[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2024217495A1 - 信息处理方法、信息处理装置、终端及网络侧设备 - Google Patents

信息处理方法、信息处理装置、终端及网络侧设备 Download PDF

Info

Publication number
WO2024217495A1
WO2024217495A1 PCT/CN2024/088579 CN2024088579W WO2024217495A1 WO 2024217495 A1 WO2024217495 A1 WO 2024217495A1 CN 2024088579 W CN2024088579 W CN 2024088579W WO 2024217495 A1 WO2024217495 A1 WO 2024217495A1
Authority
WO
WIPO (PCT)
Prior art keywords
unit
information
target
terminal
data set
Prior art date
Application number
PCT/CN2024/088579
Other languages
English (en)
French (fr)
Inventor
任千尧
谢天
Original Assignee
维沃软件技术有限公司
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 维沃软件技术有限公司 filed Critical 维沃软件技术有限公司
Publication of WO2024217495A1 publication Critical patent/WO2024217495A1/zh

Links

Definitions

  • the present application belongs to the field of communication technology, and specifically relates to an information processing method, an information processing device, a terminal and a network side device.
  • the terminal can use an encoding artificial intelligence (AI) model to compress channel information and report the compressed channel characteristic information.
  • AI artificial intelligence
  • the network side can use a decoding AI model to restore the channel characteristic information reported by the terminal to obtain the terminal channel information. In this way, the overhead of the terminal reporting channel information can be reduced.
  • the performance of the AI model will fluctuate as the terminal position, channel environment, etc. change, and may not match the current channel state, resulting in inaccurate compressed and recovered channel information.
  • the AI model is used to compress and recover the channel information, resulting in inaccurate channel information obtained by the network side, that is, the reliability of channel information reporting is low.
  • the embodiments of the present application provide an information processing method, an information processing device, a terminal, and a network-side device, which can train a target AI unit related to a decoding AI model on the terminal side during the process of reporting channel information that compresses and restores channel information based on an AI model, and use the target AI unit to supervise the performance of the channel information reported based on the AI model, so as to promptly discover the problem of inaccurate compressed and restored channel information.
  • an information processing method comprising:
  • the terminal obtains a first data set
  • the terminal trains a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal in reporting target channel information to a network side device, the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit of the network side device, the first channel characteristic information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • an information processing method comprising:
  • the first device obtains downlink channel state information
  • the first device trains a first AI unit and a second AI unit based on the downlink channel state information
  • the first device sends second information to the terminal, where the second information includes the first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set;
  • the first data set is used to train a target AI unit, and the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, wherein the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit, and the first channel characteristic information is information obtained based on the first processing of the first channel information by the first AI unit of the terminal.
  • an information processing device comprising:
  • a first acquisition module used to acquire a first data set
  • a first training module is used to train a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit of the network side device, the first channel characteristic information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • an information processing device comprising:
  • a third acquisition module is used to acquire downlink channel state information
  • a third training module configured to train the first AI unit and the second AI unit based on the downlink channel state information
  • a first sending module configured to send second information to a terminal, where the second information includes a first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set;
  • the first data set is used to train a target AI unit, and the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, wherein the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit, and the first channel characteristic information is information obtained based on the first processing of the first channel information by the first AI unit of the terminal.
  • a terminal comprising a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.
  • a terminal including a processor and a communication interface, wherein the communication interface is used to obtain a first data set; the processor is used to train a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal reporting target channel information to a network side device, and the target channel information is based on the
  • the target AI unit is obtained by first processing of the first channel information by the first AI unit of the terminal and second processing of the first channel characteristic information by the second AI unit of the network side device, wherein the first channel characteristic information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • a network side device which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the first aspect are implemented.
  • a network side device including a processor and a communication interface, the communication interface being used to obtain downlink channel state information; the processor being used to train a first AI unit and a second AI unit based on the downlink channel state information; the communication interface being further used to send second information to a terminal, the second information including a first data set, or the second information including an identifier of the first data set, and the terminal being aware of an association relationship between the first data set and the identifier of the first data set;
  • the first data set is used to train a target AI unit, and the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, wherein the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit, and the first channel characteristic information is information obtained based on the first processing of the first channel information by the first AI unit of the terminal.
  • a readable storage medium on which a program or instruction is stored.
  • the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.
  • a wireless communication system comprising: a terminal and a first device, wherein the terminal can be used to execute the steps of the method described in the first aspect, and the first device can be used to execute the steps of the method described in the second aspect.
  • a chip comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method described in the first aspect, or to implement the method described in the second aspect.
  • a computer program/program product is provided, wherein the computer program/program product is stored in a storage medium, and the program/program product is executed by at least one processor to implement the steps of the information processing method as described in the first aspect, or to implement the steps of the information processing method as described in the second aspect.
  • a target AI unit in an AI-based channel information reporting scenario, is trained on the terminal side, and the target AI unit is related to a second AI unit used on the network side.
  • the AI unit can be used to detect the performance of the terminal in reporting the target channel information to the network side device, so as to promptly discover the problem of inaccurate AI-based compression-recovery channel information.
  • FIG1 is a schematic diagram of the structure of a wireless communication system to which an embodiment of the present application can be applied;
  • FIG2 is a schematic diagram of the architecture of a neural network model
  • Figure 3 is a schematic diagram of a neuron
  • FIG. 4 is a flow chart of an information processing method provided in an embodiment of the present application.
  • FIG5 is a flow chart of another information processing method provided in an embodiment of the present application.
  • FIG6 is a schematic diagram of the structure of an information processing device provided in an embodiment of the present application.
  • FIG7 is a schematic diagram of the structure of another information processing device provided in an embodiment of the present application.
  • FIG8 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application.
  • FIG9 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application.
  • FIG10 is a schematic diagram of the structure of a network side device provided in an embodiment of the present application.
  • FIG. 11 is a schematic diagram of the structure of another network-side device provided in an embodiment of the present application.
  • first, second, etc. of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by “first” and “second” are generally of one type, and the number of objects is not limited, for example, the first object can be one or more.
  • “or” in the present application represents at least one of the connected objects.
  • “A or B” covers three schemes, namely, Scheme 1: including A but not including B; Scheme 2: including B but not including A; Scheme 3: including both A and B.
  • the character "/" generally indicates that the objects associated with each other are in an "or” relationship.
  • indication in this application can be a direct indication (or explicit indication) or an indirect indication (or implicit indication).
  • a direct indication can be understood as the sender explicitly informing the receiver of specific information, operations to be performed, or request results in the sent indication;
  • an indirect indication can be understood as the receiver determining the corresponding information according to the indication sent by the sender, or making a judgment and determining the operation to be performed or the request result according to the judgment result.
  • LTE Long Term Evolution
  • LTE-A Long Term Evolution-Advanced
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency-Division Multiple Access
  • NR New Radio
  • FIG1 shows a block diagram of a wireless communication system applicable to the embodiment of the present application.
  • the wireless communication system includes a terminal 11 and a network side device 12 .
  • the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer), a notebook computer, a personal digital assistant (PDA), a handheld computer, a netbook, an ultra-mobile personal computer (Ultra-mobile Personal Computer, UMPC), a mobile Internet device (Mobile Internet Device, MID), an augmented reality (Augmented Reality, AR), a virtual reality (Virtual Reality, VR) device, a robot, a wearable device (Wearable Device), a flight vehicle (flight vehicle), a vehicle user equipment (VUE), a shipborne equipment, a pedestrian terminal (Pedestrian User Equipment, PUE), a smart home (home appliances with wireless communication functions, such as refrigerators, televisions, washing machines or furniture, etc.), a game console, a personal computer (Personal Computer, PC
  • Wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc.
  • the vehicle-mounted device can also be called a vehicle-mounted terminal, a vehicle-mounted controller, a vehicle-mounted module, a vehicle-mounted component, a vehicle-mounted chip or a vehicle-mounted unit, etc. It should be noted that the specific type of the terminal 11 is not limited in the embodiment of the present application.
  • the network side device 12 may include an access network device or a core network device, wherein the access network device may also be called a radio access network (Radio Access Network, RAN) device, a radio access network function or a radio access network unit.
  • the access network device may include a base station, a wireless local area network (Wireless Local Area Network, WLAN) access point (Access Point, AS) or a wireless fidelity (Wireless Fidelity, WiFi) node, etc.
  • WLAN wireless Local Area Network
  • AS Access Point
  • WiFi wireless Fidelity
  • the base station can be called Node B (Node B, NB), Evolved Node B (Evolved Node B, eNB), the next generation Node B (the next generation Node B, gNB), New Radio Node B (New Radio Node B, NR Node B), access point, Relay Base Station (Relay Base Station, RBS), Serving Base Station (Serving Base Station, SBS), Base Transceiver Station (Base Transceiver Station, BTS), radio base station, radio transceiver, base
  • the base station is not limited to specific technical terms as long as the same technical effect is achieved. It should be noted that in the embodiments of the present application, only the base station in the NR system is taken as an example for introduction, and the specific type of the base station is not limited.
  • the core network device may include the core network device may include but is not limited to at least one of the following: core network node, core network function, location management function (Location Management Function, LMF), mobility management entity (Mobility Management Entity, MME), access mobility management function (Access and Mobility Management Function, AMF), session management function (Session Management Function, SMF), user plane function (User Plane Function, UPF), policy control function (Policy Control Function, PCF), policy and charging rules function unit (Policy and Charging Rules Function, PCRF), edge application service discovery function (Edge Application Server Discovery Function, EASDF), unified data management (Unified Data Management, UDM), unified data storage (Unified Data Repository, UDR), home user server (Home Subscriber Server, HSS), centralized network configuration (CNC), network repository function (NRF), network exposure function (NEF), local NEF (Local NEF, or L-NEF), binding support function (BSF), application function (AF), etc.
  • LMF Location Management Function
  • MME
  • the transmitter can optimize the signal transmission based on CSI to make it more compatible with the channel state.
  • the Channel Quality Indicator CQI
  • MCS modulation and coding scheme
  • PMI Precoding Matrix Indicator
  • MIMO Multi-Input Multi-Output
  • the network-side device sends CSI-Reference Signals (CSI-RS) on certain time-frequency resources in a certain time slot.
  • CSI-RS CSI-Reference Signals
  • the terminal performs channel estimation based on the CSI-RS, calculates the channel information on this slot, and feeds back the PMI to the base station through the codebook.
  • the network-side device combines the channel information based on the codebook information fed back by the terminal, and uses this channel information for data precoding and multi-user scheduling before the terminal reports the CSI next time.
