optimal receive beam indexes
  • the second sub-model on the network device side outputs K optimal transmit beam indexes and K link qualities.
  • the K optimal receiving beam indexes are denoted as ⁇ n1, n2,...nK ⁇
  • the K optimal transmitting beam indexes are denoted as ⁇ m1, m2,...mK ⁇
  • the K link qualities are denoted as ⁇ r1, r2,..., rK ⁇ .
  • the link quality can be sorted from high to low, that is, ⁇ r1>r2>...>rK ⁇ .
  • This correspondence consists of the ordering of the K optimal receive beam indexes and the ordering of the K optimal transmit beam indexes.
  • the implicit indication of the sorting of K link quality that is, the optimal receiving beam index, the optimal transmitting beam index and the link quality sorted at the same position are corresponding.
  • the first sub-model output on the UE side The K optimal receive beam indexes are sorted according to the corresponding link quality (such as from high to low); the K optimal transmit beam indexes output by the second sub-model on the network device side are sorted according to the corresponding link quality. sort (such as from high to low).
  • the K optimal receiving beam indexes, K optimal transmitting beam indexes and K link qualities can be expressed as a set of triples according to the corresponding relationship, denoted as ⁇ (m1,n1,r1),(m2,n2 ,r2),...,(mK,nK,rK) ⁇ .
  • the input of the first sub-model on the UE side is the link quality of multiple beams or beam pairs measured by the UE.
  • the output of the first sub-model on the UE side includes at least one receive beam index and/or uplink feedback information.
  • the first sub-model on the UE side is at least one receive beam index
  • the first sub-model has an output layer
  • the UE outputs at least one receive beam index through the output layer.
  • the first sub-model For the case where the output of the first sub-model on the UE side is uplink feedback information, the first sub-model has an output layer, and the UE outputs first information through the output layer, and the first information is used to determine the uplink feedback information. Feedback. Further, the first sub-model further includes a quantization layer, and the UE performs quantization processing on the first information through the quantization layer to obtain the uplink feedback information.
  • the output layer of the first sub-model may have the following implementation.
  • the first sub-model includes an output layer; the UE uses the first sub-model to process the link quality of the multiple beams or beam pairs, and outputs at least one receiving node through the first part of the output layer. Beam index, and output first information through the second part of nodes of the output layer, where the first information is used to determine uplink feedback information. Further, the first sub-model further includes a quantization layer, and the UE performs quantization processing on the first information through the quantization layer to obtain the uplink feedback information.
  • the first sub-model includes a first output layer and a second output layer; the UE uses the first sub-model to process the link quality of the multiple beams or beam pairs, and through the first output The layer outputs at least one receive beam index, and outputs first information through the second output layer, where the first information is used to determine uplink feedback information. Further, the first sub-model further includes a quantization layer, and the UE performs quantization processing on the first information through the quantization layer to obtain the uplink feedback information.
  • the first output layer is adjacent to the second output layer, or the first output layer and the second output layer are separated by at least one neural network layer.
  • the first output layer is located before the second output layer, or the first output layer is located after the second output layer.
  • the first output layer is located before the second output layer, and the second output layer is the last neural network layer of the first sub-model; The first output layer is the middle neural network layer of the first sub-model.
  • the first output layer may be all or part of the middle neural network layer of the first sub-model.
  • the middle neural network layer of the first sub-model refers to the hidden layer.
  • the first output layer is located after the second output layer, and the first output layer is the last neural network layer of the first sub-model;
  • the second output layer is the middle neural network layer of the first sub-model.
  • the second output layer may be all or part of the middle neural network layer of the first sub-model.
  • the middle neural network layer of the first sub-model refers to the hidden layer.
  • the first output layer is located after the second output layer, and the first output layer is the last neural network layer of the first sub-model;
  • the second output layer is the first neural network layer of the first sub-model.
  • the second output layer may be the entire first neural network layer.
  • the first neural network layer of the first sub-model refers to the input layer, and the input of the first neural network layer is the link quality of the multiple beams or beam pairs.
  • the association between the first output layer and the second output layer means that there is an association between the nodes of the first output layer and the nodes of the second output layer.
  • the correlation between the node of the first output layer and the node of the second output layer means that the output of the node of the first output layer is correlated with the input of the node of the second output layer ( This is for the case where the first output layer is located before the second output layer), or the output of the node of the second output layer has a correlation with the input of the node of the first output layer (this is for the The second output layer is located before the first output layer).
  • this association can be a direct association (corresponding to the situation where two layers are adjacent) or an indirect association (corresponding to the situation where two layers are not adjacent).
  • the uplink feedback information obtained by the UE after quantization through the quantization layer may be a bit sequence.
  • the uplink feedback information is a first bit sequence, and the first bit sequence is used to determine at least one transmission A beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the uplink feedback information is a second bit sequence, and the second bit sequence is used to indicate the link quality of the multiple beams or beam pairs.
  • the link quality of the multiple beams or beam pairs is used to determine at least one transmit beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the UE uses the receive beam indicated by the one receive beam index to receive downlink transmission.
  • the network device uses the transmit beam indicated by the one transmit beam index to send downlink transmission.
  • the UE determines a first receive beam index from the multiple receive beam indexes, using the first receive beam index.
  • the receiving beam indicated by the beam index receives the downlink transmission.
  • the UE receives the first indication information sent by the network device, the first indication information is used to indicate the first transmit beam index adopted by the network device; the UE is based on the at least one receive beam index and the The corresponding relationship of at least one transmit beam index determines the first receive beam index corresponding to the first transmit beam index.
  • the first indication information is carried in MAC CE or DCI.
  • the network device selects a first transmit beam index from the multiple transmit beam indexes, using the first transmit beam index.
  • the transmit beam indicated by the transmit beam index performs downlink transmission.
  • the network device sends first indication information to the UE, where the first indication information is used to indicate the first transmit beam index used by the network device.
  • the first indication information is carried in MAC CE or DCI.
  • the first sub-model in the above solution is pre-trained, and the first sub-model can be trained through the network device, and the trained first sub-model can be delivered to the UE.
  • the second sub-model in the above solution is also pre-trained, and can be trained through the network device.
  • the training of the first sub-model and/or the second sub-model is not limited to the network device.
  • the first sub-model and/or the second sub-model can also be trained by other devices with processing capabilities. submodel.
  • training the first sub-model and/or the second sub-model by a network device includes: the UE reporting a data set to the network device; and the network device receiving the data reported by the UE.
  • a data set The network device uses the data set to train the first sub-model and/or the second sub-model.
  • the first sub-model is used for the UE based on multiple beams or beam pairs.
  • the link quality determines at least one receive beam index and/or uplink feedback information
  • the second sub-model is used by the network device to determine at least one transmit beam index and/or the at least one transmit beam index based on the uplink feedback information. Corresponding to at least one link quality.
  • the data set reported by the UE to the network device is used by the network device to train the first sub-model and/or the second sub-model.
  • the data set is used by a network device to train the first sub-model or the first sub-model plus the second sub-model (or the network uses the data set to train the third sub-model).
  • the network device sends the trained first sub-model to the UE; the UE receives the Issue the trained first sub-model.
  • the data set includes:
  • Input data of the first sub-model the input data being the link quality of at least some beams or beam pairs in the first set of beams or beam pairs;
  • the label of the output data of the first sub-model is the index of the K receiving beams with the best link quality in the first beam or beam pair set;
  • the label of the output data of the second sub-model is the index of the K transmit beams with the best link quality in the first beam or beam pair set and the corresponding K link qualities, K is Positive integer.
  • the first beam or beam pair set refers to the first beam set or the first beam pair set.
  • the first beam set includes all transmit beams or all receive beams, that is, all transmit beams of the network device or all receive beams of the UE.
  • the first beam pair set includes all transmit beams and receive beam pairs, that is, transmit beams and receive beam pairs formed by all possible combinations of all receive beams of the UE and all transmit beams of the network device.
  • the first sub-model and the second sub-model are trained jointly; or, the first sub-model and the second sub-model are trained independently.
  • the first sub-model and the second sub-model share a loss function, which can be understood as the first sub-model and the second sub-model
  • the joint loss function is used to jointly optimize the weights of the two sub-models. Specifically, the joint loss function is used to calculate the loss value between the output data of the entire model and the label, and determine whether the loss value meets the preset conditions. If not, the first sub-model and the second sub-model are updated based on the loss value.
  • the output data of the entire model includes the indices of K receive beams, the indices of K transmit beams, and the corresponding K link qualities.
  • the first sub-model and the second sub-model have independent loss functions, and the weights of the two sub-models are optimized through the two loss functions.
  • the first loss function is used to calculate the loss value between the output data of the first sub-model and the label of the output data of the first sub-model, and determine whether the loss value satisfies the preset condition. If not, the loss value is calculated according to the loss value. Update the first sub-model with the value (including but not limited to updating the weight of the first sub-model), and continue to iteratively perform the above steps until the loss value meets the preset conditions, and the training of the first sub-model is completed.
  • the loss value between the output data of the second sub-model and the label of the output data of the second sub-model is calculated through the second loss function, and it is judged whether the loss value satisfies the preset condition. If not, the loss value is calculated based on the loss value.
  • Update the second sub-model (including but not limited to updating the weight of the second sub-model), and continue to iteratively perform the above steps until the loss value meets the preset conditions, and the training of the second sub-model is completed.
  • the output data of the first sub-model includes the indices of K receive beams
  • the output data of the second sub-model includes the indices of K transmit beams and the corresponding K link qualities.
  • the second sub-model i.e., the uplink feedback information
  • the first sub-model for multiple beams Or obtained after processing the link quality of the beam pair here refers to processing other than the quantization layer
  • the second sub-model and the first sub-model need to be paired for training, that is, the first sub-model and the third sub-model need to be paired for training.
  • the two sub-models need to be trained jointly.
  • the first sub-model and the second sub-model also need to be deployed in pairs.
  • the link quality of multiple beams or beam pairs (here refers to processing other than the quantization layer), so the second sub-model and the first sub-model do not need to be paired for training, and the first sub-model and the second sub-model can be trained independently. Furthermore, for this case, the first sub-model and the second sub-model can also be deployed independently.
  • the technical solution of the embodiment of this application implements AI/ML-based beam management, and deploys trained neural networks on the network equipment and/or UE side respectively.
  • the beam management process includes at least one of the following: UE measurement process, UE inference (i.e. prediction) process, UE reporting and network device inference (i.e. prediction) process.
  • UE measurement process the UE measures part of all beams or beam pairs or beam pairs;
  • UE inference (i.e. prediction) process the UE predicts the optimal value based on the measurement results through the local neural network (i.e.
  • the first sub-model Receive beam index and/or uplink feedback information; for the UE reporting and network device inference (i.e., prediction) process, the UE reports the uplink feedback information to the network device, and the network device uses the local neural network (i.e., the second sub-model) according to the uplink
  • the feedback information predicts the optimal transmit beam and corresponding link quality.
  • a beam pair refers to a pair of transmitting beams and receiving beams.
  • the link quality is "L1-RSRP" as an example. Of course, it is not limited to this, and the link quality can also be L1-SINR or L1-RSRQ.
  • NW network
  • the model is deployed separately on the UE side and the NW side.
  • the model (the part included in the dotted box) includes the UE side sub-model (that is, the first sub-model) and the NW side sub-model (that is, the second sub-model).
  • the input of the UE side sub-model It is the L1-RSRP of the beam pair measured by the UE.
  • the measured beam pair is a subset of all beam pairs.
  • all beam pairs refer to all transmit beam and receive beam pairs, that is, all the receive beams of the UE and all the NW
  • the transmit beams undergo all possible combinations to form pairs of transmit beams and receive beams.
  • the UE side sub-model outputs K optimal receiving beam indexes, and at the same time, the UE side outputs uplink feedback information, which can be reported to the NW side through MAC CE (corresponding to PUSCH) or DCI (corresponding to PDCCH).
  • the input of the NW side sub-model is the uplink feedback information, and the output is the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes, where K is a positive integer.
  • the UE side sub-model outputs an optimal receive beam index
  • the NW side sub-model outputs an optimal transmit beam index and the L1-RSRP corresponding to the optimal transmit beam index.
  • the UE selects an optimal receiving beam output by the UE side sub-model for receiving downlink transmission
  • the NW selects an optimal transmit beam output by the NW side sub-model for transmitting downlink transmission.
  • the optimal receive beam indicated by an optimal receive beam index output by the UE side sub-model and the optimal transmit beam indicated by an optimal transmit beam index output by the NW side sub-model automatically realize the optimal beam pair. Matching, the UE and NW side no longer need to indicate the beam used.
  • the link quality of the optimal beam pair is given by the L1-RSRP output by the NW side sub-model.
  • the UE side sub-model outputs K (K>1) optimal receive beam indexes, and the NW side sub-model outputs K optimal transmit beam indexes and L1-RSRP corresponding to the K optimal transmit beam indexes.
  • the K optimal receive beam indexes output by the UE side sub-model (here represented by n) and the K optimal transmit beam indexes output by the NW side sub-model (here represented by m) and L1-RSRP (here Represented by the symbol r) contains an implicit relationship of one-to-one pairing, that is, K beam pairs and the corresponding L1-RSRP can be expressed as a triple set ⁇ (m1, n1, r1), (m2, n2, r2),... ,(mK,nK,rK) ⁇ , where ⁇ m1,m2,...mK ⁇ represents K optimal transmit beam indexes, ⁇ n1,n2,...nK ⁇ represents K optimal receive beam indexes, ⁇ r1,r2, ...,rK ⁇ represents the L1-RSRP corresponding to the K optimal transmit beam indexes, optionally, ⁇ r1>r2>...>rK ⁇ .
  • the pairing relationship (that is, the corresponding relationship) between the optimal transmit beam index and the optimal receive beam index is implicitly indicated by the order of the beam index sequences output by the UE side sub-model and the NW side sub-model, and is not explicitly indicated by the uplink feedback information.
  • Indicates that the K optimal receive beam indexes output by the UE side sub-model and the K optimal transmit beam indexes output by the NW side sub-model are arranged in descending order of the corresponding L1-RSRP.
  • the structure of the UE side sub-model (that is, the first sub-model) can be implemented in multiple ways.
  • the structures of several UE side sub-models are given below. It should be noted that this example uses the implementation method of deep neural network to illustrate the structure of the UE side sub-model, but it is not limited to this. Other implementation methods such as convolutional neural network and recurrent neural network can also implement the UE side sub-model. Structure. Implementation methods using other neural networks can be extended based on this example and using the same input/output interface.
  • the UE side sub-model includes an input layer, a hidden layer and an output layer.
  • the first information and the K optimal receiving beam indexes are output in parallel at the output layer, and share all hidden layer weights.
  • the input of the input layer is a matrix in which the measured L1-RSRPs of M beam pairs are arranged in a specified order.
  • the hidden layer consists of multiple fully connected layers.
  • the output layer is composed of two parts, in which K nodes of the output layer output K optimal receiving beam indexes, other nodes of the output layer output the first information, and the first information outputs uplink feedback information after passing through the quantization layer. After the uplink feedback information is reported to the NW side, it is used as the input of the NW side sub-model.
  • the uplink feedback information is a bit sequence.
  • This bit sequence does not explicitly indicate the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes, but implicitly indicates the K optimal transmit beam indexes.
  • the beam index and the L1-RSRP information corresponding to the K optimal transmit beam indexes can be extracted by the matching NW side sub-model. Therefore, the UE side sub-model and the NW side sub-model need to be paired for training and deployment, so that the NW can correctly predict the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes from the uplink feedback information. .
  • the UE side sub-model includes an input layer, a hidden layer, an uplink feedback information output layer and an optimal receiving beam index output layer.
  • the uplink feedback information output layer and the optimal receiving beam index output layer are not in the same layer, and only share some hidden layer weights.
  • the input of the input layer is a matrix in which the measured L1-RSRPs of M beam pairs are arranged in a specified order.
  • the hidden layer is composed of multiple fully connected layers.
  • the uplink feedback information output layer is part of a neural network layer of the hidden layer, and the uplink feedback information output layer is before the optimal receiving beam index output layer, and the optimal receiving beam index output layer outputs K optimal receiving beam indexes, the uplink feedback information output layer outputs the first series of information, and the first information outputs the uplink feedback information after passing through the quantization layer.
  • the uplink feedback information is a bit sequence. This bit sequence does not explicitly indicate the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes, but implicitly indicates the K optimal transmit beam indexes.
  • the beam index and the L1-RSRP information corresponding to the K optimal transmit beam indexes can be extracted by the matching NW side sub-model. Therefore, the UE side sub-model and the NW side sub-model need to be paired for training and deployment, so that the NW can correctly predict the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes from the uplink feedback information. .
  • the uplink feedback information output layer in Figure 9-2 is part of a neural network layer of the hidden layer, it is not limited to this.
  • the uplink feedback information output layer can also be a part of a neural network layer of the hidden layer. all.
  • the uplink feedback information output layer in Figure 9-2 is adjacent to the optimal receive beam index output layer, it is not limited to this.
  • the uplink feedback information output layer can also be separated from the optimal receive beam index output layer by one or more nerves. Network layer.
  • the output of the uplink feedback information output layer can act or not act (dotted line weight connection in Figure 9-2) to obtain part of the implicit information of the K optimal receiving beam indexes.
  • the UE side sub-model includes an input layer, a hidden layer, an optimal receiving beam index output layer and an uplink feedback information output layer.
  • the optimal receiving beam index output layer and the uplink feedback information output layer are not in the same layer, and only share some hidden layer weights.
  • the input of the input layer is a matrix in which the measured L1-RSRPs of M beam pairs are arranged in a specified order.
  • the hidden layer consists of multiple fully connected layers.
  • the optimal receiving beam index output layer is part of a neural network layer of the hidden layer, and the optimal receiving beam index output layer is before the uplink feedback information output layer.
  • the optimal receiving beam index output The layer outputs K optimal receiving beam indexes, the uplink feedback information output layer outputs the first series of information, and the first information outputs the uplink feedback information after passing through the quantization layer.
  • the uplink feedback information is a bit sequence. This bit sequence does not explicitly indicate the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes, but implicitly indicates the K optimal transmit beam indexes.
  • the beam index and the L1-RSRP information corresponding to the K optimal transmit beam indexes can be extracted by the matching NW side sub-model.
  • the UE side sub-model and the NW side sub-model need to be paired for training and deployment, so that the NW can correctly predict the K optimal transmit beam indexes and the L1-RSRP corresponding to the K optimal transmit beam indexes from the uplink feedback information.
  • the optimal receiving beam index output layer in Figure 9-3 is part of a neural network layer of the hidden layer, it is not limited to this.
  • the optimal receiving beam index output layer can also be a part of the hidden layer. All neural network layers.
  • the optimal receiving beam index output layer in Figure 9-3 is adjacent to the uplink feedback information output layer, it is not limited to this.
  • the optimal receiving beam index output layer can also be separated from the uplink feedback information output layer by one or more nerves. Network layer.
  • the output of the optimal receiving beam index output layer may or may not act as (dashed line weight connection in Figure 9-3) to obtain part of the implicit information of the uplink feedback information.
  • the UE side sub-model includes an input layer, a hidden layer and an output layer.
  • the input of the input layer is a matrix in which the measured L1-RSRPs of M beam pairs are arranged in a specified order.
  • the hidden layer consists of multiple fully connected layers.
  • the output layer outputs K optimal receiving beam indices.
  • the L1-RSRPs of the M beam pairs of the input layer are used as the first information to pass through the quantization layer and then the uplink feedback information is output. After the uplink feedback information is reported to the NW side, it is used as the input of the NW side sub-model.
  • the uplink feedback information is directly obtained from the L1-RSRP of the M beam pairs in the input layer through the quantization layer, implicit feature extraction is not performed through any hidden layer. Therefore, the uplink feedback information is a bit sequence that explicitly indicates the measured L1-RSRP of the M beam pairs and is reported directly to the NW side.
  • the NW side sub-model obtains K from the measured L1-RSRP of the beam pairs. L1-RSRP corresponding to the optimal transmit beam index and the K optimal transmit beam indexes. Therefore, the UE side sub-model and the NW side sub-model do not need a strict pairing relationship and can be trained and deployed independently.
  • the input of the UE side sub-model is the L1-RSRP value of the direct measurement beam pair, and there is no quantization error; while the input of the NW side sub-model is obtained from the uplink feedback information, so there is a quantization error.
  • the UE side sub-model and NW side sub-model are trained offline. That is, the sub-models deployed on the UE side and NW side are models that have been adapted to the current cell or communication scenario.
  • the data set required by the model includes the following parts:
  • All beam pairs include transmit beams and receive beam pairs formed by all possible combinations of all receive beams of the UE and all transmit beams of the network device.