  • the terminal can change the PMI reported in each subband to reporting PMI in the delay domain (delay domain, i.e., frequency domain). Since the channels in the delay domain are more concentrated, the PMIs with fewer delays can approximately represent the PMIs of all subbands, which can be regarded as compressing the delay domain information before reporting.
  • delay domain i.e., frequency domain
  • the network side device can pre-code the CSI-RS in advance and send the encoded CSI-RS to the terminal.
  • the terminal sees the channel corresponding to the encoded CSI-RS.
  • the terminal only needs to select several ports with higher strength from the ports indicated by the network side device and report the coefficients corresponding to these ports.
  • the compression effect of channel characteristic information can be improved by compressing channel information using an AI unit.
  • the terminal can estimate the CSI reference signal (CSI Reference Signal, CSI-RS) or the tracking reference signal (Tracking Reference Signal, TRS), calculate according to the estimated channel information, obtain the calculated channel information, and then encode the calculated channel information or the original estimated channel information through an encoder to obtain the encoding result, and finally send the encoding result to the base station.
  • the base station can input the encoded result into the decoder after receiving it, and use the decoder to restore the channel information.
  • the CSI compression feedback scheme based on the neural network is that the terminal uses the encoding network to compress and encode the channel information, sends the compressed content to the base station, and decodes the compressed content using the decoding network at the base station to restore the channel information.
  • the decoding network of the base station and the encoding network of the terminal need to be jointly trained to achieve a reasonable matching degree.
  • the input of the encoding network is the channel information
  • the output is the encoding information, that is, the channel characteristic information.
  • the input of the decoding network is the encoding information, and the output is the restored channel information.
  • a target AI unit is trained on the terminal side, and the target AI unit is related to a second AI unit (such as a decoder) used on the network side.
  • the target AI unit can be used to detect the performance of the terminal in reporting the target channel information to the network side device, such as: using the target AI unit to decode or restore the first channel feature information output by the first AI unit, and comparing the decoded or restored channel information with the first channel information.
  • the main evaluation indicator of the AI unit is the correlation between the channel information input to the encoder and the channel information recovered by the decoder. If the two are exactly the same, it means that the AI unit has achieved perfect compression and decompression. Usually, if the correlation loss is within a certain degree, the AI unit can be considered effective. In related technologies, in order to detect the AI unit, it is necessary to simultaneously obtain the first AI unit of the encoder and the second AI unit of the decoder on a single side (network side or terminal side).
  • the second AI unit can be used to restore the channel characteristic information compressed by the first AI unit, and the restored channel information is compared with the channel information input to the first AI unit, and the performance of the first AI unit and the second AI unit is judged based on the correlation between the two.
  • the network side will pass the second AI unit used to the terminal, so that the terminal uses the second AI unit to detect the performance of the terminal reporting the target channel information to the network side device.
  • this method is subject to many limitations, such as: due to model privacy, compatibility and other aspects, some model training collaboration methods (type2 or type3) do not support the problem of obtaining a complete model at one end, and the terminal cannot obtain the first AI unit and the second AI unit at the same time, making the solution impossible to implement.
  • the AI unit in the embodiments of the present application may also be referred to as an AI model, an AI structure, etc., or the AI unit may also refer to a processing unit capable of implementing specific algorithms, formulas, processing procedures, capabilities, etc. related to AI, or the AI unit may be a processing method, algorithm, function, module or unit for a specific data set, or the AI unit may be a processing method, algorithm, function, module or unit running on AI-related hardware such as a graphics processor (GPU), a network processor (NPU), a tensor processor (TPU), an application specific integrated circuit (ASIC), etc., and the present application does not make specific limitations on this.
  • the specific data set includes the input and/or output of the AI unit.
  • the identifier of the AI unit may be an AI model identifier, an AI structure identifier, an AI algorithm identifier, or an identifier of a specific data set associated with the AI unit, or an identifier of a specific scene, environment, or channel feature associated with the AI.
  • the present application does not specifically limit the identification of the AI-related function, feature, capability or module.
  • the AI unit is taken as a neural network in the embodiment of this application for illustration, but the specific type of the AI unit is not limited.
  • the neural network model includes an input layer, a hidden layer and an output layer, which can predict possible output results (Y) based on the input and output information ( X1 ⁇ Xn ) obtained by the input layer.
  • the neural network model is composed of a large number of neurons.
  • k the total number of input parameters.
  • the parameters of the neural network are optimized through optimization algorithms.
  • An optimization algorithm is a type of algorithm that can help us minimize or maximize an objective function (sometimes called a loss function).
  • the objective function is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we build a neural network model f(.). With the neural network model, we can get the predicted output f(x) based on the input x, and we can calculate the difference between the predicted value and the true value (f(x)-Y), which is the loss function. Our goal is to find the right W and b to minimize the value of the above loss function. The smaller the loss value, the closer our model is to the actual situation.
  • the common optimization algorithms are basically based on the error back propagation (BP) algorithm.
  • BP error back propagation
  • the basic idea of the BP algorithm is that the learning process consists of two processes: the forward propagation of the signal and the back propagation of the error.
  • the input sample is transmitted from the input layer, processed by each hidden layer layer by layer, and then transmitted to the output layer. If the actual output of the output layer does not match the expected output, it will enter the error back propagation stage.
  • Error back propagation is to propagate the output error layer by layer through the hidden layer to the input layer in some form, and distribute the error to all units in each layer, so as to obtain the error signal of each layer unit, and this error signal is used as the basis for correcting the weights of each unit.
  • This process of adjusting the weights of each layer of the signal forward propagation and error back propagation is repeated.
  • the process of continuous adjustment of weights is the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level, or until the pre-set number of learning times is reached.
  • GD Gradient Descent
  • SGD Stochastic Gradient Descent
  • mini-batch gradient descent Momentum
  • Nesterov which stands for stochastic gradient descent with momentum
  • Adagrad Adaptive gradient descent
  • Adadelta Adaptive learning rate adjustment
  • RMSprop root mean square error prop
  • Adam Adaptive Moment Estimation
  • these optimization algorithms calculate the derivative/partial derivative of the current neuron based on the error/loss obtained by the loss function, add the learning rate, the previous gradient/derivative/partial derivative, etc., and get the gradient, which is passed to the previous neuron. layer.
  • the first AI unit i.e., the AI unit corresponding to the encoder, is deployed on the terminal side. Its input information is channel information, and its output information is channel feature information.
  • the second AI unit i.e., the AI unit corresponding to the decoder, is deployed on the network side. Its input information is the channel characteristic information, and its output information is the recovered channel information.
  • the target AI unit i.e. the proxy AI unit, is used to emulate the functionality of the second AI unit.
  • the first AI unit and the second AI unit may be trained independently, such as the terminal trains the first AI unit and the network side trains the second AI unit.
  • the target AI unit needs to match the performance of the second AI unit actually used on the network side, so the target AI unit needs to be trained on the end with the second AI unit (i.e., the network side).
  • the target AI unit is trained by the terminal.
  • FIG 4 An information processing method provided in an embodiment of the present application, the execution subject of which can be a terminal, wherein the terminal can be various types of terminals 11 listed in Figure 1, or other terminals except the terminal types listed in the embodiment shown in Figure 1, which are not specifically limited here.
  • the information processing method may include the following steps:
  • Step 401 The terminal obtains a first data set.
  • the first data set may be received from a network-side device, such as a terminal receiving the first data set from a base station or a core network device.
  • a network-side device such as a terminal receiving the first data set from a base station or a core network device.
  • the first data set may be received from a third-party node, such as a terminal receiving the first data set from a designated server.
  • Step 402 The terminal trains a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal in reporting target channel information to a network side device, the target channel information is obtained based on a first processing of the first channel information by a first AI unit of the terminal and a second processing of the first channel characteristic information by a second AI unit of the network side device, the first channel characteristic information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • the target AI unit is related to the second AI unit, and may be: the target AI unit and the second AI unit are trained based on at least partially identical sample data, for example: the network side jointly trains the first AI unit and the second AI unit based on 100,000 sample data, and the terminal trains the target AI unit based on 10,000 sample data therein, so that the target AI unit trained based on at least partially identical sample data has the same or similar function as the second AI unit.
  • the target AI unit is related to the second AI unit
  • the first data set may include input information and output information of the second AI unit of the network side device, so that the target AI unit trained based on the input information and output information of the second AI unit has the same or similar function as the second AI unit.
  • the model structure of the target AI unit is simpler than the model structure of the second AI unit, or the parameter scale of the target AI unit is smaller than the parameter scale of the second AI unit.
  • the target AI unit in the embodiment of the present application can simulate the function of the second AI unit, so that the performance of the target AI unit and the performance of the second AI unit can be scaled in a certain proportion, thereby supporting the use of the second channel information restored by the target AI unit to simulate the target channel information restored by the second AI unit, wherein the correlation between the second channel information and the first channel information input to the first AI unit, and the correlation between the target channel information and the first channel information input to the first AI unit, are related or can be linearly converted, that is, the higher the correlation between the second channel information and the first channel information input to the first AI unit, the higher the correlation between the target channel information and the first channel information input to the first AI unit, that is, the better the performance of the terminal in reporting the target channel information to the network side device.
  • a network side device may receive downlink channel state information from a terminal or other node, and jointly train a first AI unit and a second AI unit based on the downlink channel state information, and send at least part of the input information and output information of the second AI unit during the training process as a first data set to the terminal, and the terminal may train a target AI unit based on the first data set.
  • the model structure of the target AI unit may be agreed upon in the protocol, and the terminal may train the target AI unit based on the first data set, which may be the weight parameters of the target AI unit trained by the terminal based on the first data set.
  • the method further includes:
  • the terminal determines first information based on the target AI unit, wherein the first information is related to the performance of the terminal reporting target channel information to a network side device.
  • the terminal determines the first information based on the target AI unit, including:
  • the terminal performs the first processing on the first channel information based on the first AI unit to obtain the first channel characteristic information
  • the terminal performs a third process on the first channel characteristic information based on the target AI unit to obtain second channel information;
  • the terminal determines first information based on a correlation between the second channel information and the first channel information.
  • the first processing may be at least one of compression processing, quantization processing, and normalization processing.
  • the third processing is similar to the recovery processing of the second AI unit of the network side device, for example, the third processing may include at least one of decompression processing, dequantization processing, and denormalization processing.
  • the target AI unit is used to simulate the second AI unit of the network side device to recover the first channel characteristic information reported by the terminal, and the obtained second channel information is the same or similar to the target channel information recovered by the network side device based on the second AI unit on the first channel characteristic information reported by the terminal.
  • the correlation between the target channel information and the first channel information can be reflected.
  • Correlation is used to determine the performance of the terminal in reporting the target channel information to the network side device. That is, the higher the correlation between the second channel information and the first channel information, the better the performance of the terminal in reporting the target channel information to the network side device.
  • the first information includes at least one of the following:
  • the output information of the target AI unit that is, the channel information restored by the target AI unit, the higher the correlation between the channel information restored by the target AI unit and the channel information input to the first AI unit, the better the performance of the terminal in reporting the target channel information to the network side device;
  • a correlation value between the output information of the target AI unit and the input information of the first AI unit for example, a correlation value such as matrix similarity or similarity coefficient of a channel matrix or Euclidean distance or Manhattan distance or square generalized cosine similarity (SGCS);
  • SGCS generalized cosine similarity
  • the terminal reports a performance value of the target channel information to the network side device, wherein the performance value of the target channel information reported by the terminal to the network side device may be positively correlated to a correlation value between the output information of the target AI unit and the input information of the first AI unit, for example: the correlation value between the output information of the target AI unit and the input information of the first AI unit is quantized or normalized to obtain the performance value of the target channel information reported by the terminal to the network side device;
  • Target indication information where the target indication information is used to indicate that the first AI unit or the second AI unit is invalid, wherein the terminal can report the target indication information to the network side device when it is determined that the performance value of the terminal reporting the target channel information to the network side device is lower than the first threshold.
  • the first threshold may be agreed upon in a protocol, determined according to business requirements, set by a user, or indicated by a network-side device.
  • the first threshold can be determined according to a non-AI CSI reporting method, for example: under a codebook-based CSI reporting method, the correlation value between the channel information before compression by the terminal and the channel information after decompression based on the codebook on the network side is used as the first threshold; if the correlation value between the output information of the target AI unit and the input information of the first AI unit is less than or equal to the first threshold, the terminal reports the target indication information to the network side device.
  • a non-AI CSI reporting method for example: under a codebook-based CSI reporting method, the correlation value between the channel information before compression by the terminal and the channel information after decompression based on the codebook on the network side is used as the first threshold; if the correlation value between the output information of the target AI unit and the input information of the first AI unit is less than or equal to the first threshold, the terminal reports the target indication information to the network side device.
  • the terminal may use the target AI unit to detect the performance of the terminal in reporting the target channel information to the network side device.
  • the terminal may also send the first information to a network side device.
  • the network side device can promptly learn whether the first AI unit and the second AI unit are invalid according to the first information. If it is determined that the first AI unit and the second AI unit are invalid, the network side device can instruct the terminal to report the channel information in a non-AI manner (such as a codebook), or update or retrain the first AI unit and the second AI unit to improve the performance of the terminal in reporting channel information to the network side device.
  • a non-AI manner such as a codebook
  • the method further includes:
  • the terminal receives fourth indication information, where the fourth indication information is used to indicate at least one of the following:
  • the terminal determines the first information based on the target AI unit
  • the terminal sends the first information to the network side device
  • the fourth indication information may come from a network side device.
  • the fourth indication information can instruct the terminal to perform an operation of “determining the first information based on the target AI unit”.
  • the fourth indication information can indicate the content of the first information reported by the terminal, such as: the output information of the target AI unit, the correlation value between the output information of the target AI unit and the input information of the first AI unit, the performance value of the terminal reporting the target channel information to the network side device, target indication information, etc.
  • the fourth indication information can indicate whether the terminal reports the first information to the network side device, for example: instructing the terminal not to report the first information to the network side device when it is determined that the first AI unit or the second AI unit is valid; and reporting the first information to the network side device when it is determined that the first AI unit or the second AI unit is invalid.
  • the fourth indication information can indicate a manner in which the terminal reports the first information, such as time domain or frequency domain resources used to report the first information, signaling that carries the first information, and the like.
  • the terminal and the network side device independently train the first AI unit and the second AI unit
  • the terminal and the network side device in order to achieve a reasonable match between the first AI unit trained by the terminal and the second AI unit trained by the network side device, the terminal and the network side device can respectively train the first AI unit and the second AI unit based on the same sample data, or the terminal can receive a second data set from the network side device or a third-party node, wherein the second data set includes: input information of the first AI unit and output information of the first AI unit generated by the network side device in the process of jointly training the first AI unit and the second AI unit.
  • the second data set and the first data set may be indicated independently.
  • the network side device sends the first data set and the second data set to the terminal respectively.
  • the number of data contained in the second data set and the first data set may be different.
  • the number of data in the second data set is greater than the number of data in the first data set.
  • the method further comprises:
  • the terminal obtains a second data set, wherein the second data set includes a first two-tuple corresponding to the sample data one by one, and the first two-tuple includes: input information of the first AI unit and output information of the first AI unit;
  • the terminal trains the first AI unit based on the second data set.
  • the sample data may be channel information.
  • a sample data may be channel information obtained at a sampling point in the time domain or frequency domain.
  • the second data set may be received from a network-side device or from a third-party node, such as a terminal receiving a second data set from a base station or a core network device or a designated server.
  • the network side device jointly trains the first AI unit and the second AI unit based on a large amount of sample data.
  • the training process for each sample data, there will be the following training process: input the sample data into the first AI unit, obtain the channel characteristic information output by the first AI unit, input the channel characteristic information output by the first AI unit into the second AI unit, and obtain the channel information output by the second AI unit.
  • the network side device sends at least part of the input information and output information of the first AI unit as the second data set to the terminal, so that the terminal can independently train the first AI unit.
  • the terminal independently trains the first AI unit, and the terminal receives the second data set from the network side device, so that the first AI unit trained based on the second data set can match the second AI unit trained by the network side device.
  • the second data set and the first data set may be jointly indicated, wherein, given that the first data set may include input information of the second AI unit and output information of the second AI unit generated by the network side device during the joint training of the first AI unit and the second AI unit, the second data set includes: input information of the first AI unit and output information of the first AI unit generated by the network side device during the joint training of the first AI unit and the second AI unit.
  • the second data set and the first data set may be jointly indicated by a triplet, wherein the triplet may include: the input information of the first AI unit, the input information of the second AI unit, and the output information of the second AI unit; or, the triplet may include: the input information of the first AI unit, the output information of the first AI unit, and the output information of the second AI unit.
  • the first data set includes at least one of the following:
  • a second two-tuple corresponding to the sample data one by one comprising: input information of the second AI unit and output information of the second AI unit;
  • a triplet corresponding to the sample data one by one including: input information of the first AI unit, output information of the first AI unit, and output information of the second AI unit, or the triplet including: input information of the first AI unit, input information of the second AI unit, and output information of the second AI unit;
  • a four-tuple corresponding to the sample data one by one comprising: input information of the first AI unit, output information of the first AI unit, input information of the second AI unit, and output information of the second AI unit.
  • the first data set when the first data set and the second data set are independently indicated, the first data set includes input information of the second AI unit and output information of the second AI unit generated by the network side device during the joint training of the first AI unit and the second AI unit.
  • the first data set when the first data set and the second data set are jointly indicated, the first data set includes a triplet corresponding one-to-one to the sample data, and the first two bits of the triplet are used as the terminal training first AI unit data set, that is, the input information of the first AI unit and the output information of the first AI unit, or the input information of the first AI unit and the input information of the second AI unit are used for the terminal training the first AI unit; the last two bits of the triplet are used as the terminal training target AI unit data set, that is, the output information of the first AI unit and the output information of the second AI unit, or the input information of the second AI unit and the output information of the second AI unit are used for the terminal training the target AI unit.
  • the first data set when the first data set and the second data set are jointly indicated, includes A four-tuple corresponding to the sample data, the first two bits of which are used for the terminal to train the first AI unit, and the last two bits are used for the terminal to train the target AI unit.
  • the output information of the first AI unit in the above-mentioned quaternary group and the input information of the second AI unit may be different.
  • the terminal uses a specific quantization scheme for the output information of the first AI unit, and reports the quantized channel characteristic information to the network side. Then, the output information of the first AI unit and the input information of the second AI unit may be different.
  • the terminal obtains the first data set, including:
  • the terminal receives second information from the first device, where the second information includes a first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set.
  • the first device may be a network-side device, which may be used to jointly train the first AI unit and the second AI unit.
  • the network-side device sends the input information and output information of the second AI unit during the joint training of the first AI unit and the second AI unit to the terminal as training data, so that the terminal can train the target AI unit.
  • the first data set may also include the input information and output information of the first AI unit during the joint training of the first AI unit and the second AI unit by the network-side device, and the terminal may independently train the first AI unit based on the information.
  • the first device may be a third-party node, such as a server.
  • the third-party node may obtain the first data set from the network-side device, or the third-party node may jointly train the first AI unit and the second AI unit.
  • the terminal may directly obtain the first data set from the first device.
  • the terminal may pre-acquire the association relationship between the first data set and the identifier of the first data set, such as: the protocol stipulates each first data set and its corresponding identifier, or the network side device pre-configures each first data set and its corresponding identifier. In this way, when the second information includes the identifier of the first data set, the terminal may determine the first data set according to the association relationship between the first data set and the identifier of the first data set.
  • the terminal may have at least two first AI units.
  • the method further includes:
  • the terminal determines, based on a first association relationship, that the target AI unit is used to determine first information related to the first AI unit, wherein the first association relationship indicates an association relationship between the target AI unit and the first AI unit.
  • the target AI unit is used to detect the performance of the channel information processed by the associated first AI unit. For example, assuming that the target AI unit is associated with the first AI unit A, and the first AI unit A matches the second AI unit A, the target AI unit can be used to detect the correlation between the input information of the first AI unit A and the output information of the second AI unit A, thereby determining the performance of the channel information reported to the network side using the first AI unit A and the second AI unit A.
  • the first association relationship includes at least one of the following:
  • the target AI unit is used to train at least one of the first AI units, and the target AI unit is used to determine first information related to a first AI unit trained based on the second data set;
  • the target AI unit may be associated with the first AI unit, or, given that the first AI unit matches the corresponding second AI unit, the target AI unit may also be associated with the second AI unit.
  • the target AI unit can be associated with a second data set or a second data set ID, wherein a second data set can be used to train one or at least two first AI units, which implicitly implies that the target AI unit is associated with one or at least two first AI units trained based on the same second data set.