  • the output of the model is divided into the following two parts: the output of the UE side sub-model and the output of the NW side sub-model.
  • the output of the UE side sub-model is the K optimal receive beam indexes
  • the output of the NW side sub-model is the K optimal transmit beam indexes
  • the model can infer the situation of less than K optimal beam pairs.
  • the NW side saves the complete structure of the UE side sub-model and the NW side sub-model, and after the training is completed, the trained UE side sub-model (specifically including the structure and parameters of the UE side sub-model) ) is delivered to the UE through the downlink channel.
  • the training process is shown in Figure 10, including the following steps:
  • Step 1001 NW performs downlink beam scanning.
  • the NW scans all downlink beams (that is, all transmission beams of network equipment), such as the beams used by 64 SSB/CSI-RS.
  • the UE traverses all its receive beams to measure L1-RSRP, and finally obtains the L1-RSRP of all beam pairs.
  • the UE determines the indexes of the K beam pairs with the highest L1-RSRP (i.e., K optimal transmit beam indexes and K optimal receive beam indexes) and the corresponding K L1-RSRPs based on the L1-RSRPs of all beam pairs.
  • Step 1002 The UE reports the data set.
  • the data set includes the following two parts: the first part is the input, that is, the input of the UE side sub-model, specifically the L1-RSRP of some of the beam pairs measured among all beam pairs; the second part is the label, which is the UE
  • the labels of the side sub-model and the NW side sub-model are specifically the indexes of the K optimal transmit beams, the K optimal receive beam indexes and the corresponding K L1-RSRPs.
  • Step 1003 The NW trains the UE side sub-model and the NW side sub-model according to the data set reported by the UE.
  • the uplink feedback information is processed by the UE side sub-model and serves as the input of the NW side sub-model. Therefore, the UE side The sub-model and the NW-side sub-model must be jointly trained to ensure the matching of the UE-side sub-model and the NW-side sub-model.
  • the UE side sub-model and the NW side sub-model share the input layer. The UE side sub-model and the NW side sub-model can be trained separately, without Rigorous model matching.
  • Step 1004 The NW delivers the UE side sub-model.
  • the NW delivers the trained UE side sub-model to the UE through the downlink channel to complete the online deployment of the model.
  • the NW and UE can implement AI/ML-based beam management according to the aforementioned related solutions.
  • this example based on the NW side sub-model and UE side sub-model that were trained offline and deployed online, AI/ML-based beam management was implemented using online reasoning, and K optimal receive beam indexes and K optimal transmit beams were obtained. index and the L1-RSRP corresponding to the K optimal transmit beam indexes, where K is a positive integer.
  • this example also provides a beam indication method based on AI/ML, which uses necessary signaling support to assist the NW and UE side to cooperate in matching the optimal transmit beam and the optimal receive beam.
  • the UE side sub-model outputs an optimal receive beam index
  • the NW side sub-model outputs an optimal transmit beam index and the L1-RSRP corresponding to the optimal transmit beam index.
  • the UE selects an optimal receiving beam output by the UE side sub-model for receiving downlink transmission
  • the NW selects an optimal transmit beam output by the NW side sub-model for transmitting downlink transmission.
  • the optimal receive beam indicated by an optimal receive beam index output by the UE side sub-model and the optimal transmit beam indicated by an optimal transmit beam index output by the NW side sub-model automatically realize the optimal beam pair. Matching, the UE and NW side no longer need to indicate the beam used.
  • the UE side sub-model outputs K (K>1) optimal receive beam indexes, and the NW side sub-model outputs K optimal transmit beam indexes and L1-RSRP corresponding to the K optimal transmit beam indexes.
  • the NW selects an optimal transmit beam index from the K optimal transmit beam indexes and uses the optimal transmit beam indicated by the optimal transmit beam index for downlink transmission.
  • the NW indicates to the UE the used transmit beam index.
  • Optimal transmit beam index the UE can select the optimal receive beam index that matches the optimal transmit beam index used by the NW based on the corresponding relationship between the optimal transmit beam index and the optimal receive beam index, and then use the optimal receive beam index The indicated optimal receiving beam receives the downlink transmission.
  • the NW can indicate the optimal transmit beam index used by the NW through MAC CE or DCI.
  • the beam indication process is shown in Figure 11, including the following steps:
  • Step 1101 NW performs downlink beam scanning.
  • the NW scans all or part of the downlink beam (that is, all or part of the transmission beam of the network device).
  • the UE traverses all or part of the receive beams to measure the L1-RSRP, and finally obtains the L1-RSRP of some of the beam pairs among all the beam pairs.
  • Step 1102 UE side sub-model inference.
  • the UE infers (that is, predicts) the K optimal receiving beam indexes and uplink feedback information based on the L1-RSRP of some beam pairs through the UE side sub-model.
  • Step 1103 The UE reports uplink feedback information.
  • Step 1104 NW side sub-model inference.
  • the NW infers (that is, predicts) the K optimal transmit beam indexes and the corresponding K L1-RSRPs based on the uplink feedback information through the NW side sub-model.
  • the NW selects an optimal transmit beam index from the K optimal transmit beam indexes as the transmit beam index for downlink transmission.
  • Step 1105 The NW indicates the optimal transmit beam index to the UE.
  • the UE determines the corresponding optimal receive beam index according to the optimal transmit beam index indicated by the NW, and uses the receive beam indicated by the optimal receive beam index to receive downlink transmission.
  • the size of the sequence numbers of the above-mentioned processes does not mean the order of execution.
  • the execution order of each process should be determined by its functions and internal logic, and should not be used in this application.
  • the implementation of the examples does not constitute any limitations.
  • the terms “downlink”, “uplink” and “sidelink” are used to indicate the transmission direction of signals or data, where “downlink” is used to indicate that the transmission direction of signals or data is from the station.
  • uplink is used to indicate that the transmission direction of the signal or data is the second direction from the user equipment of the cell to the site
  • sidelink is used to indicate that the transmission direction of the signal or data is A third direction sent from User Device 1 to User Device 2.
  • downlink signal indicates that the transmission direction of the signal is the first direction.
  • the term “and/or” is only an association relationship describing associated objects, indicating that three relationships can exist. Specifically, A and/or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" in this article generally indicates that the related objects are an "or" relationship.
  • FIG 12 is a schematic structural diagram of the beam management device provided by the embodiment of the present application. It is applied to user equipment. As shown in Figure 12, the beam management device includes:
  • Measurement unit 1201 used to measure the link quality of multiple beams or beam pairs
  • Processing unit 1202 configured to use the first sub-model to determine at least one receive beam index and/or uplink feedback information based on the link quality of the multiple beams or beam pairs; wherein the uplink feedback information is used to determine at least one transmit The beam index and/or at least one link quality corresponding to the at least one transmit beam index.
  • the apparatus further includes: a sending unit 1203, configured to send the uplink feedback information to a network device, where the uplink feedback information is used by the network device to determine at least one transmission using the second sub-model.
  • the beam index and/or at least one link quality corresponding to the at least one transmit beam index are used by the network device to determine at least one transmission using the second sub-model.
  • the multiple beams are multiple receive beams or multiple transmit beams
  • the multiple beam pairs are multiple transmit beam and receive beam pairs.
  • the at least one receiving beam index is an index of K receiving beams with the best link quality among the multiple beams or beam pairs, and K is a positive integer.
  • the at least one transmit beam index is an index of K transmit beams with the best link quality among the multiple beams or beam pairs, and K is a positive integer.
  • the corresponding relationship is determined by the ordering of the at least one receive beam index, the order of the at least one transmit beam index, and the order of the at least one link quality.
  • the first sub-model includes an output layer; the processing unit 1202 is configured to use the first sub-model to process the link quality of the multiple beams or beam pairs, through the A first part of nodes in the output layer outputs at least one receive beam index, and a second part of nodes in the output layer outputs first information, where the first information is used to determine uplink feedback information.
  • the first sub-model includes a first output layer and a second output layer; the processing unit 1202 is configured to use the first sub-model to analyze the links of the multiple beams or beam pairs.
  • the quality is processed, at least one receive beam index is output through the first output layer, and first information is output through the second output layer, and the first information is used to determine uplink feedback information.
  • the first output layer is adjacent to the second output layer, or the first output layer and the second output layer are spaced apart. At least one neural network layer.
  • the first output layer is located before the second output layer, or the first output layer is located after the second output layer.
  • the second output layer is the last neural network layer of the first sub-model; the first output The layer is the middle neural network layer of the first sub-model.
  • the first output layer when the first output layer is located after the second output layer, the first output layer is the last neural network layer of the first sub-model; the second output The layer is the middle neural network layer of the first sub-model.
  • the first output layer when the first output layer is located after the second output layer, the first output layer is the last neural network layer of the first sub-model; the second output The layer is the first neural network layer of the first sub-model.
  • the input of the first neural network layer is the link quality of the multiple beams or beam pairs.
  • the first sub-model further includes a quantization layer
  • the processing unit 1202 is configured to perform quantization processing on the first information through the quantization layer to obtain the uplink feedback information.
  • the uplink feedback information is a first bit sequence
  • the first bit sequence is used to determine at least one transmit beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the uplink feedback information is a second bit sequence
  • the second bit sequence is used to indicate the link quality of the multiple beams or beam pairs
  • the link quality of the multiple beams or beam pairs is The link quality is used to determine at least one transmit beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the device further includes: a receiving unit 1204, configured to use the receiving beam indicated by the one receiving beam index to perform downlink when the at least one receiving beam index is one receiving beam index. Reception of transmission.
  • the device further includes: a receiving unit 1204, configured to determine the first receiving beam index from the multiple receiving beam indexes when the at least one receiving beam index is multiple receiving beam indexes. Receive beam index, use the receive beam indicated by the first receive beam index to receive downlink transmission.
  • the receiving unit 1204 is configured to receive first indication information sent by a network device, where the first indication information is used to indicate the first transmit beam index used by the network device; based on the The corresponding relationship between at least one receive beam index and the at least one transmit beam index determines the first receive beam index corresponding to the first transmit beam index.
  • the first indication information is carried in MAC CE or DCI.
  • the sending unit 1203 is used to report a data set to a network device, where the data set is used by the network device to train the first sub-model; the receiving unit 1204 is used to receive The trained first sub-model delivered by the network device.
  • the data set is also used by the network device to train a second sub-model, and the second sub-model is used by the network device to determine at least one transmit beam index and sum based on the uplink feedback information. /or at least one link quality corresponding to the at least one transmit beam index.
  • the first sub-model and the second sub-model are trained jointly; or, the first sub-model and the second sub-model are trained independently.
  • the data set includes:
  • Input data of the first sub-model the input data being the link quality of at least some beams or beam pairs in the first set of beams or beam pairs;
  • the label of the output data of the first sub-model is the index of the K receiving beams with the best link quality in the first beam or beam pair set;
  • the label of the output data of the second sub-model is the index of the K transmit beams with the best link quality in the first beam or beam pair set and the corresponding K link qualities, K is Positive integer.
  • the link quality includes at least one of the following: L1-RSRP, L1-SINR, and L1-RSRQ.
  • FIG 13 is a schematic diagram 2 of the structure of a beam management device provided by an embodiment of the present application. It is applied to network equipment. As shown in Figure 13, the beam management device includes:
  • Receiving unit 1301, configured to receive uplink feedback information sent by the UE
  • the processing unit 1302 is configured to use the second sub-model to determine at least one transmit beam index and/or at least one link quality corresponding to the at least one transmit beam index based on the uplink feedback information.
  • the uplink feedback information is a first bit sequence
  • the first bit sequence is used to determine at least one transmit beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the uplink feedback information is a second bit sequence
  • the second bit sequence is used to indicate the link quality of multiple beams or beam pairs
  • the link quality of the multiple beams or beam pairs The quality is used to determine at least one transmit beam index and at least one link quality corresponding to the at least one transmit beam index.
  • the device further includes: a sending unit 1303, configured to use the transmit beam indicated by the one transmit beam index for downlink when the at least one transmit beam index is one transmit beam index. The transmission is sent.
  • the device further includes: a sending unit 1303, configured to select a first transmit beam index from the plurality of transmit beam indexes when the at least one transmit beam index is a plurality of transmit beam indexes. Transmit beam index, use the transmit beam indicated by the first transmit beam index to send downlink transmission.
  • a sending unit 1303 configured to select a first transmit beam index from the plurality of transmit beam indexes when the at least one transmit beam index is a plurality of transmit beam indexes. Transmit beam index, use the transmit beam indicated by the first transmit beam index to send downlink transmission.
  • the sending unit 1303 is configured to send first indication information to the UE, where the first indication information is used to indicate the first transmit beam index used by the network device.
  • the first indication information is carried in MAC CE or DCI.
  • the receiving unit 1301 is configured to receive the data set reported by the UE; the processing unit 1302 is used to train the second sub-model using the data set.
  • the data set is also used for the network device to train a first sub-model, and the first sub-model is used for the UE to determine at least one based on the link quality of multiple beams or beam pairs. Receive beam index and/or uplink feedback information.
  • the at least one receiving beam index is an index of K receiving beams with the best link quality among the multiple beams or beam pairs, and K is a positive integer.
  • the at least one transmit beam index is an index of K transmit beams with the best link quality among the multiple beams or beam pairs, and K is a positive integer.
  • the corresponding relationship is determined by the ordering of the at least one receive beam index, the order of the at least one transmit beam index, and the order of the at least one link quality.
  • the first sub-model and the second sub-model are trained jointly; or, the first sub-model and the second sub-model are trained independently.
  • the sending unit 1303 is configured to deliver the trained first sub-model to the UE.
  • the data set includes:
  • Input data of the first sub-model the input data being the link quality of at least some beams or beam pairs in the first set of beams or beam pairs;
  • the label of the output data of the first sub-model is the index of the K receiving beams with the best link quality in the first beam or beam pair set;
  • the label of the output data of the second sub-model is the index of the K transmit beams with the best link quality in the first beam or beam pair set and the corresponding K link qualities, K is Positive integer.
  • the link quality includes at least one of the following: L1-RSRP, L1-SINR, and L1-RSRQ.
  • Figure 14 is a schematic structural diagram of a communication device 1400 provided by an embodiment of the present application.
  • the communication device can be user equipment or network equipment.
  • the communication device 1400 shown in Figure 14 includes a processor 1410.
  • the processor 1410 can call and run a computer program from the memory to implement the method in the embodiment of the present application.
  • the communication device 1400 may further include a memory 1420.
  • the processor 1410 can call and run the computer program from the memory 1420 to implement the method in the embodiment of the present application.
  • the memory 1420 may be a separate device independent of the processor 1410, or may be integrated into the processor 1410.
  • the communication device 1400 may also include a transceiver 1430, and the processor 1410 may control the transceiver 1430 to communicate with other devices. Specifically, it may send information or data to other devices, or receive other devices. Information or data sent by the device.
  • the transceiver 1430 may include a transmitter and a receiver.
  • the transceiver 1430 may further include an antenna, and the number of antennas may be one or more.
  • the communication device 1400 can be specifically a network device according to the embodiment of the present application, and the communication device 1400 can implement the corresponding processes implemented by the network device in the various methods of the embodiment of the present application. For the sake of brevity, details will not be described here. .
  • the communication device 1400 may specifically be the user equipment in the embodiment of the present application, and the communication device 1400 may implement the corresponding processes implemented by the user equipment in the various methods of the embodiment of the present application. For the sake of brevity, details will not be repeated here. .
  • Figure 15 is a schematic structural diagram of a chip according to an embodiment of the present application.
  • the chip 1500 shown in Figure 15 includes a processor 1510.
  • the processor 1510 can call and run a computer program from the memory to implement the method in the embodiment of the present application.
  • the chip 1500 may also include a memory 1520.
  • the processor 1510 can call and run the computer program from the memory 1520 to implement the method in the embodiment of the present application.
  • the memory 1520 may be a separate device independent of the processor 1510, or may be integrated into the processor 1510.
  • the chip 1500 may also include an input interface 1530.
  • the processor 1510 can control the input interface 1530 to communicate with other devices or chips. Specifically, it can obtain information or data sent by other devices or chips.
  • the chip 1500 may also include an output interface 1540.
  • the processor 1510 can control the output interface 1540 to communicate with other devices or chips. Specifically, it can output information or data to other devices or chips.
  • the chip can be applied to the network device in the embodiment of the present application, and the chip can implement the corresponding processes implemented by the network device in the various methods of the embodiment of the present application.
  • the details will not be described again.
  • the chip can be applied to the user equipment in the embodiment of the present application, and the chip can implement the corresponding processes implemented by the user equipment in the various methods of the embodiment of the present application.
  • the details will not be described again.
  • chips mentioned in the embodiments of this application may also be called system-on-chip, system-on-a-chip, system-on-chip or system-on-chip, etc.
  • Figure 16 is a schematic block diagram of a communication system 1600 provided by an embodiment of the present application. As shown in Figure 16, the communication system 1600 includes user equipment 1610 and network equipment 1620.
  • the user equipment 1610 can be used to implement the corresponding functions implemented by the user equipment in the above method
  • the network device 1620 can be used to implement the corresponding functions implemented by the network equipment in the above method.
  • no further details will be given here. .
  • the processor in the embodiment of the present application may be an integrated circuit chip and has signal processing capabilities.
  • each step of the above method embodiment can be completed through an integrated logic circuit of hardware in the processor or instructions in the form of software.
  • the above-mentioned processor can be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), an off-the-shelf programmable gate array (Field Programmable Gate Array, FPGA) or other available processors.
  • DSP Digital Signal Processor
  • ASIC Application Specific Integrated Circuit
  • FPGA Field Programmable Gate Array
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
  • the steps of the method disclosed in conjunction with the embodiments of the present application can be directly implemented by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other mature storage media in this field.
  • the storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
  • non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electrically removable memory. Erase programmable read-only memory (Electrically EPROM, EEPROM) or flash memory. Volatile memory may be Random Access Memory (RAM), which is used as an external cache.
  • RAM Random Access Memory
  • RAM static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM DDR SDRAM
  • enhanced SDRAM ESDRAM
  • Synchlink DRAM SLDRAM
  • Direct Rambus RAM Direct Rambus RAM
  • the memory in the embodiment of the present application can also be a static random access memory (static RAM, SRAM), a dynamic random access memory (dynamic RAM, DRAM), Synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous connection Dynamic random access memory (synch link DRAM, SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DR RAM) and so on. That is, memories in embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.
  • Embodiments of the present application also provide a computer-readable storage medium for storing computer programs.
  • the computer-readable storage medium can be applied to the network device in the embodiment of the present application, and the computer program causes the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiment of the present application. For the sake of simplicity, here No longer.
  • the computer-readable storage medium can be applied to the user equipment in the embodiment of the present application, and the computer program causes the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiment of the present application. For the sake of simplicity, here No longer.
  • An embodiment of the present application also provides a computer program product, including computer program instructions.
  • the computer program product can be applied to the network device in the embodiment of the present application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiment of the present application. For the sake of brevity, they are not included here. Again.
  • the computer program product can be applied to the user equipment in the embodiments of the present application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiments of the present application. For the sake of brevity, they are not included here. Again.
  • An embodiment of the present application also provides a computer program.
  • the computer program can be applied to the network device in the embodiment of the present application.
  • the computer program When the computer program is run on the computer, it causes the computer to execute the corresponding processes implemented by the network device in each method of the embodiment of the present application.
  • the computer program For the sake of simplicity , which will not be described in detail here.
  • the computer program can be applied to the user equipment in the embodiments of the present application.
  • the computer program When the computer program is run on the computer, it causes the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiments of the present application. For the sake of simplicity , which will not be described in detail here.
  • the disclosed systems, devices and methods can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented.
  • the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, indirect coupling or communication connection of devices or units, which may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application can be integrated into one processing unit, each unit can exist physically alone, or two or more units can be integrated into one unit.
  • the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product.
  • the computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory,) ROM, random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program code. .