  • a protocol agreement or network-side pre-configuration method can be adopted to enable the terminal to obtain the second data set corresponding to the second data set ID.
  • the first data set is associated with the first AI unit or the first AI unit ID.
  • the first data set only includes the input and output of the decoder, it is necessary to additionally inform which first AI units the first data set is associated with, that is, which encoder input and output data set the first data set corresponds to.
  • a first data set can be used to train at least one target AI unit. At this time, it is implicit that one or at least two target AI units trained based on the same first data set are associated with the first AI unit corresponding to the first data set.
  • the first data set is associated with the second AI unit or the second AI unit ID, wherein, given that the first AI unit matches the corresponding second AI unit, a first data set can be used to train at least one target AI unit. In this case, it is implicitly associated with one or at least two target AI units trained based on the same first data set and the second AI unit corresponding to the first data set.
  • the first data set is associated with the second data set or the second data set ID.
  • the target AI unit trained based on the first data set can be used to detect the performance of the first AI unit trained based on the associated second data set.
  • the second data set is used to train M first AI units
  • the first data set is used to train N target AI units
  • the first association relationship includes an association relationship between the N target AI units and the M first AI units, where N and M are positive integers respectively.
  • N is greater than M, and one of the N target AI units is associated with at least one first AI unit of the M first AI units;
  • N is less than M, at least one of the N target AI units is connected to the M first AI A first AI unit in the unit is associated;
  • N is equal to M, and the N target AI units are associated with the M first AI units in a one-to-one correspondence.
  • the above-mentioned first association relationship may be indicated by a network-side device or a third-party node, or determined by the terminal.
  • the method further comprises at least one of the following:
  • the terminal receives first indication information, where the first indication information indicates the first association relationship
  • the terminal associates a target AI unit trained based on first data with a first AI unit trained based on second data, wherein the first data and the second data correspond to the same sample data.
  • the first indication information comes from at least one of the following:
  • a network side device such as a base station or a core network device, for example: a third-party node is responsible for transmitting at least one of the first data set and the second data set, and the base station or the core network device controls the first association relationship.
  • the first indication information may be indicated by a data set ID, for example: the first association relationship includes an association relationship between a first data set ID and a first AI unit ID.
  • the first indication information may be indicated by at least one of downlink control information (Downlink Control Information, DCI), medium access control layer control element (Medium Access Control Control Element, MAC CE), and radio resource control (Radio Resource Control, RRC) signaling; or,
  • DCI Downlink Control Information
  • MAC CE Medium Access Control Control Element
  • RRC Radio Resource Control
  • the first indication information may be carried in the relevant information of the AI unit, for example, in the process of transmitting or configuring the second data set, associating the second data set with the corresponding first data set.
  • the terminal may autonomously decide to associate the first data and the second data corresponding to the same sample data, that is, to associate the target AI unit trained based on the first data and the first AI unit trained based on the second data.
  • the first data table is data used to train the target AI unit, and may specifically include the input and output of the second AI unit; the second data table is data used to train the first AI unit, and may specifically include the input and output of the first AI unit.
  • the terminal can automatically associate the target AI unit trained by the last two digits with the first AI unit trained by the first two digits.
  • the method further includes:
  • the terminal sends target capability information, wherein the target capability information indicates at least one of the following:
  • the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit;
  • the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device;
  • the network side device will send the first data set to the terminal only when the target capability information reported by the terminal to the network side device indicates that the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit.
  • the target capability information reported by the terminal to the network side device indicates that the terminal uses the
  • the network side device can configure or instruct the terminal to report the first information.
  • the network side device will send the first data set to the terminal only when the target capability information reported by the terminal to the network side device indicates that the terminal supports training the target AI unit.
  • the terminal can train the target AI unit if the terminal supports it, or detect the performance of the terminal reporting target channel information to the network side device based on the target AI unit, thereby avoiding unnecessary waste of computing power and air interface resources due to lack of terminal support.
  • the method further includes:
  • the terminal If the terminal supports independent training of the first AI unit, the terminal confirms support for training the target AI unit;
  • the terminal confirms that training of the target AI unit is not supported.
  • the terminal when the terminal supports independent training of the first AI unit, it is assumed that the terminal supports training of the target AI unit; when the terminal does not support independent training of the first AI unit and only supports AI unit transfer (i.e., the network side jointly trains the first AI unit and the second AI unit, and transfers the first AI unit to the terminal or transfers the first AI unit and the second AI unit), it is assumed that the terminal does not support training of the target AI unit. In this way, the terminal can directly determine whether the terminal supports training of the target AI unit based on whether it supports independent training capabilities.
  • the method further includes:
  • the terminal sends target status information related to the target AI unit.
  • the target status information can be used by the network side device to determine whether the terminal needs to train a new target AI unit, thereby sending the corresponding first data set to the terminal accordingly.
  • the target state information indicates at least one of the following:
  • each second data set has an associated target AI unit
  • each first AI unit has an associated target AI unit
  • each first AI unit requires a new target AI unit.
  • the target AI unit when each second data set has an associated target AI unit, can be used to perform performance detection on the first AI unit trained based on the respective associated second data sets.
  • the network side device can send the first data sets associated with these second data sets to the terminal, so that the terminal can train the target AI units associated with these second data sets accordingly.
  • the target AI unit when each first AI unit has an associated target AI unit, can be used to perform performance detection on the first AI units based on their respective associations.
  • the network-side device can send the first data sets associated with these first AI units to the terminal, so that the terminal can train the target AI units associated with these first AI units accordingly.
  • the terminal may directly determine which second data sets require new target AI units, for example For example, some second data sets have no associated target AI units, or some target AI units associated with the second data sets become invalid due to the movement of the terminal location, changes in the communication environment, etc.
  • the network-side device can send the first data sets associated with these second data sets to the terminal, so that the terminal can train the target AI units associated with these second data sets accordingly.
  • the terminal may directly determine which first AI units need new target AI units, for example, some first AI units have no associated target AI units, or the target AI units associated with some first AI units are invalid due to the movement of the terminal, changes in the communication environment, etc.
  • the network-side device may send the first data sets associated with these first AI units to the terminal, so that the terminal can train the target AI units associated with these first AI units accordingly.
  • the terminal may periodically report target status information.
  • the terminal sends the target status information related to the target AI unit to the network side device or the third-party node, so that the network side device or the third-party node can promptly know which second data sets or target AI units associated with the first AI unit that the terminal needs to train or update, so that the corresponding first data set can be sent to the terminal accordingly.
  • the method further includes:
  • the terminal sends second indication information, where the second indication information indicates first identification information of a target AI unit required by the terminal, and the first identification information is used to identify a second data set or a first AI unit.
  • the network side device may determine which second data sets or first AI units can use the target AI unit based on the second indication information, thereby issuing the corresponding first data set so that the terminal can train the corresponding target AI unit.
  • the network side device may determine which second data sets or first AI units have associated target AI units based on the second indication information, thereby instructing the terminal to use the corresponding target AI unit to perform performance detection on the target channel information reported by the terminal.
  • the terminal may have multiple second data sets.
  • the terminal will train an encoder for each second data set, and use the corresponding first AI unit to compress the CSI in a certain environment.
  • not all second data sets can have a target AI unit.
  • some channel environments are complex, and simple target AI units are not sufficient for decoding; or, some second data sets may lack the corresponding target AI unit data set, resulting in the inability to train the target AI unit; or, some target AI units are difficult to train and cannot be implemented by the terminal. Therefore, for terminals that support target AI units, not all second data sets can train the target AI unit.
  • the terminal can directly report the second data set or the first AI unit that can be detected using the target AI unit, and the base station indicates which second data sets or first AI units use the target AI unit.
  • the method further includes:
  • the terminal sends target request information, wherein the target request information is used to request training of a target AI unit associated with the first AI unit;
  • the terminal obtains a first data set, including:
  • the terminal receives a first data set corresponding to the first AI unit.
  • the terminal requests to train a target AI unit for a certain group of first AI units, and the network side device sends the corresponding first data set after permission.
  • the method further includes:
  • the terminal receives third indication information, wherein the third indication information is used to instruct the terminal to train a target AI unit associated with the first AI unit, or to train a target AI unit associated with a second data set, where the second data set is used to train at least one of the first AI units.
  • the network side device instructs the terminal to train the target AI units associated with those first AI units or the second data set.
  • a target AI unit in an AI-based channel information reporting scenario, is trained on the terminal side, and the target AI unit is related to a second AI unit used on the network side.
  • the AI unit can be used to detect the performance of the terminal in reporting the target channel information to the network side device, so as to promptly discover the problem of inaccurate AI-based compression-recovery channel information.
  • the execution subject of the information processing method is a first device
  • the first device may include a network side device, such as a base station or a core network device, or the first device may include a third-party node, such as a server.
  • a network side device such as a base station or a core network device
  • the first device may include a third-party node, such as a server.
  • the information processing method may include the following steps:
  • Step 501 A first device obtains downlink channel state information.
  • the first device may receive downlink channel state information from the terminal.
  • the first device may obtain downlink channel state information that is pre-stored or agreed upon by protocol.
  • Step 502 The first device trains a first AI unit and a second AI unit based on the downlink channel state information.
  • Step 503 The first device sends second information to the terminal, where the second information includes the first data set, or the second information includes the identifier of the first data set, and the terminal learns the association relationship between the first data set and the identifier of the first data set.
  • the first data set is used to train a target AI unit, and the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, wherein the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit, and the first channel characteristic information is information obtained based on the first processing of the first channel information by the first AI unit of the terminal.
  • first AI unit, second AI unit, second information, first data set, identifier of the first data set, target channel information, first channel information, first channel characteristic information, first processing, and second processing have the same meanings as the first AI unit, second AI unit, second information, first data set, identifier of the first data set, target channel information, first channel information, first channel characteristic information, first processing, and second processing in the method embodiment shown in Figure 4, and are not repeated here.
  • the method further includes:
  • the first device sends a second data set to the terminal, wherein the second data set is used by the terminal to train the first AI unit, wherein the second data set includes a first two-tuple corresponding one-to-one to the sample data, and the first two-tuple includes: input information of the first AI unit and output information of the first AI unit.
  • the second data set has the same meaning and function as the second data set in the method embodiment shown in FIG. 4 , and will not be described in detail herein.
  • the first device sends to the terminal: input information of the first AI unit and output information of the first AI unit generated during the joint training of the first AI unit and the second AI unit, which can improve the matching degree between the first AI unit trained by the terminal and the second AI unit of the first device.
  • the first data set includes at least one of the following:
  • a second two-tuple corresponding to the sample data one by one comprising: input information of the second AI unit and output information of the second AI unit;
  • a triplet corresponding to the sample data one by one including: input information of the first AI unit, output information of the first AI unit, and output information of the second AI unit, or the triplet including: input of the first AI unit, input of the second AI unit, and output of the second AI unit;
  • a four-tuple corresponding to the sample data one by one comprising: input information of the first AI unit, output information of the first AI unit, input information of the second AI unit, and output information of the second AI unit.
  • the method further includes:
  • the first device sends first indication information to the terminal, where the first indication information indicates a first association relationship, wherein the first association relationship indicates an association relationship between the target AI unit and the first AI unit.
  • the first association relationship includes at least one of the following:
  • the target AI unit is used to train at least one of the first AI units, and the target AI unit is used to determine first information related to a first AI unit trained based on the second data set;
  • the target AI unit when the target AI unit is associated with a second data set, the second data set is used to train M first AI units, and the first data set is used to train N target AI units, and the first association relationship includes an association relationship between the N target AI units and the M first AI units, where N and M are positive integers.
  • N is greater than M, and one of the N target AI units is associated with at least one first AI unit of the M first AI units;
  • N is less than M, at least one of the N target AI units is associated with one of the M first AI units;
  • N is equal to M, and the N target AI units are associated with the M first AI units in a one-to-one correspondence.
  • the method further includes:
  • the first device receives target capability information sent from the terminal, wherein the target capability information indicates at least one of the following:
  • the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit;
  • the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device;
  • the first device sends second information to the terminal, including:
  • the target capability information indicates that the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit, or the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device, or the terminal supports training the target AI unit
  • the first device sends second information to the terminal.
  • the method further includes:
  • the first device receives target state information related to the target AI unit from the terminal.
  • the target state information indicates at least one of the following:
  • each second data set has an associated target AI unit
  • each first AI unit has an associated target AI unit
  • each first AI unit requires a new target AI unit.
  • the method further includes:
  • the first device receives second indication information from the terminal, where the second indication information indicates first identification information of a target AI unit required by the terminal, and the first identification information is used to identify a second data set or a first AI unit.
  • the method further includes:
  • the first device receives target request information from the terminal, wherein the target request information is used to request training of a target AI unit associated with the first AI unit;
  • the second information includes a first data set corresponding to the first AI unit, or an identifier of the first data set corresponding to the first AI unit.
  • the method further includes:
  • the first device sends third indication information to the terminal, wherein the third indication information is used to instruct the terminal to train a target AI unit associated with the first AI unit, or to train a target AI unit associated with a second data set, wherein the second data set is used to train at least one of the first AI units.
  • the method further includes:
  • the first device receives first information from the terminal, wherein the first information is related to the performance of the terminal reporting target channel information to a network side device.
  • the method further includes:
  • the first device sends fourth indication information to the terminal, where the fourth indication information is used to indicate at least one of the following:
  • the terminal determines the first information based on the target AI unit
  • the terminal sends the first information to the network side device
  • the first device can jointly train the first AI unit and the second AI unit, and send a first data set generated in the process of jointly training the first AI unit and the second AI unit to the terminal, such as the input and output of the second AI unit.
  • the terminal can train the target AI unit related to the above-mentioned second AI unit based on this. It has similar beneficial effects to the method embodiment shown in Figure 4. To avoid repetition, it will not be repeated here.
  • the information interaction process between the base station and the terminal is taken as an example to illustrate the information processing method provided in the embodiment of the present application:
  • Scenario 1 Independently indicate the first data set and manage the first association relationship based on the model ID
  • a typical AI unit independent training process includes the following steps:
  • Step 1a During the capability reporting phase, the terminal informs the base station that it can support separate training and can train the proxy AI model (i.e., the target AI unit) under the separate training framework.
  • the proxy AI model i.e., the target AI unit
  • the base station collects downlink channel data.
  • the base station trains a set of encoder-decoder models for CSI compression based on the collected data, and obtains a second data set (ie, the input and output of the encoder) based on the trained encoder model.
  • the base station sends the second data set to the terminal.
  • the terminal receives the second data set, and trains one or more encoders on the terminal side based on the second data set, and then assigns a model ID to each encoder on the terminal side, and then reports the model ID to the base station.
  • the base station associates the (encoder) model ID reported by the terminal with the local decoder model ID, thereby ensuring that the encoder on the terminal side and the decoder measured by the base station can be used in pair.
  • the terminal requests the base station to train the relevant proxy AI model for one or more local encoders.
  • the base station sends one or more first data sets (ie, input and output of the decoder) and a decoder ID and/or encoder ID associated with each first data set to the terminal according to the proxy model training request of the terminal.
  • first data sets ie, input and output of the decoder
  • decoder ID and/or encoder ID associated with each first data set to the terminal according to the proxy model training request of the terminal.
  • the terminal trains the agent AI model according to each first data set.
  • the terminal associates each proxy AI model with a local encoder model according to the decoder ID and/or encoder ID associated with the corresponding first data set.
  • the terminal reports to the base station whether the proxy AI model is successfully trained on each first data set.
  • the base station instructs the terminal to use the proxy AI model for performance detection based on the proxy AI model status reported by the user.
  • the content of the instruction includes the reporting content and reporting method of the performance monitoring.
  • the user uses the proxy AI model to perform model monitoring.
  • a specific implementation method is as follows: the terminal inputs the downlink channel measurement into the encoder-proxy AI model, obtains the output of the proxy AI model, calculates the square generalized cosine similarity (SGCS) value between the output and the CSI measurement value, and processes the SGCS value according to the instructions of the base station and reports it.
  • SGCS square generalized cosine similarity
  • the terminal periodically reports whether there is an available proxy AI model for each local encoder.
  • the encoder can be updated periodically. After the encoder is updated, its corresponding proxy AI model may also need to be updated, or the proxy AI model may become invalid as the terminal changes position or communication environment, etc.
  • the terminal periodically reports whether there is an available proxy AI model for each local encoder, so that the network-side device can know whether the encoder's proxy AI model needs to be updated or trained.
  • a typical AI unit independent training process includes the following steps:
  • the terminal informs the base station that it can support separate training and can train the proxy AI model (i.e., the target AI unit) under the separate training framework.
  • the proxy AI model i.e., the target AI unit
  • the base station collects downlink channel data.
  • the base station trains a set of encoder-decoder models for CSI compression based on the collected data, and obtains the second data set (i.e., the input and output of the encoder) and the first data set (i.e., the input and output of the decoder) based on the trained encoder model.
  • the base station sends the second data set and the first data set jointly to the terminal.
  • One way of jointly sending is to send data using a triplet (encoder input-encoder output-decoder output). This method assumes that the decoder input is the same as the encoder output.
  • the terminal receives the first data set and the second data set, trains one or more terminal-side encoders based on the encoder input-output data set (second data set), and then trains one or more proxy AI models based on the decoder input-output data set (first data set).
  • the terminal associates the trained proxy AI model with the local encoder and assigns a model ID to each encoder, and then reports the encoder model ID, the second data set corresponding to the encoder model ID, and whether an available proxy AI model has been trained for the model to the base station.
  • the base station associates the reported encoder ID with the local decoder ID and pairs the encoder and decoder.
  • the base station instructs the terminal to use the proxy AI model for performance detection based on the proxy AI model status reported by the user.
  • the content of the instruction includes the reporting content and reporting method of the performance monitoring.
  • the user uses the proxy AI model to perform model monitoring.
  • a specific implementation method is as follows: the terminal inputs the downlink channel measurement into the encoder-proxy model, obtains the output of the proxy model, calculates the square cosine similarity (SGCS) between the output and the CSI measurement value, and processes the SGCS value according to the instructions of the base station and reports it.
  • SGCS square cosine similarity
  • the terminal periodically reports whether there is an available proxy AI model for each local encoder.
  • Scenario 3 Indicate the first data set independently and manage the first association relationship based on the function ID (also called data set ID)
  • a typical AI unit independent training process includes the following steps:
  • the terminal informs the base station that it can support separate training and can train the proxy AI model (target AI unit) under the separate training framework.
  • the base station collects downlink channel data.
  • the base station trains a set of encoder-decoder models for CSI compression based on the collected data, and obtains a second data set (ie, a data set consisting of the input and output of the encoder) based on the trained encoder model.
  • the base station sends the second data set and the ID of the second data set to the terminal.
  • the terminal receives the second data set, and trains one or more encoders on the terminal side based on the second data set.
  • the terminal requests the base station to train a related proxy AI model for one or more second data sets.
  • the base station sends one or more first data sets (i.e., the input and output of the decoder) and a second data set ID associated with each first data set to the terminal according to the terminal's proxy AI model training request.
  • first data sets i.e., the input and output of the decoder
  • second data set ID associated with each first data set
  • the terminal trains the proxy AI model according to each first data set, associates each trained proxy AI model with the corresponding second data set ID, and reports to the base station whether the proxy AI model is successfully trained on each first data set.
  • the base station instructs the terminal to use the proxy AI model to perform model performance detection based on the proxy AI model status reported by the user.
  • the instructions include the reporting content and reporting method of performance monitoring.
  • the user uses the proxy AI model to perform model monitoring.
  • a specific implementation method is: the terminal inputs the downlink channel measurement into the encoder-proxy AI model, obtains the output of the proxy AI model, calculates the square cosine similarity (SGCS) between the output and the CSI measurement value, and processes the SGCS value according to the instructions of the base station and reports it.
  • SGCS square cosine similarity
  • the terminal periodically reports whether there is an available proxy AI model for each local second data set.
  • Scenario 4 Jointly indicating the first data set and the second data set, and managing the first association relationship based on the function ID (also referred to as the data set ID)
  • a typical AI unit independent training process includes the following steps:
  • the terminal informs the base station that it can support separate training and can train the proxy AI model (i.e., the target AI unit) under the separate training framework.
  • the proxy AI model i.e., the target AI unit
  • the base station collects downlink channel data.
  • the base station trains a set of encoder-decoder models for CSI compression based on the collected data, and obtains the second data set (ie, the input and output of the encoder) and the first data set (ie, the input and output of the decoder) based on the trained encoder model.
  • the base station sends the second data set and the first data set jointly to the terminal.
  • One way of jointly sending is to send data using a triplet (encoder input-encoder output-decoder output). This method assumes that the decoder input is the same as the encoder output.