Landscapes

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

Abstract

本申请实施例提供一种波束管理方法及装置、用户设备、网络设备,该方法包括:UE测量多个波束或波束对的链路质量;所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。网络设备接收UE发送的上行反馈信息;所述网络设备利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。

Description

一种波束管理方法及装置、用户设备、网络设备 技术领域
本申请实施例涉及移动通信技术领域,具体涉及一种波束管理方法及装置、用户设备、网络设备。
背景技术
模拟波束赋形技术,可以在增强网络覆盖的同时,降低设备的实现复杂度。模拟波束赋形不仅可以用于网络设备,也可以用于用户设备(User Equipment,UE)。
为了获得良好的通信质量,需要确定出最优发射波束和/或最优接收波束。目前的下行波束管理流程中,为了确定出最优发射波束和/或最优接收波束,UE需要遍历全部的发射波束和/或遍历全部的接收波束,来进行相应的链路质量测量,这将带来大量的开销和时延。
发明内容
本申请实施例提供一种波束管理方法及装置、用户设备、网络设备、芯片、计算机可读存储介质、计算机程序产品、计算机程序。
本申请实施例提供的波束管理方法,包括:
UE测量多个波束或波束对的链路质量;
所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
本申请实施例提供的波束管理方法,包括:
网络设备接收UE发送的上行反馈信息;
所述网络设备利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
本申请实施例提供的波束管理装置,应用于UE,所述装置包括:
测量单元,用于测量多个波束或波束对的链路质量;
处理单元,用于利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
本申请实施例提供的波束管理装置,应用于网络设备,所述装置包括:
接收单元,用于接收UE发送的上行反馈信息;
处理单元,用于利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
本申请实施例提供的用户设备,包括处理器和存储器。该存储器用于存储计算机程序,该处理器用于调用并运行该存储器中存储的计算机程序,执行上述的波束管理方法。
本申请实施例提供的网络设备,包括处理器和存储器。该存储器用于存储计算机程序,该处理器用于调用并运行该存储器中存储的计算机程序,执行上述的波束管理方法。
本申请实施例提供的芯片,用于实现上述的波束管理方法。
具体地,该芯片包括:处理器,用于从存储器中调用并运行计算机程序,使得安装有该芯片的设备执行上述的波束管理方法。
本申请实施例提供的计算机可读存储介质,用于存储计算机程序,该计算机程序使得计算机执行上述的波束管理方法。
本申请实施例提供的计算机程序产品,包括计算机程序指令,该计算机程序指令使得计算机执行上述的波束管理方法。
本申请实施例提供的计算机程序,当其在计算机上运行时,使得计算机执行上述的波束管理 方法。
本申请实施例的技术方案,在UE侧部署第一子模型,使得UE可以在测量少量波束或波束对的链路质量的情况下,通过第一子模型基于测量到的波束或波束对的链路质量预测(也即推断)出至少一个最优接收波束索引和/或上行反馈信息。另一方面,在网络设备侧部署第二子模型,使得网络设备可以通过第二子模型基于UE的上行反馈信息预测出至少一个最优发射波束索引和/或对应的至少一个链路质量。通过采用本申请实施例的技术方案,可以快速地确定出最优发射波束和/或最优接收波束,并显著地减小波束扫描过程中的资源开销和时延。
附图说明
此处所说明的附图用来提供对本申请的进一步理解,构成本申请的一部分,本申请的示意性实施例及其说明用于解释本申请,并不构成对本申请的不当限定。在附图中:
图1是本申请实施例的一个应用场景的示意图;
图2是一种神经元结构的示意图;
图3是一种深度神经网络的示意图;
图4是一种卷积神经网络的示意图;
图5是一种LSTM单元结构的示意图;
图6是一种下行波束扫描过程的示意图;
图7是本申请实施例提供的一种波束管理方法的流程示意图;
图8是本申请实施例提供的模型在UE侧和NW侧分离部署的示意图;
图9-1是本申请实施例提供的UE侧子模型的结构示意图一;
图9-2是本申请实施例提供的UE侧子模型的结构示意图二;
图9-3是本申请实施例提供的UE侧子模型的结构示意图三;
图9-4是本申请实施例提供的UE侧子模型的结构示意图四;
图10是本申请实施例提供的模型训练过程的示意图;
图11是本申请实施例提供的波束指示过程的示意图;
图12是本申请实施例提供的波束管理装置的结构组成示意图一;
图13是本申请实施例提供的波束管理装置的结构组成示意图二;
图14是本申请实施例提供的一种通信设备示意性结构图;
图15是本申请实施例的芯片的示意性结构图;
图16是本申请实施例提供的一种通信系统的示意性框图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
图1是本申请实施例的一个应用场景的示意图。如图1所示,通信系统可以包括用户设备110和网络设备120。网络设备120可以通过空口与用户设备110进行通信。
应理解,本申请实施例虽然以图1所示的通信系统进行示例性说明,但本申请实施例不限定于此。也就是说,本申请实施例的技术方案可以应用于各种通信系统,例如:5G通信系统(也称为新无线(New Radio,NR)通信系统),或未来的通信系统等。
在图1所示的通信系统中,网络设备120可以是与用户设备110通信的接入网设备。接入网设备可以为特定的地理区域提供通信覆盖,并且可以与位于该覆盖区域内的用户设备110进行通信。作为示例:接入网设备可以是NR系统中的基站(gNB)。
在图1所示的通信系统中,用户设备110可以是任意用户设备,例如手机、具有无线通信功能的手持设备、车载设备、可穿戴设备等。
图1所示的无线通信系统还可以包括与接入网设备进行通信的核心网设备130,在网络演进过程中,核心网设备130的名字可能有所不同。
需要说明的是,本文中术语“系统”和“网络”在本文中常被可互换使用。本文中术语“和/或”,仅仅是一种描述关联对象的关联关系,表示可以存在三种关系,例如,A和/或B,可以表示:单独 存在A,同时存在A和B,单独存在B这三种情况。另外,本文中字符“/”,一般表示前后关联对象是一种“或”的关系。还应理解,在本申请的实施例中提到的“指示”可以是直接指示,也可以是间接指示,还可以是表示具有关联关系。举例说明,A指示B,可以表示A直接指示B,例如B可以通过A获取;也可以表示A间接指示B,例如A指示C,B可以通过C获取;还可以表示A和B之间具有关联关系。还应理解,在本申请的实施例中提到的“对应”可表示两者之间具有直接对应或间接对应的关系,也可以表示两者之间具有关联关系,也可以是指示与被指示、配置与被配置等关系。还应理解,在本申请的实施例中提到的“预定义”或“预定义规则”可以通过在设备(例如,包括用户设备和网络设备)中预先保存相应的代码、表格或其他可用于指示相关信息的方式来实现,本申请对于其具体的实现方式不做限定。比如预定义可以是指协议中定义的。还应理解,本申请实施例中,所述"协议"可以指通信领域的标准协议,例如可以包括NR协议以及应用于未来的通信系统中的相关协议,本申请对此不做限定。
为便于理解本申请实施例的技术方案,以下对本申请实施例的相关技术进行说明,以下相关技术作为可选方案与本申请实施例的技术方案可以进行任意结合,其均属于本申请实施例的保护范围。
神经网络与机器学习
神经网络是一种由多个神经元(简称为节点)相互连接构成的运算模型,其中节点间的连接代表加权值,称为权重。一个节点可以与多个上一级节点连接,将多个上一级节点的输出作为该节点的输入,该节点利用权重对输入进行加权求和,并将得到的结果通过特定的激活函数进行处理。作为示例:神经元结构如图2所示,神经元利用n个权重(记作w1、w2…、wn)对n个输入(记作a1、a2、…、an)进行加权求和,并将得到的结果通过激活函数(记作f)进行处理,最终输出t;可选地,神经元在进行加权求和后,还可以在得到的结果上增加偏置b。
神经网络的类型有多种,深度神经网络(Deep Neural Network,DNN)是一种常见的神经网络。除了深度神经网络以外,神经网络的类型还有卷积神经网络(Convolutional Neural Network,CNN)、循环神经网络(Recurrent Neural Network,RNN)等。
图3是一种典型的深度神经网络,如图3所示,深度神经网络的基本结构包括:输入层、隐藏层和输出层,输入层的输入即为深度神经网络的输入,输出层的输出即为深度神经网络的输出。对于深度神经网络来说,通过设计不同的节点连接方式,不同的权重,不同的激活函数,可以使得深度神经网络产生不同的输出,进而拟合从输入到输出的映射关系。图3所示的深度神经网络属于全连接模型,每一个节点都与其全部的下一级节点相连。
图4是一种典型的卷积神经网络,如图4所示,卷积神经网络的基本结构包括:输入层、多个卷积层、多个池化层、全连接层及输出层,输入层的输入即为卷积神经网络的输入,输出层的输出即为卷积神经网络的输出。对于卷积神经网络来说,卷积层中卷积核的每个神经元与其输入进行局部连接,并通过引入池化层提取某一层局部的最大值或者平均值特征,有效减少了网络的参数,并挖掘了局部特征,使得卷积神经网络能够快速收敛,获得优异的性能。
循环神经网络是一种对序列数据建模的神经网络,在自然语言处理领域,如机器翻译、语音识别等应用取得显著成绩。具体表现为,循环神经网络对过去时刻的信息进行记忆,并用于当前输出的计算中,即隐藏层之间的节点不再是无连接的而是有连接的,并且隐藏层的输入不仅包括上一层的输入还包括上一时刻隐藏层的输出。循环神经网络一般包括长短期记忆人工神经网络(Long-Short Term Memory,LSTM)单元结构,图5是一种典型的LSTM单元结构,如图5所示,LSTM单元结构包括:tanh节点以及σ节点。不同于卷积神经网络只考虑最近的状态,LSTM的细胞状态会决定哪些状态应该被留下来,哪些状态应该被遗忘,解决了传统卷积神经网络在长期记忆上存在的缺陷。
需要说明的是,本申请实施例的技术方案中的神经网络可以是深度神经网络,但不局限于此,也可以将深度神经网络替换为卷积神经网络或者循环神经网络等来作为本申请实施例的技术方案中的神经网络。
波束管理
波束管理机制分为上行波束管理机制和下行波束管理机制。对于下行波束管理机制,包括下行波束扫描,UE进行最优波束上报,网络设备对于下行波束的指示等过程。需要说明的是,对于网络设备来说,下行波束是指网络设备的发射波束;对于UE来说,下行波束是指UE的接收波束。
如图6所示,下行波束扫描过程有3个过程,分别称为P1过程、P2过程和P3过程。对于P1过程,网络设备遍历全部的发射波束以及UE遍历全部的接收波束,从而可以遍历全部的波束对组合,UE针对每个波束对测量相应的链路质量。对于P2过程,UE的接收波束是特定的,网络设备遍历全部的发射波束,从而可以对特定的接收波束来遍历全部的发射波束,UE针对每个发射波束测 量相应的链路质量。对于P3过程,网络设备的发射波束是特定的,UE遍历全部的接收波束,从而可以对特定的发射波束来遍历全部的接收波束,UE针对每个接收波束测量相应的链路质量。
UE进行链路质量测量后,通过比较测量到的全部波束或者波束对的链路质量,选择链路质量最高的K个波束或者波束对(也即K个最优波束或波束对),然后通过上行控制信息上报给网络设备,K为正整数。
网络设备获取K个最优波束或波束对后,从K个最优波束或波束对中选择进行下行传输所使用的发射波束,并通过媒体介入控制控制元素(Media Access Control-Control Element,MAC CE)或下行控制信息(Downlink Control Information,DCI)中携带的传输配置指示(Transmission Configuration Indication,TCI)状态,来完成对发射波束的指示。UE使用该发射波束对应的接收波束来进行下行传输的接收。
对于上述波束扫描过程来说,需要遍历全部的发射波束和/或接收波束,因此会带来大量的开销和时延。举例来说,假设网络设备有64个发射波束,UE有4个接收波束,那么,对于P1过程,UE需要测量256(即64×4)个波束对,从而需要256个下行资源开销。从时间角度说,每个波束对的测量都需要一定时长来完成,256个波束对的测量需要的时长将会非常大。可见,目前的波束扫描过程,将带来大量的开销和时延。
本申请实施例的技术方案,使用人工智能/机器学习(Artificial Intelligence/Machine Learning,AI/ML)技术,通过构建数据集训练神经网络,并在UE侧和/或网络设备侧进行部署,使得UE可以在部分波束或波束对的测量结果作为神经网络的输入的情况下,通过神经网络预测(也即推断)出至少一个最优接收波束索引和/或上行反馈信息。网络设备通过神经网络根据UE上报的上行反馈信息预测出至少一个最优发射波束索引和/或对应的至少一个链路质量。如此,可以使得网络设备和UE快速地匹配出最优发射波束和/或最优接收波束,并显著地减小波束扫描过程中的资源开销和时延。
为便于理解本申请实施例的技术方案,以下通过具体实施例详述本申请的技术方案。以上相关技术作为可选方案与本申请实施例的技术方案可以进行任意结合,其均属于本申请实施例的保护范围。本申请实施例包括以下内容中的至少部分内容。
需要说明的是,本申请实施例中描述的波束指发射波束或接收波束,波束对指一对发射波束和接收波束。
需要说明的是,本申请实施例中描述的“模型”也可以称为“神经网络”,可以但不局限于是深度神经网络、卷积神经网络、循环神经网络。
图7是本申请实施例提供的一种波束管理方法的流程示意图,如图7所示,所述波束管理方法包括以下步骤:
步骤701:UE测量多个波束或波束对的链路质量。
步骤702:所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,UE测量的多个波束为所有波束的一个子集。这里,所有波束指全部的发射波束或者全部的接收波束,即网络设备全部的发射波束或者UE全部的接收波束。
在一些可选实施方式中,UE测量的波束对为所有波束对的一个子集。这里,所有波束对指全部的发射波束和接收波束对,即UE全部的接收波束和网络设备全部的发射波束进行所有可能的组合形成的发射波束和接收波束对。
本申请实施例中,波束的链路质量指发射波束对应的链路质量。波束对的链路质量指发射波束和接收波束对对应的链路质量。
本申请实施例中,针对波束或波束对的测量指针对波束或波束对关联的参考信号的测量。
在一些可选实施方式中,所述参考信号包括以下至少之一:同步信号块(Synchronization Signal and PBCH Block,SSB)、信道状态信息-参考信号(Channel State Information-Reference Signal,CSI-RS)。
在一些可选实施方式中,所述链路质量包括以下至少之一:层1-参考信号接收功率(Layer1-Reference Signal Receiving Power,L1-RSRP)、层1-信号与干扰加噪声比(Layer1-Signal to Interference plus Noise Ratio,L1-SINR)、层1-参考信号接收质量(Layer1-Reference Signal Receiving Quality,L1-RSRQ)。
本申请实施例中,UE测量多个波束或波束对的链路质量后,利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息。这里,所述多个波束为多个接 收波束或者多个发射波束,所述多个波束对为多个发射波束和接收波束对。
作为一种实现方式,UE测量多个波束(这里指多个发射波束)的链路质量后,利用第一子模型基于所述多个波束的链路质量确定上行反馈信息。
作为另一种实现方式,UE测量多个波束(这里指多个接收波束)的链路质量后,利用第一子模型基于所述多个波束的链路质量确定至少一个接收波束索引。