  • the terminal receives the first data set and the second data set, trains one or more terminal-side encoders based on the encoder input-output data set (second data set), and then trains one or more proxy AI models based on the decoder input-output data set (first data set).
  • the terminal associates the trained proxy AI model with the second data set ID, and reports to the base station whether an available proxy AI model has been trained for the second data set.
  • the base station instructs the terminal to use the proxy AI model for performance detection based on the proxy AI model status reported by the user.
  • the instructions include the reporting content and reporting method of the performance monitoring.
  • the user uses the proxy AI model to perform model monitoring.
  • a specific implementation method is: the terminal inputs the downlink channel measurement into the encoder-proxy AI model, obtains the output of the proxy AI model, calculates the square cosine similarity (SGCS) between the output and the CSI measurement value, and processes the SGCS value according to the instructions of the base station and reports it.
  • SGCS square cosine similarity
  • the terminal periodically reports whether there is an available proxy AI model for each local encoder.
  • the information processing method provided in the embodiment of the present application can be executed by an information processing device.
  • the information processing device provided in the embodiment of the present application is described by taking the information processing device executing the information processing method as an example.
  • the information processing device 600 provided in the embodiment of the present application may be a device in a terminal. As shown in FIG6 , the information processing device 600 may include the following modules:
  • a first acquisition module 601 is used to acquire a first data set
  • the first training module 602 is used to train a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit of the network side device, the first channel characteristic information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • the information processing device 600 further includes:
  • a second acquisition module is used to acquire a second data set, wherein the second data set includes a first two-tuple corresponding to the sample data one by one, and the first two-tuple includes: input information of the first AI unit and output information of the first AI unit;
  • the second training module is used to train the first AI unit based on the second data set.
  • the information processing device 600 further includes:
  • the first determination module is used to determine first information based on the target AI unit, wherein the first information is related to the performance of the terminal reporting target channel information to the network side device.
  • the first determining module is specifically used to:
  • the first channel information is subjected to the first processing to obtain the first channel characteristic. Solicit information;
  • first information is determined.
  • the first data set includes at least one of the following:
  • a second two-tuple corresponding to the sample data one by one comprising: input information of the second AI unit and output information of the second AI unit;
  • a triplet corresponding to the sample data one by one including: input information of the first AI unit, output information of the first AI unit, and output information of the second AI unit, or the triplet including: input information of the first AI unit, input information of the second AI unit, and output information of the second AI unit;
  • a four-tuple corresponding to the sample data one by one comprising: input information of the first AI unit, output information of the first AI unit, input information of the second AI unit, and output information of the second AI unit.
  • the first acquisition module 601 is specifically configured to:
  • Second information is received from a first device, where the second information includes a first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set.
  • the information processing device 600 further includes:
  • the second determination module is used to determine the target AI unit based on a first association relationship to determine first information related to the first AI unit, wherein the first association relationship indicates an association relationship between the target AI unit and the first AI unit.
  • the first association relationship includes at least one of the following:
  • the target AI unit is used to train at least one of the first AI units, and the target AI unit is used to determine first information related to a first AI unit trained based on the second data set;
  • the information processing device 600 further includes at least one of the following:
  • a first receiving module configured to receive first indication information, where the first indication information indicates the first association relationship
  • An association module is used to associate a target AI unit trained based on first data with a first AI unit trained based on second data, wherein the first data and the second data correspond to the same sample data.
  • the second data set is used to train M first AI units
  • the first data set is used to train N target AI units
  • the first association relationship includes an association relationship between the N target AI units and the M first AI units, where N and M are positive integers respectively.
  • N is greater than M, and one of the N target AI units is associated with at least one first AI unit of the M first AI units;
  • N is less than M, at least one of the N target AI units is associated with one of the M first AI units;
  • N is equal to M, and the N target AI units are associated with the M first AI units in a one-to-one correspondence.
  • the information processing device 600 further includes:
  • the third sending module is configured to send target capability information, wherein the target capability information indicates at least one of the following:
  • the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit;
  • the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device;
  • the information processing device 600 further includes:
  • a third determination module configured to confirm support for training the target AI unit if the terminal supports independent training of the first AI unit
  • the fourth determination module is used to confirm that training of the target AI unit is not supported when the terminal only supports AI unit transmission.
  • the information processing device 600 further includes:
  • the fourth sending module is used to send target state information related to the target AI unit.
  • the target state information indicates at least one of the following:
  • each second data set has an associated target AI unit
  • each first AI unit has an associated target AI unit
  • each first AI unit requires a new target AI unit.
  • the information processing device 600 further includes:
  • the fifth sending module is used to send second indication information, where the second indication information indicates first identification information of the target AI unit required by the terminal, and the first identification information is used to identify the second data set or the first AI unit.
  • the information processing device 600 further includes:
  • a sixth sending module configured to send target request information, wherein the target request information is used to request training of a target AI unit associated with the first AI unit;
  • the first acquisition module 601 is specifically used for:
  • a first data set corresponding to the first AI unit is received.
  • the information processing device 600 further includes:
  • the second receiving module is used to receive third indication information, wherein the third indication information is used to instruct the terminal to train a target AI unit associated with the first AI unit, or to train a target AI unit associated with a second data set, and the second data set is used to train at least one of the first AI units.
  • the information processing device 600 further includes:
  • the fifth sending module is used to send the first information to the network side device.
  • the information processing device 600 further includes:
  • the third receiving module is configured to receive fourth indication information, wherein the fourth indication information is used to indicate at least one of the following:
  • the terminal determines the first information based on the target AI unit
  • the terminal sends the first information to the network side device
  • the information processing device 600 in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or a component in an electronic device, such as an integrated circuit or a chip.
  • the electronic device may be a terminal.
  • the terminal may include but is not limited to the types of the terminal 11 listed above, and the embodiment of the present application does not specifically limit this.
  • the information processing device 600 provided in the embodiment of the present application can implement each process implemented by the method embodiment shown in Figure 4 and achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the first device may be a network side device, such as a base station or a core network device, or the first device may be a third-party device, such as a server.
  • the information processing device 700 may include the following modules:
  • the third acquisition module 701 is used to acquire downlink channel state information
  • a third training module 702 configured to train the first AI unit and the second AI unit based on the downlink channel state information
  • a first sending module 703 is configured to send second information to a terminal, where the second information includes a first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set;
  • the first data set is used to train a target AI unit, and the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, wherein the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel characteristic information by the second AI unit, and the first channel characteristic information is information obtained based on the first processing of the first channel information by the first AI unit of the terminal.
  • the information processing device 700 further includes:
  • a second sending module configured to send a second data set to the terminal, wherein the second data set is used by the terminal to train the first AI unit, wherein the second data set includes a first two-tuple corresponding to the sample data in a one-to-one manner,
  • the first tuple includes: input information of the first AI unit and output information of the first AI unit.
  • the first data set includes at least one of the following:
  • a second two-tuple corresponding to the sample data one by one comprising: input information of the second AI unit and output information of the second AI unit;
  • a triplet corresponding to the sample data one by one including: input information of the first AI unit, output information of the first AI unit, and output information of the second AI unit, or the triplet including: input of the first AI unit, input of the second AI unit, and output of the second AI unit;
  • a four-tuple corresponding to the sample data one by one comprising: input information of the first AI unit, output information of the first AI unit, input information of the second AI unit, and output information of the second AI unit.
  • the information processing device 700 further includes:
  • a seventh sending module is used to send first indication information to the terminal, where the first indication information indicates a first association relationship, wherein the first association relationship indicates an association relationship between the target AI unit and the first AI unit.
  • the first association relationship includes at least one of the following:
  • the target AI unit is used to train at least one of the first AI units, and the target AI unit is used to determine first information related to a first AI unit trained based on the second data set;
  • the target AI unit when the target AI unit is associated with a second data set, the second data set is used to train M first AI units, the first data set is used to train N target AI units, and the first association relationship includes an association relationship between the N target AI units and the M first AI units, where N and M are positive integers respectively.
  • N is greater than M, and one of the N target AI units is associated with at least one first AI unit of the M first AI units;
  • N is less than M, at least one of the N target AI units is associated with one of the M first AI units;
  • N is equal to M, and the N target AI units are associated with the M first AI units in a one-to-one correspondence.
  • the information processing device 700 further includes:
  • the fourth receiving module is used to receive target capability information sent from the terminal, wherein the target capability information Indicate at least one of the following:
  • the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit;
  • the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device;
  • the first sending module 703 is specifically used for:
  • the target capability information indicates that the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit, or the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device, or the terminal supports training the target AI unit
  • the second information is sent to the terminal.
  • the information processing device 700 further includes:
  • the fifth receiving module is used to receive target state information related to the target AI unit from the terminal.
  • the target state information indicates at least one of the following:
  • each second data set has an associated target AI unit
  • each first AI unit has an associated target AI unit
  • each first AI unit requires a new target AI unit.
  • the information processing device 700 further includes:
  • the sixth receiving module is used to receive second indication information from the terminal, where the second indication information indicates first identification information of a target AI unit required by the terminal, and the first identification information is used to identify a second data set or a first AI unit.
  • the information processing device 700 further includes:
  • a seventh receiving module configured to receive target request information from the terminal, wherein the target request information is used to request training of a target AI unit associated with the first AI unit;
  • the second information includes a first data set corresponding to the first AI unit, or an identifier of the first data set corresponding to the first AI unit.
  • the information processing device 700 further includes:
  • a seventh sending module configured to send third indication information to the terminal, wherein the third indication information is used to instruct the terminal to train a target AI unit associated with the first AI unit, or to train a target AI unit associated with a second data set, wherein the second data set is used to train at least one of the first AI units.
  • the information processing device 700 further includes:
  • An eighth receiving module is used to receive first information from the terminal, wherein the first information is related to the performance of the terminal reporting target channel information to a network side device.
  • the information processing device 700 further includes:
  • an eighth sending module configured to send fourth indication information to the terminal, wherein the fourth indication information is used to indicate at least one of the following:
  • the terminal determines the first information based on the target AI unit
  • the terminal sends the first information to the network side device
  • the information processing device 700 provided in the embodiment of the present application can implement each process implemented by the method embodiment shown in Figure 5 and achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the embodiment of the present application further provides a communication device 800, including a processor 801 and a memory 802, wherein the memory 802 stores a program or instruction that can be run on the processor 801.
  • the communication device 800 is a terminal
  • the program or instruction is executed by the processor 801 to implement the various steps of the information processing method embodiment shown in FIG4 , and can achieve the same technical effect.
  • the communication device 800 is a first device
  • the program or instruction is executed by the processor 801 to implement the various steps of the information processing method embodiment shown in FIG5 , and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the embodiment of the present application also provides a terminal, including a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps in the method embodiment shown in Figure 4.
  • This terminal embodiment corresponds to the above-mentioned terminal side method embodiment, and each implementation process and implementation method of the above-mentioned method embodiment can be applied to the terminal embodiment and can achieve the same technical effect.
  • Figure 9 is a schematic diagram of the hardware structure of a terminal implementing an embodiment of the present application.
  • the terminal 900 includes but is not limited to: a radio frequency unit 901, a network module 902, an audio output unit 903, an input unit 904, a sensor 905, a display unit 906, a user input unit 907, an interface unit 908, a memory 909 and at least some of the components of a processor 910.
  • the terminal 900 may also include a power source (such as a battery) for supplying power to each component, and the power source may be logically connected to the processor 910 through a power management system, so as to implement functions such as managing charging, discharging, and power consumption management through the power management system.
  • a power source such as a battery
  • the terminal structure shown in FIG9 does not constitute a limitation on the terminal, and the terminal may include more or fewer components than shown in the figure, or combine certain components, or arrange components differently, which will not be described in detail here.
  • the input unit 904 may include a graphics processing unit (GPU) 9041 and a microphone 9042, and the graphics processor 9041 processes the image data of the static picture or video obtained by the image capture device (such as a camera) in the video capture mode or the image capture mode.
  • the display unit 906 may include a display panel 9061, and the display panel 9061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc.
  • the user input unit 907 includes a touch panel 9071 and at least one of other input devices 9072.
  • the touch panel 9071 is also called a touch screen.
  • the touch panel 9071 may include two parts: a touch detection device and a touch controller.
  • Other input devices 9072 may include, but are not limited to, a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.
  • the RF unit 901 can transmit the data to the processor 910 for processing; in addition, the RF unit 901 can send uplink data to the network side device.
  • the RF unit 901 includes but is not limited to an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc.
  • the memory 909 can be used to store software programs or instructions and various data.
  • the memory 909 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instruction required for at least one function (such as a sound playback function, an image playback function, etc.), etc.
  • the memory 909 may include a volatile memory or a non-volatile memory.
  • the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.
  • the volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM).
  • RAM random access memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • SDRAM synchronous dynamic random access memory
  • DDRSDRAM double data rate synchronous dynamic random access memory
  • ESDRAM enhanced synchronous dynamic random access memory
  • SLDRAM synchronous link dynamic random access memory
  • DRRAM direct memory bus random access memory
  • the processor 910 may include one or more processing units; optionally, the processor 910 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 910.
  • the radio frequency unit 901 is used to obtain a first data set
  • Processor 910 is used to train a target AI unit based on the first data set, wherein the target AI unit is used to detect the performance of the terminal in reporting target channel information to the network side device, the target channel information is obtained based on a first processing of the first channel information by the first AI unit of the terminal and a second processing of the first channel feature information by the second AI unit of the network side device, the first channel feature information is information obtained by the first processing of the first channel information by the first AI unit of the terminal, and the target AI unit is related to the second AI unit.
  • the RF unit 901 is further configured to acquire a second data set, wherein the second data set includes a first two-tuple corresponding to the sample data one by one, and the first two-tuple includes: input information of the first AI unit and output information of the first AI unit;
  • Processor 910 is further configured to train the first AI unit based on the second data set.
  • the processor 910 is further used to determine first information based on the target AI unit, wherein the first information is related to the performance of the terminal reporting target channel information to the network side device.
  • the determining of the first information based on the target AI unit performed by the processor 910 includes:
  • first information is determined.
  • the first data set includes at least one of the following:
  • a second two-tuple corresponding to the sample data one by one comprising: input information of the second AI unit and output information of the second AI unit;
  • a triplet corresponding to the sample data one by one including: input information of the first AI unit, output information of the first AI unit, and output information of the second AI unit, or the triplet including: input information of the first AI unit, input information of the second AI unit, and output information of the second AI unit;
  • a four-tuple corresponding to the sample data one by one comprising: input information of the first AI unit, output information of the first AI unit, input information of the second AI unit, and output information of the second AI unit.
  • the acquiring of the first data set performed by the radio frequency unit 901 includes:
  • Second information is received from a first device, where the second information includes a first data set, or the second information includes an identifier of the first data set, and the terminal learns an association relationship between the first data set and the identifier of the first data set.
  • the processor 910 is further configured to determine, based on a first association relationship, that the target AI unit is used to determine first information related to the first AI unit, wherein the first association relationship indicates an association relationship between the target AI unit and the first AI unit.
  • the first association relationship includes at least one of the following:
  • the target AI unit is used to train at least one of the first AI units, and the target AI unit is used to determine first information related to a first AI unit trained based on the second data set;
  • the radio frequency unit 901 is further configured to receive first indication information, where the first indication information indicates the first association relationship; or,
  • the processor 910 is further configured to associate a target AI unit trained based on the first data with a first AI unit trained based on the second data, wherein the first data and the second data correspond to the same sample data.
  • the second data set is used to train M first AI units
  • the first data set is used to train N target AI units
  • the first association relationship includes an association relationship between the N target AI units and the M first AI units, where N and M are positive integers respectively.
  • N is greater than M, and one of the N target AI units is associated with at least one first AI unit of the M first AI units;
  • N is less than M, at least one of the N target AI units is associated with one of the M first AI units;
  • N is equal to M, and the N target AI units are associated with the M first AI units in a one-to-one correspondence.
  • the radio frequency unit 901 is further configured to send target capability information, wherein the target capability information indicates at least one of the following:
  • the terminal has the ability to train the target AI unit within the framework of independently training the first AI unit;
  • the terminal uses the target AI unit to detect the performance of the terminal reporting target channel information to the network side device;
  • the processor 910 is further configured to, when the terminal supports independent training of the first AI unit, confirm support for training the target AI unit;
  • the processor 910 is further configured to, when the terminal only supports AI unit transmission, confirm that training of the target AI unit is not supported.
  • the RF unit 901 is also used to send target status information related to the target AI unit.
  • the target state information indicates at least one of the following:
  • each second data set has an associated target AI unit
  • each first AI unit has an associated target AI unit
  • each first AI unit requires a new target AI unit.
  • the radio frequency unit 901 is further used to send second indication information, where the second indication information indicates first identification information of a target AI unit required by the terminal, and the first identification information is used to identify a second data set or a first AI unit.
  • the RF unit 901 is further configured to send target request information, wherein the target request information is used to request training of a target AI unit associated with the first AI unit;
  • the obtaining of the first data set performed by the radio frequency unit 901 includes:
  • a first data set corresponding to the first AI unit is received.
  • the RF unit 901 is further used to receive third indication information, wherein the third indication information is used to instruct the terminal to train a target AI unit associated with the first AI unit, or to train a target AI unit associated with a second data set, and the second data set is used to train at least one of the first AI units.
  • the radio frequency unit 901 is further configured to send the first information to a network side device.
  • the radio frequency unit 901 is further configured to receive fourth indication information, where the fourth indication information is used to indicate at least one of the following:
  • the terminal determines the first information based on the target AI unit
  • the terminal sends the first information to the network side device
  • the embodiment of the present application also provides a network side device, including a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps of the method embodiment shown in Figure 5.
  • the network side device embodiment corresponds to the above-mentioned network side device method embodiment, and each implementation process and implementation method of the above-mentioned method embodiment can be applied to the network side device embodiment, and can achieve the same technical effect.
  • the embodiment of the present application also provides a network side device.
  • the network side device 1000 includes: an antenna 1001, a radio frequency device 1002, a baseband device 1003, a processor 1004 and a memory 1005.
  • the antenna 1001 is connected to the radio frequency device 1002.
  • the radio frequency device 1002 receives information through the antenna 1001 and sends the received information to the baseband device 1003 for processing.
  • the baseband device 1003 processes the information to be sent and sends it to the radio frequency device 1002.
  • the radio frequency device 1002 processes the received information and sends it out through the antenna 1001.
  • the method executed by the network-side device in the above embodiment may be implemented in the baseband device 1003, which includes a baseband processor.
  • the baseband device 1003 may include, for example, at least one baseband board, on which a plurality of chips are arranged, as shown in FIG10 , wherein one of the chips is, for example, a baseband processor, which is connected to the memory 1005 through a bus interface to call a program in the memory 1005 and execute the network device operations shown in the above method embodiment.
  • the network side device may also include a network interface 1006, which is, for example, a Common Public Radio Interface (CPRI).
  • CPRI Common Public Radio Interface
  • the network side device 1000 of the embodiment of the present application also includes: instructions or programs stored in the memory 1005 and executable on the processor 1004.
  • the processor 1004 calls the instructions or programs in the memory 1005 to execute the method executed by each module shown in Figure 7 and achieves the same technical effect. To avoid repetition, it will not be repeated here.
  • the embodiment of the present application further provides a network side device.
  • the network side device 1100 includes: a processor 1101, a network interface 1102, and a memory 1103.
  • the network interface 1102 is, for example, a Common Public Radio Interface (CPRI).
  • CPRI Common Public Radio Interface
  • the network side device 1100 of the embodiment of the present application also includes: instructions or programs stored in the memory 1103 and executable on the processor 1101.
  • the processor 1101 calls the instructions or programs in the memory 1103 to execute the method executed by each module as shown in Figure 7 and achieves the same technical effect. To avoid repetition, it will not be repeated here.
  • An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored.
  • a program or instruction is stored.
  • the various processes of the method embodiment shown in Figure 4 or Figure 5 are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
  • the processor is the processor in the terminal described in the above embodiment.
  • the readable storage medium includes Computer-readable storage media, such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disk, etc. In some examples, the readable storage medium may be a non-transitory readable storage medium.
  • An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the method embodiment shown in Figure 4 or Figure 5, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.
  • the embodiments of the present application further provide a computer program/program product, which is stored in a storage medium, and is executed by at least one processor to implement the various processes of the method embodiments shown in Figures 4 or 5, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • An embodiment of the present application also provides a communication system, including: a terminal and a first device, wherein the terminal can be used to execute the steps of the information processing method shown in FIG. 4 , and the first device can be used to execute the steps of the information processing method shown in FIG. 5 .