作为另一种实现方式,UE测量多个波束对(这里指多对发射波束和接收波束)的链路质量后,利用第一子模型基于所述多个波束对的链路质量确定至少一个接收波束索引和上行反馈信息。
在一些可选实施方式中,所述UE确定所述上行反馈信息的情况下,所述UE向网络设备发送所述上行反馈信息,相应地,网络设备接收UE发送的上行反馈信息,所述网络设备利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
这里,所述上行反馈信息用于所述网络设备利用第二子模型确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
上述方案中,所述至少一个接收波束索引为所述多个波束或波束对中的链路质量最好的K个接收波束的索引,K为正整数。这里,可以将所述至少一个接收波束索引记作K个最优接收波束索引。
上述方案中,所述至少一个发射波束索引为所述多个波束或波束对中的链路质量最好的K个发射波束的索引,K为正整数。这里,可以将所述至少一个发射波束索引记作K个最优发射波束索引。
这里,K的取值可以通过网络设备配置的或者是预定义的或者是基于UE自身实现确定的。
本申请实施例中,由于链路质量是波束或波束对的链路质量,因此,发射波束与链路质量之间具有对应关系,或者接收波束与链路质量之间具有对应关系,或者发射波束和接收波束对与链路质量之间具有对应关系。基于此,上述方案中的所述至少一个接收波束索引、所述至少一个发射波束索引以及所述至少一个链路质量中的任意两者之间是具有对应关系的。
在一些可选实施方式中,所述对应关系通过所述至少一个接收波束索引的排序、所述至少一个发射波束索引的排序以及所述至少一个链路质量的排序确定。
作为示例:UE侧的第一子模型输出K个最优接收波束索引,网络设备侧的第二子模型输出K个最优发射波束索引以及K个链路质量。K个最优接收波束索引记作{n1,n2,…nK},K个最优发射波束索引记作{m1,m2,…mK},K个链路质量记作{r1,r2,…,rK}。可选地,链路质量可以按照从高到低的顺序进行排序,即{r1>r2>…>rK}。K个最优接收波束索引、K个最优发射波束索引以及K个链路质量具有一一对应关系,该对应关系由K个最优接收波束索引的排序、K个最优发射波束索引的排序以及K个链路质量的排序隐式指示,即排序在相同位置的最优接收波束索引、最优发射波束索引以及链路质量是对应的,具体实现时,UE侧的第一子模型输出的K个最优接收波束索引按照对应的链路质量的排序(如从高到低的顺序)进行排序;网络设备侧的第二子模型输出的K个最优发射波束索引按照对应的链路质量的排序(如从高到低的顺序)进行排序。如此,K个最优接收波束索引、K个最优发射波束索引以及K个链路质量按照对应关系可表示为一个三元组集合,记为{(m1,n1,r1),(m2,n2,r2),…,(mK,nK,rK)}。需要指明的是,{n1,n2,…nK}内的元素之间不存在互斥关系,即K个最优接收波束索引内的最优接收波束索引可以是相同的,也可以是不同的;同理,{m1,m2,…mK}内的元素之间不存在互斥关系,即K个最优发射波束索引内的最优发射波束索引可以是相同的,也可以是不同的;而{(m1,n1),(m2,n2),…,(mK,nK)}内的元素之间存在互斥关系,即K个最优波束对内的最优波束对是两两不同的。
本申请实施例中,UE侧的第一子模型的输入为所述UE测量的多个波束或波束对的链路质量。UE侧的第一子模型的输出包括至少一个接收波束索引和/或上行反馈信息。
对于UE侧的第一子模型的输出为至少一个接收波束索引的情况,所述第一子模型具有一个输出层,所述UE通过所述输出层输出至少一个接收波束索引。
对于UE侧的第一子模型的输出为上行反馈信息的情况,所述第一子模型具有一个输出层,所述UE通过所述输出层输出第一信息,所述第一信息用于确定上行反馈信息。进一步,所述第一子模型还包括量化层,所述UE通过所述量化层对所述第一信息进行量化处理,得到所述上行反馈信息。
对于UE侧的第一子模型的输出包括至少一个接收波束索引和上行反馈信息的情况,所述第一子模型的输出层可以有如下实现方式。
方式A:所述第一子模型包括输出层;所述UE利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述输出层的第一部分节点输出至少一个接收波束索引,以及通过所述输出 层的第二部分节点输出第一信息,所述第一信息用于确定上行反馈信息。进一步,所述第一子模型还包括量化层,所述UE通过所述量化层对所述第一信息进行量化处理,得到所述上行反馈信息。
方式B:所述第一子模型包括第一输出层和第二输出层;所述UE利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述第一输出层输出至少一个接收波束索引,以及通过所述第二输出层输出第一信息,所述第一信息用于确定上行反馈信息。进一步,所述第一子模型还包括量化层,所述UE通过所述量化层对所述第一信息进行量化处理,得到所述上行反馈信息。
这里,在所述第一子模型中,所述第一输出层与所述第二输出层相邻,或者所述第一输出层与所述第二输出层之间间隔至少一个神经网络层。
这里,在所述第一子模型中,所述第一输出层位于所述第二输出层之前,或者所述第一输出层位于所述第二输出层之后。
方案B-1)在一些可选实施方式中,所述第一输出层位于所述第二输出层之前的情况,所述第二输出层为所述第一子模型的最后一个神经网络层;所述第一输出层为所述第一子模型的中间一个神经网络层。
这里,所述第一输出层可以是第一子模型的中间一个神经网络层的全部或者部分。以深度神经网络为例,第一子模型的中间一个神经网络层是指隐藏层。
方案B-2)在一些可选实施方式中,所述第一输出层位于所述第二输出层之后的情况,所述第一输出层为所述第一子模型的最后一个神经网络层;所述第二输出层为所述第一子模型的中间一个神经网络层。
这里,所述第二输出层可以是第一子模型的中间一个神经网络层的全部或者部分。以深度神经网络为例,第一子模型的中间一个神经网络层是指隐藏层。
方案B-3)在一些可选实施方式中,所述第一输出层位于所述第二输出层之后的情况,所述第一输出层为所述第一子模型的最后一个神经网络层;所述第二输出层为所述第一子模型的第一个神经网络层。
这里,所述第二输出层可以是第一个神经网络层的全部。以深度神经网络为例,第一子模型的第一个神经网络层是指输入层,所述第一个神经网络层的输入为所述多个波束或波束对的链路质量。
上述方案中,所述第一输出层和所述第二输出层之间具有关联,或者所述第一输出层和所述第二输出层之间不具有关联。
这里,所述第一输出层和所述第二输出层之间具有关联的含义是:所述第一输出层的节点与所述第二输出层的节点之间具有关联。这里,所述第一输出层的节点与所述第二输出层的节点之间具有关联是指:所述第一输出层的节点的输出与所述第二输出层的节点的输入具有关联(这是针对所述第一输出层位于所述第二输出层之前的情况),或者,所述第二输出层的节点的输出与所述第一输出层的节点的输入具有关联(这是针对所述第二输出层位于所述第一输出层之前的情况)。需要指出的是,这种关联可以是直接关联(对应两层相邻的情况)或者间接关联(对应两层不相邻的情况)。此外,可以是所述第一输出层的全部节点或部分节点与所述第二输出层的全部节点或部分节点之间具有关联。
本申请实施例中,UE通过量化层量化后得到的上行反馈信息可以是一个比特序列。
在一些可选实施方式中,对于上述方案A、方案B-1)和方案B-2)的情况,所述上行反馈信息为第一比特序列,所述第一比特序列用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,对于上述方案B-3)的情况,所述上行反馈信息为第二比特序列,所述第二比特序列用于指示所述多个波束或波束对的链路质量,所述多个波束或波束对的链路质量用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述至少一个接收波束索引为一个接收波束索引的情况,所述UE采用所述一个接收波束索引所指示的接收波束进行下行传输的接收。
在一些可选实施方式中,所述至少一个发射波束索引为一个发射波束索引的情况,所述网络设备采用所述一个发射波束索引所指示的发射波束进行下行传输的发送。
在一些可选实施方式中,所述至少一个接收波束索引为多个接收波束索引的情况,所述UE从所述多个接收波束索引中确定出第一接收波束索引,采用所述第一接收波束索引所指示的接收波束进行下行传输的接收。
这里,所述UE接收网络设备发送的第一指示信息,所述第一指示信息用于指示所述网络设备采用的第一发射波束索引;所述UE基于所述至少一个接收波束索引和所述至少一个发射波束索引 的对应关系,确定所述第一发射波束索引对应的第一接收波束索引。可选地,所述第一指示信息携带在MAC CE或者DCI中。
在一些可选实施方式中,所述至少一个发射波束索引为多个发射波束索引的情况,所述网络设备从所述多个发射波束索引中选择出第一发射波束索引,采用所述第一发射波束索引所指示的发射波束进行下行传输的发送。
这里,所述网络设备向所述UE发送第一指示信息,所述第一指示信息用于指示所述网络设备采用的第一发射波束索引。可选地,所述第一指示信息携带在MAC CE或者DCI中。
上述方案中的第一子模型是预先训练好的,可以通过所述网络设备训练所述第一子模型,并将训练好的所述第一子模型下发给所述UE。
同样,上述方案中的第二子模型也是预先训练好的,可以通过所述网络设备训练所述第二子模型。
当然,不局限于由所述网络设备训练所述第一子模型和/或所述第二子模型,也可以由其他具有处理能力的设备训练所述第一子模型和/或所述第二子模型。
在一些可选实施方式中,由网络设备训练所述第一子模型和/或所述第二子模型,包括:所述UE向网络设备上报数据集;所述网络设备接收所述UE上报的数据集,所述网络设备利用所述数据集训练所述第一子模型和/或所述第二子模型,这里,所述第一子模型用于所述UE基于多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息,所述第二子模型用于所述网络设备基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
这里,所述UE向网络设备上报的数据集用于网络设备训练所述第一子模型和/或所述第二子模型。
作为一种实现方式,所述数据集用于网络设备训练所述第一子模型或者所述第一子模型加所述第二子模型(或者说所述网络利用所述数据集训练所述第一子模型或者所述第一子模型加所述第二子模型)的情况下,所述网络设备向所述UE下发训练好的所述第一子模型;所述UE接收所述网络设备下发的训练好的所述第一子模型。
在一些可选实施方式中,所述数据集包括:
所述第一子模型的输入数据,所述输入数据为第一波束或波束对集合中的至少部分波束或波束对的链路质量;
所述第一子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个接收波束的索引;
所述第二子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个发射波束的索引以及对应的K个链路质量,K为正整数。
这里,第一波束或波束对集合是指第一波束集合或第一波束对集合。这里,第一波束集合包括全部的发射波束或者全部的接收波束,即网络设备全部的发射波束或者UE全部的接收波束。第一波束对集合包括全部的发射波束和接收波束对,即UE全部的接收波束和网络设备全部的发射波束进行所有可能的组合形成的发射波束和接收波束对。
本申请实施例中,所述第一子模型和所述第二子模型联合训练;或者,所述第一子模型和所述第二子模型独立训练。
这里,在对第一子模型训练和第二子模型训练联合训练的过程中,第一子模型和第二子模型共用一个损失函数,该损失函数可以理解为第一子模型和第二子模型的联合损失函数,通过该联合损失函数共同优化这两个子模型的权重。具体地,通过联合损失函数计算整个模型的输出数据与标签之间的损失值,判断损失值是否满足预设条件,若不满足,则根据损失值对第一子模型和第二子模型进行更新(包括但不限于更新第一子模型和第二子模型的权重),并不断迭代执行上述步骤,直到损失值满足预设条件,完成第一子模型和第二子模型的联合训练。这里,整个模型的输出数据包括K个接收波束的索引、K个发射波束的索引以及对应的K个链路质量。
这里,对第一子模型训练和第二子模型独立训练的过程中,第一子模型和第二子模型拥有独立的损失函数,通过两个损失函数分别优化这两个子模型的权重。具体地,通过第一损失函数计算第一子模型的输出数据与所述第一子模型的输出数据的标签之间的损失值,判断损失值是否满足预设条件,若不满足,则根据损失值对第一子模型进行更新(包括但不限于更新第一子模型的权重),并不断迭代执行上述步骤,直到损失值满足预设条件,完成第一子模型的训练。同样,通过第二损失函数计算第二子模型的输出数据与所述第二子模型的输出数据的标签之间的损失值,判断损失值是 否满足预设条件,若不满足,则根据损失值对第二子模型进行更新(包括但不限于更新第二子模型的权重),并不断迭代执行上述步骤,直到损失值满足预设条件,完成第二子模型的训练。这里,第一子模型的输出数据包括K个接收波束的索引,第二子模型的输出数据包括K个发射波束的索引以及对应的K个链路质量。
在一些可选实施方式中,对于上述方案A、方案B-1)和方案B-2)的情况,由于第二子模型的输入(即上行反馈信息)是由第一子模型对多个波束或波束对的链路质量进行处理后(这里指除了量化层以外的处理)得到的,因而第二子模型与第一子模型需要配对进行训练,也即所述第一子模型和所述第二子模型需要联合训练。此外,对于这种情况,所述第一子模型和所述第二子模型也需要配对部署。
在一些可选实施方式中,对于上述方案B-3)的情况,由于第二子模型的输入(即上行反馈信息)无需由第一子模型对多个波束或波束对的链路质量进行处理后(这里指除了量化层以外的处理)得到,因而第二子模型与第一子模型不需要配对进行训练,所述第一子模型和所述第二子模型可以独立训练。此外,对于这种情况,所述第一子模型和所述第二子模型也可以独立部署。
本申请实施例的技术方案,实现了基于AI/ML的波束管理,在网络设备设备和或UE侧分别部署训练好的神经网络,该波束管理过程包括以下至少之一:UE测量过程、UE推理(也即预测)过程、UE上报和网络设备推理(也即预测)过程。对于UE测量过程,UE测量所有波束或波束对中的部分波束或波束对;对于UE推理(也即预测)过程,UE通过本地的神经网络(也即第一子模型)根据测量结果预测最优接收波束索引和/或上行反馈信息;对于UE上报和网络设备推理(也即预测)过程,UE向网络设备上报上行反馈信息,网络设备通过本地的神经网络(也即第二子模型)根据上行反馈信息预测最优发射波束及对应的链路质量。通过该方案,可以使得网络设备和UE选择到信道质量较好的波束或波束对上,并且减少大量的波束扫描过程中产生的开销和时延。
以下结合具体实例对本申请实施例的上述技术方案中的全部方案或部分方案进行解释说明。
需要说明的是,以下实例中是以“波束对”作为例子进行说明的,波束对指一对发射波束和接收波束。
需要说明的是,以下实例中是以链路质量为“L1-RSRP”作为例子进行说明的,当然不局限于此,链路质量还可以是L1-SINR、L1-RSRQ。
需要说明的是,以下实例中的网络(NW)对应于上述方案中的“网络设备”。
实例一
本实例中,模型在UE侧和NW侧分离部署。如图8所示,模型(虚线框所包含的部分)包含UE侧子模型(也即第一子模型)和NW侧子模型(也即第二子模型),其中,UE侧子模型的输入为UE所测量波束对的L1-RSRP,该所测量波束对为所有波束对的一个子集,这里,所有波束对指全部的发射波束和接收波束对,即UE全部的接收波束和NW全部的发射波束进行所有可能的组合形成的发射波束和接收波束对。UE侧子模型输出K个最优接收波束索引,同时UE侧输出上行反馈信息,该上行反馈信息可通过MAC CE(对应PUSCH)或者DCI(对应PDCCH)上报至NW侧。NW侧子模型的输入为上行反馈信息,输出为K个最优发射波束索引及K个最优发射波束索引对应的L1-RSRP,K为正整数。
情况1)K=1
UE侧子模型输出1个最优接收波束索引,NW侧子模型输出1个最优发射波束索引及该最优发射波束索引对应的L1-RSRP。
这种情况,UE选择UE侧子模型输出的1个最优接收波束用于下行传输的接收,NW选择NW侧子模型输出的1个最优发射波束用于下行传输的发送。这里,UE侧子模型输出的1个最优接收波束索引所指示的最优接收波束和NW侧子模型输出的1个最优发射波束索引所指示的最优发射波束自动实现最优波束对的匹配,不再需要UE和NW侧对所使用的波束进行指示。该最优波束对的链路质量由NW侧子模型输出的L1-RSRP给出。
情况2)K>1
UE侧子模型输出K(K>1)个最优接收波束索引,NW侧子模型输出K个最优发射波束索引及K个最优发射波束索引对应的L1-RSRP。
这种情况,UE侧子模型的输出的K个最优接收波束索引(这里用n表示)与NW侧子模型输出的K个最优发射波束索引(这里用m表示)及L1-RSRP(这里用符号r表示)包含一一配对的隐含关系,即K个波束对及对应L1-RSRP可表示为一个三元组集合{(m1,n1,r1),(m2,n2,r2),…,(mK,nK,rK)},其中{m1,m2,…mK}表示K个最优发射波束索引,{n1,n2,…nK} 表示K个最优接收波束索引,{r1,r2,…,rK}表示K个最优发射波束索引对应的L1-RSRP,可选地,{r1>r2>…>rK}。最优发射波束索引和最优接收波束索引的配对关系(也即对应关系)由UE侧子模型和NW侧子模型的输出的波束索引序列的顺序隐式指示,而不通过上行反馈信息显性指示,即UE侧子模型输出的K个最优接收波束索引以及NW侧子模型输出的K个最优发射波束索引按照对应的L1-RSRP的降序进行排列。
实例二
本实例中,UE侧子模型(也即第一子模型)的结构可以有多种实现方式。以下给出几种UE侧子模型的结构。需要说明的是,本实例采用深度神经网络的实现方式对UE侧子模型的结构进行说明,但不局限于此,其他如卷积神经网络、循环神经网络等实现方式也可以实现UE侧子模型的结构。采用其他神经网络的实现方法,可在基于本实例并采用相同输入/输出接口的基础上进行扩展。
作为一种实现方式,如图9-1所示,UE侧子模型包含输入层,隐藏层与输出层。第一信息与K个最优接收波束索引在输出层并行输出,且共用所有的隐藏层权重。输入层的输入为测量到的M个波束对的L1-RSRP按照规定顺序排列成的矩阵。隐藏层由多个全连接层构成。输出层由两部分拼接组成,其中输出层的K个节点输出K个最优接收波束索引,输出层的其他节点输出第一信息,第一信息经过量化层后输出上行反馈信息。上行反馈信息上报至NW侧后,作为NW侧子模型的输入。这里,上行反馈信息为一段比特序列,该比特序列并不显式地指示K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP,而是隐含了K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP的信息,该信息可由匹配的NW侧子模型进行提取。因此,UE侧子模型和NW侧子模型需要配对进行训练和部署,使得NW能够从上行反馈信息中正确的预测出K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP。
作为一种实现方式,如图9-2所示,UE侧子模型包含输入层,隐藏层、上行反馈信息输出层和最优接收波束索引输出层。上行反馈信息输出层和最优接收波束索引输出层不在同一层,只共用部分隐藏层权重。输入层的输入为测量到的M个波束对的L1-RSRP按照规定顺序排列成的矩阵。隐藏层由多个全连接层构成,上行反馈信息输出层是隐藏层的一个神经网络层的部分,且上行反馈信息输出层在最优接收波束索引输出层之前,最优接收波束索引输出层输出K个最优接收波束索引,上行反馈信息输出层输出第一系信息,第一信息经过量化层后输出上行反馈信息。上行反馈信息上报至NW侧后,作为NW侧子模型的输入。这里,上行反馈信息为一段比特序列,该比特序列并不显式地指示K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP,而是隐含了K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP的信息,该信息可由匹配的NW侧子模型进行提取。因此,UE侧子模型和NW侧子模型需要配对进行训练和部署,使得NW能够从上行反馈信息中正确的预测出K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP。需要指出的是,虽然图9-2中的上行反馈信息输出层是隐藏层的一个神经网络层的部分,但不局限于此,上行反馈信息输出层还可以是隐藏层的一个神经网络层的全部。虽然图9-2中的上行反馈信息输出层与最优接收波束索引输出层相邻,但不局限于此,上行反馈信息输出层还可以与最优接收波束索引输出层间隔一个或多个神经网络层。此外,上行反馈信息输出层的输出可作为或者不作为(图9-2中的虚线权重连接)获取K个最优接收波束索引的部分隐含信息。
作为一种实现方式,如图9-3所示,UE侧子模型包含输入层,隐藏层、最优接收波束索引输出层和上行反馈信息输出层。最优接收波束索引输出层和上行反馈信息输出层不在同一层,只共用部分隐藏层权重。输入层的输入为测量到的M个波束对的L1-RSRP按照规定顺序排列成的矩阵。隐藏层由多个全连接层构成,最优接收波束索引输出层是隐藏层的一个神经网络层的部分,且最优接收波束索引输出层在上行反馈信息输出层之前,最优接收波束索引输出层输出K个最优接收波束索引,上行反馈信息输出层输出第一系信息,第一信息经过量化层后输出上行反馈信息。上行反馈信息上报至NW侧后,作为NW侧子模型的输入。这里,上行反馈信息为一段比特序列,该比特序列并不显式地指示K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP,而是隐含了K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP的信息,该信息可由匹配的NW侧子模型进行提取。因此,UE侧子模型和NW侧子模型需要配对进行训练和部署,使得NW能够从上行反馈信息中正确的预测出K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP。需要指出的是,虽然图9-3中的最优接收波束索引输出层是隐藏层的一个神经网络层的部分,但不局限于此,最优接收波束索引输出层还可以是隐藏层的一个神经网络层的全部。虽然图9-3中的最优接收波束索引输出层与上行反馈信息输出层相邻,但不局限于此,最优接收波束索引输出层还可以与上行反馈信息输出层间隔一个或多个神经网络层。此外,最优接收波束索引输出层的输出可作为 或者不作为(图9-3中的虚线权重连接)获取上行反馈信息的部分隐含信息。
作为一种实现方式,如图9-4所示,UE侧子模型包含输入层,隐藏层与输出层。输入层的输入为测量到的M个波束对的L1-RSRP按照规定顺序排列成的矩阵。隐藏层由多个全连接层构成。输出层输出K个最优接收波束索引。输入层的M个波束对的L1-RSRP作为第一信息经过量化层后输出上行反馈信息。上行反馈信息上报至NW侧后,作为NW侧子模型的输入。由于上行反馈信息直接由输入层的M个波束对的L1-RSRP经量化层获取,并不经过任何隐藏层进行隐式特征提取。因此,上行反馈信息为一段比特序列,该序列显示地指示测量的M个波束对的L1-RSRP,并直接上报至NW侧,由NW侧子模型从测量的波束对的L1-RSRP中获取K个最优发射波束索引和K个最优发射波束索引对应的L1-RSRP。因此,UE侧子模型和NW侧子模型不需要严格的配对关系,可以独立进行训练和部署。需要注意的是,UE侧子模型的输入为直接的测量波束对的L1-RSRP值,不存在量化误差;而NW侧子模型的输入由于从上行反馈信息中获取,因此存在量化误差。
实例三
本实例中,离线训练UE侧子模型和NW侧子模型。即部署于UE侧和NW侧的子模型是已经适配当前小区或通信场景下的模型。
在离线训练过程中,模型所需的数据集包括如下部分:
输入:所有波束对中的部分波束对的L1-RSRP。这里,所有波束对包括UE全部的接收波束和网络设备全部的发射波束进行所有可能的组合形成的发射波束和接收波束对。
标签:所有波束对中L1-RSRP最高的K个波束对的索引(即K个最优发射波束索引和K个最优接收波束索引)及对应的L1-RSRP。
此外,模型的输出分为以下两部分:UE侧子模型的输出、NW侧子模型的输出。其中,UE侧子模型的输出为K个最优接收波束索引;NW侧子模型的输出为K个最优发射波束索引以及K个最优发射波束索引对应的L1-RSRP。
需要说明的是,如果输出K个最优的波束对,那么该模型能够推断小于K个的最优波束对的情况。
考虑到NW拥有更强的算力,因此在NW侧完成离线训练任务。具体的,在离线训练过程中,NW侧保存UE侧子模型与NW侧子模型的完整结构,并在训练完成后,将训练好的UE侧子模型(具体包括UE侧子模型的结构和参数)通过下行信道下发给UE。具体地,训练过程如图10所示,包括以下步骤:
步骤1001:NW进行下行波束扫描。
这里,NW在进行下行波束扫描的时候,扫描全部的下行波束(也即网络设备全部的发射波束),如64个SSB/CSI-RS所用的波束。UE针对每个下行波束(也即发射波束),遍历其全部接收波束来进行L1-RSRP的测量,最终获取所有波束对的L1-RSRP。UE根据所有波束对的L1-RSRP确定L1-RSRP最高的K个波束对的索引(即K个最优发射波束索引和K个最优接收波束索引)及对应的K个L1-RSRP。
步骤1002:UE进行数据集上报。
这里,数据集包括以下两部分:第一部分为输入,也即UE侧子模型的输入,具体为测量到的所有波束对中的部分波束对的L1-RSRP;第二部分是标签,也即UE侧子模型和NW侧子模型的标签,具体为K个最优发送波束的索引,K个最优接收波束索引以及对应的K个L1-RSRP。
步骤1003:NW根据UE上报的数据集对UE侧子模型和NW侧子模型进行训练。
这里,对于一些UE侧子模型的结构(如图9-1至图9-3所示的结构),上行反馈信息由UE侧子模型处理得到,并作为NW侧子模型的输入,因此UE侧子模型和NW侧子模型必须联合训练,以保证UE侧子模型和NW子模型的匹配。对于另一些UE侧子模型的结构(如图9-4所示的结构),UE侧子模型和NW侧子模型共用输入层,UE侧子模型和NW侧子模型可以进行分离训练,不需要严格的模型匹配。
步骤1004:NW进行UE侧子模型下发。
这里,NW将训练好的UE侧子模型通过下行信道下发至UE,完成模型的在线部署。之后,NW和UE可按照前述相关方案实现基于AI/ML的波束管理。
实例四
本实例中,基于离线训练并在线部署的NW侧子模型和UE侧子模型,采用在线推理的方法实现基于AI/ML的波束管理,得到K个最优接收波束索引、K个最优发射波束索引及K个最优发射波束索引对应的L1-RSRP,K为正整数。从外,本实例也给出基于AI/ML的波束指示方法,通过必 要的信令支撑以辅助NW和UE侧协作进行最优发射波束和最优接收波束的匹配。
情况1)K=1
UE侧子模型输出1个最优接收波束索引,NW侧子模型输出1个最优发射波束索引及该最优发射波束索引对应的L1-RSRP。
这种情况,UE选择UE侧子模型输出的1个最优接收波束用于下行传输的接收,NW选择NW侧子模型输出的1个最优发射波束用于下行传输的发送。这里,UE侧子模型输出的1个最优接收波束索引所指示的最优接收波束和NW侧子模型输出的1个最优发射波束索引所指示的最优发射波束自动实现最优波束对的匹配,不再需要UE和NW侧对所使用的波束进行指示。
情况2)K>1
UE侧子模型输出K(K>1)个最优接收波束索引,NW侧子模型输出K个最优发射波束索引及K个最优发射波束索引对应的L1-RSRP。
这种情况,NW从K个最优发射波束索引中选择一个最优发射波束索引,并使用该最优发射波束索引所指示的最优发射波束进行下行传输的发送,NW向UE指示所使用的最优发射波束索引,UE可根据最优发射波束索引和最优接收波束索引的对应关系选择与NW使用的最优发射波束索引相匹配的最优接收波束索引,进而使用该最优接收波束索引所指示的最优接收波束进行下行传输的接收。这里,NW可以通过MAC CE或DCI来指示NW所使用的最优发射波束索引。对于K个最优接收波束索引来说,如果想要从中指示出一个最优接收波束索引,需要采用log2K个比特进行指示。例如当K=4时,需要2比特进行指示。具体地,波束指示流程如图11所示,包括以下步骤:
步骤1101:NW进行下行波束扫描。
这里,NW在进行下行波束扫描的时候,扫描全部或部分的下行波束(也即网络设备全部或部分的发射波束)。UE针对全部或部分的下行波束(也即发射波束),遍历其全部或部分接收波束来进行L1-RSRP的测量,最终获取所有波束对中的部分波束对的L1-RSRP。
步骤1102:UE侧子模型推理。
这里,UE通过UE侧子模型根据部分波束对的L1-RSRP推理(也即预测)出K个最优接收波束索引和上行反馈信息。
步骤1103:UE进行上行反馈信息的上报。
步骤1104:NW侧子模型推理。
这里,NW通过NW侧子模型根据上行反馈信息推理(也即预测)出K个最优发射波束索引及对应的K个L1-RSRP。K>1的情况,NW从K个最优发射波束索引中选择一个最优发射波束索引作为用于下行传输的发射波束索引。
步骤1105:NW向UE指示最优发射波束索引。
这里,UE根据NW指示的最优发射波束索引确定出对应的最优接收波束索引,使用该最优接收波束索引所指示的接收波束进行下行传输的接收。
以上结合附图详细描述了本申请的优选实施方式,但是,本申请并不限于上述实施方式中的具体细节,在本申请的技术构思范围内,可以对本申请的技术方案进行多种简单变型,这些简单变型均属于本申请的保护范围。例如,在上述具体实施方式中所描述的各个具体技术特征,在不矛盾的情况下,可以通过任何合适的方式进行组合,为了避免不必要的重复,本申请对各种可能的组合方式不再另行说明。又例如,本申请的各种不同的实施方式之间也可以进行任意组合,只要其不违背本申请的思想,其同样应当视为本申请所公开的内容。又例如,在不冲突的前提下,本申请描述的各个实施例和/或各个实施例中的技术特征可以和现有技术任意的相互组合,组合之后得到的技术方案也应落入本申请的保护范围。
还应理解,在本申请的各种方法实施例中,上述各过程的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本申请实施例的实施过程构成任何限定。此外,在本申请实施例中,术语“下行”、“上行”和“侧行”用于表示信号或数据的传输方向,其中,“下行”用于表示信号或数据的传输方向为从站点发送至小区的用户设备的第一方向,“上行”用于表示信号或数据的传输方向为从小区的用户设备发送至站点的第二方向,“侧行”用于表示信号或数据的传输方向为从用户设备1发送至用户设备2的第三方向。例如,“下行信号”表示该信号的传输方向为第一方向。另外,本申请实施例中,术语“和/或”,仅仅是一种描述关联对象的关联关系,表示可以存在三种关系。具体地,A和/或B可以表示:单独存在A,同时存在A和B,单独存在B这三种情况。另外,本文中字符“/”,一般表示前后关联对象是一种“或”的关系。
图12是本申请实施例提供的波束管理装置的结构组成示意图一,应用于用户设备,如图12所 示,所述波束管理装置包括:
测量单元1201,用于测量多个波束或波束对的链路质量;
处理单元1202,用于利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述装置还包括:发送单元1203,用于向网络设备发送所述上行反馈信息,所述上行反馈信息用于所述网络设备利用第二子模型确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述多个波束为多个接收波束或者多个发射波束,所述多个波束对为多个发射波束和接收波束对。
在一些可选实施方式中,所述至少一个接收波束索引为所述多个波束或波束对中的链路质量最好的K个接收波束的索引,K为正整数。
在一些可选实施方式中,所述至少一个发射波束索引为所述多个波束或波束对中的链路质量最好的K个发射波束的索引,K为正整数。
在一些可选实施方式中,所述至少一个接收波束索引、所述至少一个发射波束索引以及所述至少一个链路质量中的任意两者之间具有对应关系。
在一些可选实施方式中,所述对应关系通过所述至少一个接收波束索引的排序、所述至少一个发射波束索引的排序以及所述至少一个链路质量的排序确定。
在一些可选实施方式中,所述第一子模型包括输出层;所述处理单元1202,用于利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述输出层的第一部分节点输出至少一个接收波束索引,以及通过所述输出层的第二部分节点输出第一信息,所述第一信息用于确定上行反馈信息。
在一些可选实施方式中,所述第一子模型包括第一输出层和第二输出层;所述处理单元1202,用于利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述第一输出层输出至少一个接收波束索引,以及通过所述第二输出层输出第一信息,所述第一信息用于确定上行反馈信息。
在一些可选实施方式中,在所述第一子模型中,所述第一输出层与所述第二输出层相邻,或者所述第一输出层与所述第二输出层之间间隔至少一个神经网络层。
在一些可选实施方式中,在所述第一子模型中,所述第一输出层位于所述第二输出层之前,或者所述第一输出层位于所述第二输出层之后。
在一些可选实施方式中,所述第一输出层位于所述第二输出层之前的情况,所述第二输出层为所述第一子模型的最后一个神经网络层;所述第一输出层为所述第一子模型的中间一个神经网络层。
在一些可选实施方式中,所述第一输出层位于所述第二输出层之后的情况,所述第一输出层为所述第一子模型的最后一个神经网络层;所述第二输出层为所述第一子模型的中间一个神经网络层。
在一些可选实施方式中,所述第一输出层位于所述第二输出层之后的情况,所述第一输出层为所述第一子模型的最后一个神经网络层;所述第二输出层为所述第一子模型的第一个神经网络层。
在一些可选实施方式中,所述第一个神经网络层的输入为所述多个波束或波束对的链路质量。
在一些可选实施方式中,所述第一输出层和所述第二输出层之间具有关联,或者所述第一输出层和所述第二输出层之间不具有关联。
在一些可选实施方式中,所述第一子模型还包括量化层,所述处理单元1202,用于通过所述量化层对所述第一信息进行量化处理,得到所述上行反馈信息。
在一些可选实施方式中,所述上行反馈信息为第一比特序列,所述第一比特序列用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述上行反馈信息为第二比特序列,所述第二比特序列用于指示所述多个波束或波束对的链路质量,所述多个波束或波束对的链路质量用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述装置还包括:接收单元1204,用于在所述至少一个接收波束索引为一个接收波束索引的情况,采用所述一个接收波束索引所指示的接收波束进行下行传输的接收。
在一些可选实施方式中,所述装置还包括:接收单元1204,用于在所述至少一个接收波束索引为多个接收波束索引的情况,从所述多个接收波束索引中确定出第一接收波束索引,采用所述第一接收波束索引所指示的接收波束进行下行传输的接收。
在一些可选实施方式中,所述接收单元1204,用于接收网络设备发送的第一指示信息,所述第 一指示信息用于指示所述网络设备采用的第一发射波束索引;基于所述至少一个接收波束索引和所述至少一个发射波束索引的对应关系,确定所述第一发射波束索引对应的第一接收波束索引。
在一些可选实施方式中,所述第一指示信息携带在MAC CE或者DCI中。
在一些可选实施方式中,所述发送单元1203,用于向网络设备上报数据集,所述数据集用于所述网络设备训练所述第一子模型;所述接收单元1204,用于接收所述网络设备下发的训练好的所述第一子模型。
在一些可选实施方式中,所述数据集还用于所述网络设备训练第二子模型,所述第二子模型用于所述网络设备基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述第一子模型和所述第二子模型联合训练;或者,所述第一子模型和所述第二子模型独立训练。
在一些可选实施方式中,所述数据集包括:
所述第一子模型的输入数据,所述输入数据为第一波束或波束对集合中的至少部分波束或波束对的链路质量;
所述第一子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个接收波束的索引;
所述第二子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个发射波束的索引以及对应的K个链路质量,K为正整数。
在一些可选实施方式中,所述链路质量包括以下至少之一:L1-RSRP、L1-SINR、L1-RSRQ。
本领域技术人员应当理解,本申请实施例的上述波束管理装置的相关描述可以参照本申请实施例的波束管理方法的相关描述进行理解。
图13是本申请实施例提供的波束管理装置的结构组成示意图二,应用于网络设备,如图13所示,所述波束管理装置包括:
接收单元1301,用于接收UE发送的上行反馈信息;
处理单元1302,用于利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述上行反馈信息为第一比特序列,所述第一比特序列用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述上行反馈信息为第二比特序列,所述第二比特序列用于指示多个波束或波束对的链路质量,所述多个波束或波束对的链路质量用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
在一些可选实施方式中,所述装置还包括:发送单元1303,用于在所述至少一个发射波束索引为一个发射波束索引的情况,采用所述一个发射波束索引所指示的发射波束进行下行传输的发送。
在一些可选实施方式中,所述装置还包括:发送单元1303,用于在所述至少一个发射波束索引为多个发射波束索引的情况,从所述多个发射波束索引中选择出第一发射波束索引,采用所述第一发射波束索引所指示的发射波束进行下行传输的发送。
在一些可选实施方式中,所述发送单元1303,用于向所述UE发送第一指示信息,所述第一指示信息用于指示所述网络设备采用的第一发射波束索引。
在一些可选实施方式中,所述第一指示信息携带在MAC CE或者DCI中。
在一些可选实施方式中,所述接收单元1301,用于接收所述UE上报的数据集;所述处理单元1302,用于利用所述数据集训练所述第二子模型。
在一些可选实施方式中,所述数据集还用于所述网络设备训练第一子模型,所述第一子模型用于所述UE基于多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息。
在一些可选实施方式中,所述至少一个接收波束索引为所述多个波束或波束对中的链路质量最好的K个接收波束的索引,K为正整数。
在一些可选实施方式中,所述至少一个发射波束索引为所述多个波束或波束对中的链路质量最好的K个发射波束的索引,K为正整数。
在一些可选实施方式中,所述至少一个接收波束索引、所述至少一个发射波束索引以及所述至少一个链路质量中的任意两者之间具有对应关系。
在一些可选实施方式中,所述对应关系通过所述至少一个接收波束索引的排序、所述至少一个发射波束索引的排序以及所述至少一个链路质量的排序确定。
在一些可选实施方式中,所述第一子模型和所述第二子模型联合训练;或者,所述第一子模型和所述第二子模型独立训练。
在一些可选实施方式中,所述发送单元1303,用于向所述UE下发训练好的所述第一子模型。
在一些可选实施方式中,所述数据集包括:
所述第一子模型的输入数据,所述输入数据为第一波束或波束对集合中的至少部分波束或波束对的链路质量;
所述第一子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个接收波束的索引;
所述第二子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个发射波束的索引以及对应的K个链路质量,K为正整数。
在一些可选实施方式中,所述链路质量包括以下至少之一:L1-RSRP、L1-SINR、L1-RSRQ。
本领域技术人员应当理解,本申请实施例的上述波束管理装置的相关描述可以参照本申请实施例的波束管理方法的相关描述进行理解。
图14是本申请实施例提供的一种通信设备1400示意性结构图。该通信设备可以用户设备,也可以是网络设备。图14所示的通信设备1400包括处理器1410,处理器1410可以从存储器中调用并运行计算机程序,以实现本申请实施例中的方法。
可选地,如图14所示,通信设备1400还可以包括存储器1420。其中,处理器1410可以从存储器1420中调用并运行计算机程序,以实现本申请实施例中的方法。
其中,存储器1420可以是独立于处理器1410的一个单独的器件,也可以集成在处理器1410中。
可选地,如图14所示,通信设备1400还可以包括收发器1430,处理器1410可以控制该收发器1430与其他设备进行通信,具体地,可以向其他设备发送信息或数据,或接收其他设备发送的信息或数据。
其中,收发器1430可以包括发射机和接收机。收发器1430还可以进一步包括天线,天线的数量可以为一个或多个。
可选地,该通信设备1400具体可为本申请实施例的网络设备,并且该通信设备1400可以实现本申请实施例的各个方法中由网络设备实现的相应流程,为了简洁,在此不再赘述。
可选地,该通信设备1400具体可为本申请实施例的用户设备,并且该通信设备1400可以实现本申请实施例的各个方法中由用户设备实现的相应流程,为了简洁,在此不再赘述。
图15是本申请实施例的芯片的示意性结构图。图15所示的芯片1500包括处理器1510,处理器1510可以从存储器中调用并运行计算机程序,以实现本申请实施例中的方法。
可选地,如图15所示,芯片1500还可以包括存储器1520。其中,处理器1510可以从存储器1520中调用并运行计算机程序,以实现本申请实施例中的方法。
其中,存储器1520可以是独立于处理器1510的一个单独的器件,也可以集成在处理器1510中。
可选地,该芯片1500还可以包括输入接口1530。其中,处理器1510可以控制该输入接口1530与其他设备或芯片进行通信,具体地,可以获取其他设备或芯片发送的信息或数据。
可选地,该芯片1500还可以包括输出接口1540。其中,处理器1510可以控制该输出接口1540与其他设备或芯片进行通信,具体地,可以向其他设备或芯片输出信息或数据。
可选地,该芯片可应用于本申请实施例中的网络设备,并且该芯片可以实现本申请实施例的各个方法中由网络设备实现的相应流程,为了简洁,在此不再赘述。
可选地,该芯片可应用于本申请实施例中的用户设备,并且该芯片可以实现本申请实施例的各个方法中由用户设备实现的相应流程,为了简洁,在此不再赘述。
应理解,本申请实施例提到的芯片还可以称为系统级芯片,系统芯片,芯片系统或片上系统芯片等。
图16是本申请实施例提供的一种通信系统1600的示意性框图。如图16所示,该通信系统1600包括用户设备1610和网络设备1620。
其中,该用户设备1610可以用于实现上述方法中由用户设备实现的相应的功能,以及该网络设备1620可以用于实现上述方法中由网络设备实现的相应的功能为了简洁,在此不再赘述。
应理解,本申请实施例的处理器可能是一种集成电路芯片,具有信号的处理能力。在实现过程中,上述方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现成可编程门阵列(Field Programmable Gate Array, FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器,闪存、只读存储器,可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器,处理器读取存储器中的信息,结合其硬件完成上述方法的步骤。
可以理解,本申请实施例中的存储器可以是易失性存储器或非易失性存储器,或可包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(Read-Only Memory,ROM)、可编程只读存储器(Programmable ROM,PROM)、可擦除可编程只读存储器(Erasable PROM,EPROM)、电可擦除可编程只读存储器(Electrically EPROM,EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory,RAM),其用作外部高速缓存。通过示例性但不是限制性说明,许多形式的RAM可用,例如静态随机存取存储器(Static RAM,SRAM)、动态随机存取存储器(Dynamic RAM,DRAM)、同步动态随机存取存储器(Synchronous DRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM,DDR SDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(Synchlink DRAM,SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM,DR RAM)。应注意,本文描述的系统和方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
应理解,上述存储器为示例性但不是限制性说明,例如,本申请实施例中的存储器还可以是静态随机存取存储器(static RAM,SRAM)、动态随机存取存储器(dynamic RAM,DRAM)、同步动态随机存取存储器(synchronous DRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM,DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(synch link DRAM,SLDRAM)以及直接内存总线随机存取存储器(Direct Rambus RAM,DR RAM)等等。也就是说,本申请实施例中的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
本申请实施例还提供了一种计算机可读存储介质,用于存储计算机程序。
可选的,该计算机可读存储介质可应用于本申请实施例中的网络设备,并且该计算机程序使得计算机执行本申请实施例的各个方法中由网络设备实现的相应流程,为了简洁,在此不再赘述。
可选地,该计算机可读存储介质可应用于本申请实施例中的用户设备,并且该计算机程序使得计算机执行本申请实施例的各个方法中由用户设备实现的相应流程,为了简洁,在此不再赘述。
本申请实施例还提供了一种计算机程序产品,包括计算机程序指令。
可选的,该计算机程序产品可应用于本申请实施例中的网络设备,并且该计算机程序指令使得计算机执行本申请实施例的各个方法中由网络设备实现的相应流程,为了简洁,在此不再赘述。
可选地,该计算机程序产品可应用于本申请实施例中的用户设备,并且该计算机程序指令使得计算机执行本申请实施例的各个方法中由用户设备实现的相应流程,为了简洁,在此不再赘述。
本申请实施例还提供了一种计算机程序。
可选的,该计算机程序可应用于本申请实施例中的网络设备,当该计算机程序在计算机上运行时,使得计算机执行本申请实施例的各个方法中由网络设备实现的相应流程,为了简洁,在此不再赘述。
可选地,该计算机程序可应用于本申请实施例中的用户设备,当该计算机程序在计算机上运行时,使得计算机执行本申请实施例的各个方法中由用户设备实现的相应流程,为了简洁,在此不再赘述。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,上述描述的系统、装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在本申请所提供的几个实施例中,应该理解到,所揭露的系统、装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦 合或通信连接可以是通过一些接口,装置或单元的间接耦合或通信连接,可以是电性,机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(Read-Only Memory,)ROM、随机存取存储器(Random Access Memory,RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述,仅为本申请的具体实施方式,但本申请的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本申请的保护范围之内。因此,本申请的保护范围应所述以权利要求的保护范围为准。

Claims (57)

  1. 一种波束管理方法,所述方法包括:
    用户设备UE测量多个波束或波束对的链路质量;
    所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
  2. 根据权利要求1所述的方法,其中,所述UE确定所述上行反馈信息的情况下,所述方法还包括:
    所述UE向网络设备发送所述上行反馈信息,所述上行反馈信息用于所述网络设备利用第二子模型确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
  3. 根据权利要求1或2所述的方法,其中,所述多个波束为多个接收波束或者多个发射波束,所述多个波束对为多个发射波束和接收波束对。
  4. 根据权利要求3所述的方法,其中,所述至少一个接收波束索引为所述多个波束或波束对中的链路质量最好的K个接收波束的索引,K为正整数。
  5. 根据权利要求3所述的方法,其中,所述至少一个发射波束索引为所述多个波束或波束对中的链路质量最好的K个发射波束的索引,K为正整数。
  6. 根据权利要求1至5中任一项所述的方法,其中,所述至少一个接收波束索引、所述至少一个发射波束索引以及所述至少一个链路质量中的任意两者之间具有对应关系。
  7. 根据权利要求6所述的方法,其中,所述对应关系通过所述至少一个接收波束索引的排序、所述至少一个发射波束索引的排序以及所述至少一个链路质量的排序确定。
  8. 根据权利要求1至7中任一项所述的方法,其中,所述第一子模型包括输出层;
    所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和上行反馈信息,包括:
    所述UE利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述输出层的第一部分节点输出至少一个接收波束索引,以及通过所述输出层的第二部分节点输出第一信息,所述第一信息用于确定上行反馈信息。
  9. 根据权利要求1至7中任一项所述的方法,其中,所述第一子模型包括第一输出层和第二输出层;
    所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和上行反馈信息,包括:
    所述UE利用第一子模型对所述多个波束或波束对的链路质量进行处理,通过所述第一输出层输出至少一个接收波束索引,以及通过所述第二输出层输出第一信息,所述第一信息用于确定上行反馈信息。
  10. 根据权利要求9所述的方法,其中,在所述第一子模型中,所述第一输出层与所述第二输出层相邻,或者所述第一输出层与所述第二输出层之间间隔至少一个神经网络层。
  11. 根据权利要求9或10所述的方法,其中,在所述第一子模型中,所述第一输出层位于所述第二输出层之前,或者所述第一输出层位于所述第二输出层之后。
  12. 根据权利要求9至11中任一项所述的方法,其中,所述第一输出层位于所述第二输出层之前的情况,
    所述第二输出层为所述第一子模型的最后一个神经网络层;
    所述第一输出层为所述第一子模型的中间一个神经网络层。
  13. 根据权利要求9至11中任一项所述的方法,其中,所述第一输出层位于所述第二输出层之后的情况,
    所述第一输出层为所述第一子模型的最后一个神经网络层;
    所述第二输出层为所述第一子模型的中间一个神经网络层。
  14. 根据权利要求9至11中任一项所述的方法,其中,所述第一输出层位于所述第二输出层之后的情况,
    所述第一输出层为所述第一子模型的最后一个神经网络层;
    所述第二输出层为所述第一子模型的第一个神经网络层。
  15. 根据权利要求14所述的方法,其中,所述第一个神经网络层的输入为所述多个波束或波束对的链路质量。
  16. 根据权利要求9至15中任一项所述的方法,其中,所述第一输出层和所述第二输出层之间具有关联,或者所述第一输出层和所述第二输出层之间不具有关联。
  17. 根据权利要求8至16中任一项所述的方法,其中,所述第一子模型还包括量化层,所述方法还包括:
    所述UE通过所述量化层对所述第一信息进行量化处理,得到所述上行反馈信息。
  18. 根据权利要求17所述的方法,其中,所述上行反馈信息为第一比特序列,所述第一比特序列用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
  19. 根据权利要求17所述的方法,其中,所述上行反馈信息为第二比特序列,所述第二比特序列用于指示所述多个波束或波束对的链路质量,所述多个波束或波束对的链路质量用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
  20. 根据权利要求1至19中任一项所述的方法,其中,所述方法还包括:
    所述至少一个接收波束索引为一个接收波束索引的情况,所述UE采用所述一个接收波束索引所指示的接收波束进行下行传输的接收。
  21. 根据权利要求1至19中任一项所述的方法,其中,所述方法还包括:
    所述至少一个接收波束索引为多个接收波束索引的情况,所述UE从所述多个接收波束索引中确定出第一接收波束索引,采用所述第一接收波束索引所指示的接收波束进行下行传输的接收。
  22. 根据权利要求21所述的方法,其中,所述UE从所述多个接收波束索引中确定出第一接收波束索引,包括:
    所述UE接收网络设备发送的第一指示信息,所述第一指示信息用于指示所述网络设备采用的第一发射波束索引;
    所述UE基于所述至少一个接收波束索引和所述至少一个发射波束索引的对应关系,确定所述第一发射波束索引对应的第一接收波束索引。
  23. 根据权利要求22所述的方法,其中,所述第一指示信息携带在媒体介入控制控制元素MAC CE或者下行控制信息DCI中。
  24. 根据权利要求1至23中任一项所述的方法,其中,所述UE利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息之前,所述方法还包括:
    所述UE向网络设备上报数据集,所述数据集用于所述网络设备训练所述第一子模型;
    所述UE接收所述网络设备下发的训练好的所述第一子模型。
  25. 根据权利要求24所述的方法,其中,所述数据集还用于所述网络设备训练第二子模型,所述第二子模型用于所述网络设备基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
  26. 根据权利要求25所述的方法,其中,
    所述第一子模型和所述第二子模型联合训练;或者,
    所述第一子模型和所述第二子模型独立训练。
  27. 根据权利要求24至26中任一项所述的方法,其中,所述数据集包括:
    所述第一子模型的输入数据,所述输入数据为第一波束或波束对集合中的至少部分波束或波束对的链路质量;
    所述第一子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个接收波束的索引;
    所述第二子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个发射波束的索引以及对应的K个链路质量,K为正整数。
  28. 根据权利要求1至27中任一项所述的方法,其中,所述链路质量包括以下至少之一:层1-参考信号接收功率L1-RSRP、层1-信号与干扰加噪声比L1-SINR、层1-参考信号接收质量L1-RSRQ。
  29. 一种波束管理方法,所述方法包括:
    网络设备接收UE发送的上行反馈信息;
    所述网络设备利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述 至少一个发射波束索引对应的至少一个链路质量。
  30. 根据权利要求29所述的方法,其中,所述上行反馈信息为第一比特序列,所述第一比特序列用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
  31. 根据权利要求29所述的方法,其中,所述上行反馈信息为第二比特序列,所述第二比特序列用于指示多个波束或波束对的链路质量,所述多个波束或波束对的链路质量用于确定至少一个发射波束索引以及所述至少一个发射波束索引对应的至少一个链路质量。
  32. 根据权利要求29至31中任一项所述的方法,其中,所述方法还包括:
    所述至少一个发射波束索引为一个发射波束索引的情况,所述网络设备采用所述一个发射波束索引所指示的发射波束进行下行传输的发送。
  33. 根据权利要求29至31中任一项所述的方法,其中,所述方法还包括:
    所述至少一个发射波束索引为多个发射波束索引的情况,所述网络设备从所述多个发射波束索引中选择出第一发射波束索引,采用所述第一发射波束索引所指示的发射波束进行下行传输的发送。
  34. 根据权利要求33所述的方法,其中,所述方法还包括:
    所述网络设备向所述UE发送第一指示信息,所述第一指示信息用于指示所述网络设备采用的第一发射波束索引。
  35. 根据权利要求34所述的方法,其中,所述第一指示信息携带在MAC CE或者DCI中。
  36. 根据权利要求34所述的方法,其中,所述网络设备接收UE发送的上行反馈信息之前,所述方法还包括:
    所述网络设备接收所述UE上报的数据集;
    所述网络设备利用所述数据集训练所述第二子模型。
  37. 根据权利要求36所述的方法,其中,所述数据集还用于所述网络设备训练第一子模型,所述第一子模型用于所述UE基于多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息。
  38. 根据权利要求37所述的方法,其中,所述至少一个接收波束索引为所述多个波束或波束对中的链路质量最好的K个接收波束的索引,K为正整数。
  39. 根据权利要求37所述的方法,其中,所述至少一个发射波束索引为所述多个波束或波束对中的链路质量最好的K个发射波束的索引,K为正整数。
  40. 根据权利要求37至39中任一项所述的方法,其中,所述至少一个接收波束索引、所述至少一个发射波束索引以及所述至少一个链路质量中的任意两者之间具有对应关系。
  41. 根据权利要求40所述的方法,其中,所述对应关系通过所述至少一个接收波束索引的排序、所述至少一个发射波束索引的排序以及所述至少一个链路质量的排序确定。
  42. 根据权利要求37至41中任一项所述的方法,其中,
    所述第一子模型和所述第二子模型联合训练;或者,
    所述第一子模型和所述第二子模型独立训练。
  43. 根据权利要求37至42中任一项所述的方法,其中,所述方法还包括:
    所述网络设备向所述UE下发训练好的所述第一子模型。
  44. 根据权利要求36至43中任一项所述的方法,其中,所述数据集包括:
    所述第一子模型的输入数据,所述输入数据为第一波束或波束对集合中的至少部分波束或波束对的链路质量;
    所述第一子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个接收波束的索引;
    所述第二子模型的输出数据的标签,所述标签为所述第一波束或波束对集合中的链路质量最好的K个发射波束的索引以及对应的K个链路质量,K为正整数。
  45. 根据权利要求29至44中任一项所述的方法,其中,所述链路质量包括以下至少之一:L1-RSRP、L1-SINR、L1-RSRQ。
  46. 一种波束管理装置,应用于UE,所述装置包括:
    测量单元,用于测量多个波束或波束对的链路质量;
    处理单元,用于利用第一子模型基于所述多个波束或波束对的链路质量确定至少一个接收波束索引和/或上行反馈信息;其中,所述上行反馈信息用于确定至少一个发射波束索引和/或所述 至少一个发射波束索引对应的至少一个链路质量。
  47. 一种波束管理装置,应用于网络设备,所述装置包括:
    接收单元,用于接收UE发送的上行反馈信息;
    处理单元,用于利用第二子模型基于所述上行反馈信息确定至少一个发射波束索引和/或所述至少一个发射波束索引对应的至少一个链路质量。
  48. 一种用户设备,包括:处理器和存储器,该存储器用于存储计算机程序,所述处理器用于调用并运行所述存储器中存储的计算机程序,执行如权利要求1至28中任一项所述的方法。
  49. 一种网络设备,包括:处理器和存储器,该存储器用于存储计算机程序,所述处理器用于调用并运行所述存储器中存储的计算机程序,执行如权利要求29至45中任一项所述的方法。
  50. 一种芯片,包括:处理器,用于从存储器中调用并运行计算机程序,使得安装有所述芯片的设备执行如权利要求1至28中任一项所述的方法。
  51. 一种芯片,包括:处理器,用于从存储器中调用并运行计算机程序,使得安装有所述芯片的设备执行如权利要求29至45中任一项所述的方法。
  52. 一种计算机可读存储介质,用于存储计算机程序,所述计算机程序使得计算机执行如权利要求1至28中任一项所述的方法。
  53. 一种计算机可读存储介质,用于存储计算机程序,所述计算机程序使得计算机执行如权利要求29至45中任一项所述的方法。
  54. 一种计算机程序产品,包括计算机程序指令,该计算机程序指令使得计算机执行如权利要求1至28中任一项所述的方法。
  55. 一种计算机程序产品,包括计算机程序指令,该计算机程序指令使得计算机执行如权利要求29至45中任一项所述的方法。
  56. 一种计算机程序,所述计算机程序使得计算机执行如权利要求1至28中任一项所述的方法。
  57. 一种计算机程序,所述计算机程序使得计算机执行如权利要求29至45中任一项所述的方法。
PCT/CN2022/096170 2022-05-31 2022-05-31 一种波束管理方法及装置、用户设备、网络设备 WO2023230818A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/096170 WO2023230818A1 (zh) 2022-05-31 2022-05-31 一种波束管理方法及装置、用户设备、网络设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/096170 WO2023230818A1 (zh) 2022-05-31 2022-05-31 一种波束管理方法及装置、用户设备、网络设备

Publications (1)

Publication Number Publication Date
WO2023230818A1 true WO2023230818A1 (zh) 2023-12-07

Family

ID=89026602

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/096170 WO2023230818A1 (zh) 2022-05-31 2022-05-31 一种波束管理方法及装置、用户设备、网络设备

Country Status (1)

Country Link
WO (1) WO2023230818A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109983820A (zh) * 2017-07-25 2019-07-05 联发科技股份有限公司 波束成形无线通信系统中具有波束指示的波束管理方法
CN110677180A (zh) * 2013-09-27 2020-01-10 三星电子株式会社 无线通信系统中发送和接收波束信息的装置及方法
CN111543012A (zh) * 2017-12-15 2020-08-14 高通股份有限公司 用于动态波束对确定的方法和设备
CN111713038A (zh) * 2018-02-15 2020-09-25 高通股份有限公司 用于辅助波束细化的技术
WO2022104700A1 (en) * 2020-11-20 2022-05-27 Huawei Technologies Co., Ltd. Beam set for channel state information feedback in mimo systems

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110677180A (zh) * 2013-09-27 2020-01-10 三星电子株式会社 无线通信系统中发送和接收波束信息的装置及方法
CN109983820A (zh) * 2017-07-25 2019-07-05 联发科技股份有限公司 波束成形无线通信系统中具有波束指示的波束管理方法
CN111543012A (zh) * 2017-12-15 2020-08-14 高通股份有限公司 用于动态波束对确定的方法和设备
CN111713038A (zh) * 2018-02-15 2020-09-25 高通股份有限公司 用于辅助波束细化的技术
WO2022104700A1 (en) * 2020-11-20 2022-05-27 Huawei Technologies Co., Ltd. Beam set for channel state information feedback in mimo systems

Similar Documents

Publication Publication Date Title
EP3583797B1 (en) Methods and systems for network self-optimization using deep learning
CN111967605A (zh) 无线电接入网中的机器学习
CN103716123A (zh) 一种传感器网络中分组丢失和错误的原因识别方法和系统
US20240211770A1 (en) Communication method and apparatus
EP4156631A1 (en) Reinforcement learning (rl) and graph neural network (gnn)-based resource management for wireless access networks
US12047248B2 (en) Method of controlling state control parameter for adjusting state of network of base station by using any one of plurality of models and electronic device performing the method
US12010571B2 (en) Spectral efficiency prediction with artificial intelligence for enhancing carrier aggregation and proactive radio resource management
Yan et al. Deep reinforcement learning based offloading for mobile edge computing with general task graph
Wu et al. AoI minimization for UAV-to-device underlay communication by multi-agent deep reinforcement learning
Zhao et al. Hierarchical multi-objective deep reinforcement learning for packet duplication in multi-connectivity for URLLC
CN116848828A (zh) 机器学习模型分布
CN114492841A (zh) 一种模型梯度更新方法及装置
Hajizadeh et al. Decentralized configuration of TSCH-based IoT networks for distinctive QoS: A deep reinforcement learning approach
WO2023230818A1 (zh) 一种波束管理方法及装置、用户设备、网络设备
Li et al. Learning to Transmit with Provable Guarantees in Wireless Federated Learning
Sahu et al. Communication efficient distributed estimation over directed random graphs
WO2024031535A1 (zh) 无线通信方法、终端设备和网络设备
WO2024159986A1 (zh) 一种无线局域网络动态阈值参数生成方法及装置
WO2022237865A1 (zh) 一种数据处理方法及装置
WO2024032695A1 (zh) Csi预测处理方法、装置、通信设备及可读存储介质
US20240314613A1 (en) Communication method and apparatus
WO2023231706A1 (zh) 数据信道模型的发送方法及装置、信息发送方法及装置
WO2023093725A1 (zh) 一种校准方法及装置
US20240346329A1 (en) Method and apparatus for training intelligent model
WO2024094166A1 (zh) 报告传输方法及通信装置

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: 22944158

Country of ref document: EP

Kind code of ref document: A1