Landscapes

  • Mobile Radio Communication Systems (AREA)

Abstract

本申请公开了一种信息处理方法、信息处理装置、终端及网络侧设备,属于通信技术领域,本申请实施例的信息处理方法包括:终端获取第一数据集;所述终端基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。

Description

信息处理方法、信息处理装置、终端及网络侧设备
相关申请的交叉引用
本申请主张在2023年04月20日提交的中国专利申请No.202310431601.7的优先权,其全部内容通过引用包含于此。
技术领域
本申请属于通信技术领域,具体涉及一种信息处理方法、信息处理装置、终端及网络侧设备。
背景技术
在相关技术中,终端可以采用编码人工智能(Artificial Intelligence,AI)模型对信道信息进行压缩,并上报压缩后的信道特征信息,网络侧则可以采用解码AI模型对终端上报的信道特征信息进行恢复处理,得到终端信道信息,这样,能够降低终端上报信道信息的开销。
但是,使用AI模型进行信道信息压缩的过程中,随着终端位置、信道环境等的变化,AI模型的性能会产生波动,可能无法匹配当前的信道状态,导致压缩-恢复的信道信息不准确,此时,采用该AI模型对信道信息进行压缩-恢复,造成网络侧获取的信道信息不准确,即信道信息上报的可靠性低。
发明内容
本申请实施例提供一种信息处理方法、信息处理装置、终端及网络侧设备,能够在基于AI模型对信道信息进行压缩-恢复的信道信息上报的过程中,在终端侧训练与解码AI模型相关的目标AI单元,并采用该目标AI单元对基于AI模型上报的信道信息的性能进行监督,以及时发现压缩-恢复的信道信息不准确的问题。
第一方面,提供了一种信息处理方法,该方法包括:
终端获取第一数据集;
所述终端基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
第二方面,提供了一种信息处理方法,该方法包括:
第一设备获取下行信道状态信息;
所述第一设备基于所述下行信道状态信息训练第一AI单元和第二AI单元;
所述第一设备向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向所述网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
第三方面,提供了一种信息处理装置,包括:
第一获取模块,用于获取第一数据集;
第一训练模块,用于基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
第四方面,提供了一种信息处理装置,包括:
第三获取模块,用于获取下行信道状态信息;
第三训练模块,用于基于所述下行信道状态信息训练第一AI单元和第二AI单元;
第一发送模块,用于向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向所述网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
第五方面,提供了一种终端,该终端包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤。
第六方面,提供了一种终端,包括处理器及通信接口,其中,所述通信接口用于获取第一数据集;所述处理器用于基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所 述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
第七方面,提供了一种网络侧设备,该网络侧设备包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤。
第八方面,提供了一种网络侧设备,包括处理器及通信接口,所述通信接口用于获取下行信道状态信息;所述处理器用于基于所述下行信道状态信息训练第一AI单元和第二AI单元;所述通信接口还用于向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向所述网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
第九方面,提供了一种可读存储介质,所述可读存储介质上存储程序或指令,所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤,或者实现如第二方面所述的方法的步骤。
第十方面,提供了一种无线通信系统,包括:终端和第一设备,所述终端可用于执行如第一方面所述的方法的步骤,所述第一设备可用于执行如第二方面所述的方法的步骤。
第十一方面,提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如第一方面所述的方法,或实现如第二方面所述的方法。
第十二方面,提供了一种计算机程序/程序产品,所述计算机程序/程序产品被存储在存储介质中,所述程序/程序产品被至少一个处理器执行以实现如第一方面所述的信息处理方法的步骤,或者实现如第二方面所述的信息处理方法的步骤。
在本申请实施例中,在基于AI的信道信息上报场景下,在终端侧训练目标AI单元,且该目标AI单元与网络侧使用的第二AI单元相关,这样,可以利用该AI单元来检测所述终端向网络侧设备上报目标信道信息的性能,以及时发现基于AI的压缩-恢复的信道信息不准确的问题。
附图说明
图1是本申请实施例能够应用的无线通信系统的结构示意图;
图2是神经网络模型的架构示意图;
图3是神经元的示意图;
图4是本申请实施例提供的一种信息处理方法的流程图
图5是本申请实施例提供的另一种信息处理方法的流程图;
图6是本申请实施例提供的一种信息处理装置的结构示意图;
图7是本申请实施例提供的另一种信息处理装置的结构示意图;
图8是本申请实施例提供的一种通信设备的结构示意图;
图9是本申请实施例提供的一种终端的结构示意图;
图10是本申请实施例提供的一种网络侧设备的结构示意图;
图11是本申请实施例提供的另一种网络侧设备的结构示意图。
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清楚描述,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所获得的所有其他实施例,都属于本申请保护的范围。
本申请的术语“第一”、“第二”等是用于区别类似的对象,而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换,以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施,且“第一”、“第二”所区别的对象通常为一类,并不限定对象的个数,例如第一对象可以是一个,也可以是多个。此外,本申请中的“或”表示所连接对象的至少其中之一。例如“A或B”涵盖三种方案,即,方案一:包括A且不包括B;方案二:包括B且不包括A;方案三:既包括A又包括B。字符“/”一般表示前后关联对象是一种“或”的关系。
本申请的术语“指示”既可以是一个直接的指示(或者说显式的指示),也可以是一个间接的指示(或者说隐含的指示)。其中,直接的指示可以理解为,发送方在发送的指示中明确告知了接收方具体的信息、需要执行的操作或请求结果等内容;间接的指示可以理解为,接收方根据发送方发送的指示确定对应的信息,或者进行判断并根据判断结果确定需要执行的操作或请求结果等。
值得指出的是,本申请实施例所描述的技术不限于长期演进型(Long Term Evolution,LTE)/LTE的演进(LTE-Advanced,LTE-A)系统,还可用于其他无线通信系统,诸如码分多址(Code Division Multiple Access,CDMA)、时分多址(Time Division Multiple Access,TDMA)、频分多址(Frequency Division Multiple Access,FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access,OFDMA)、单载波频分多址(Single-carrier Frequency-Division Multiple Access,SC-FDMA)或其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用,所描述的技术既可用于以上提及的系统和无线电技术,也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio,NR)系统, 并且在以下大部分描述中使用NR术语,但是这些技术也可应用于NR系统以外的系统,如第6代(6th Generation,6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的框图。无线通信系统包括终端11和网络侧设备12。其中,终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)、笔记本电脑、个人数字助理(Personal Digital Assistant,PDA)、掌上电脑、上网本、超级移动个人计算机(Ultra-mobile Personal Computer,UMPC)、移动上网装置(Mobile Internet Device,MID)、增强现实(Augmented Reality,AR)、虚拟现实(Virtual Reality,VR)设备、机器人、可穿戴式设备(Wearable Device)、飞行器(flight vehicle)、车载设备(Vehicle User Equipment,VUE)、船载设备、行人终端(Pedestrian User Equipment,PUE)、智能家居(具有无线通信功能的家居设备,如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(Personal Computer,PC)、柜员机或者自助机等终端侧设备。可穿戴式设备包括:智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。其中,车载设备也可以称为车载终端、车载控制器、车载模块、车载部件、车载芯片或车载单元等。需要说明的是,在本申请实施例并不限定终端11的具体类型。网络侧设备12可以包括接入网设备或核心网设备,其中,接入网设备也可以称为无线接入网(Radio Access Network,RAN)设备、无线接入网功能或无线接入网单元。接入网设备可以包括基站、无线局域网(Wireless Local Area Network,WLAN)接入点(Access Point,AS)或无线保真(Wireless Fidelity,WiFi)节点等。其中,基站可被称为节点B(Node B,NB)、演进节点B(Evolved Node B,eNB)、下一代节点B(the next generation Node B,gNB)、新空口节点B(New Radio Node B,NR Node B)、接入点、中继站(Relay Base Station,RBS)、服务基站(Serving Base Station,SBS)、基收发机站(Base Transceiver Station,BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set,BSS)、扩展服务集(Extended Service Set,ESS)、家用B节点(home Node B,HNB)、家用演进型B节点(home evolved Node B)、发送接收点(Transmission Reception Point,TRP)或所述领域中其他某个合适的术语,只要达到相同的技术效果,所述基站不限于特定技术词汇,需要说明的是,在本申请实施例中仅以NR系统中的基站为例进行介绍,并不限定基站的具体类型。
核心网设备可以包含核心网设备可以包含但不限于如下至少一项:核心网节点、核心网功能、定位管理功能(Location Management Function,LMF)、移动管理实体(Mobility Management Entity,MME)、接入移动管理功能(Access and Mobility Management Function,AMF)、会话管理功能(Session Management Function,SMF)、用户平面功能(User Plane Function,UPF)、策略控制功能(Policy Control Function,PCF)、策略与计费规则功能单元(Policy and Charging Rules Function,PCRF)、边缘应用服务发现功能(Edge Application Server Discovery Function,EASDF)、统一数据管理(Unified Data Management,UDM)、统一数据仓储(Unified Data Repository,UDR)、归属用户服务器(Home Subscriber Server, HSS)、集中式网络配置(Centralized network configuration,CNC)、网络存储功能(Network Repository Function,NRF)、网络开放功能(Network Exposure Function,NEF)、本地NEF(Local NEF,或L-NEF)、绑定支持功能(Binding Support Function,BSF)、应用功能(Application Function,AF)等。需要说明的是,在本申请实施例中仅以NR系统中的核心网设备为例进行介绍,并不限定核心网设备的具体类型。需要说明的是,在本申请实施例中仅以NR系统中的核心网设备为例进行介绍,并不限定核心网设备的具体类型。
在无线通信技术中,准确的信道状态信息(channel state information,CSI)反馈对信道容量至关重要。尤其是对于多天线系统来讲,发送端可以根据CSI优化信号的发送,使其更加匹配信道的状态。如:信道质量指示(Channel Quality Indicator,CQI)可以用来选择合适的调制编码方案(Modulation and Coding Scheme,MCS),以实现链路自适应;预编码矩阵指示(Precoding Matrix Indicator,PMI)可以用来实现特征波束成形(eigen beamforming),从而最大化接收信号的强度,或者用来抑制干扰(如小区间干扰、多用户之间干扰等)。因此,自从多天线技术(如:多输入多输出(Multi-Input Multi-Output,MIMO))被提出以来,CSI的获取一直都是研究热点。
通常,网络侧设备在某个时隙(slot)的某些时频资源上发送CSI参考信号(CSI-Reference Signals,CSI-RS),终端根据CSI-RS进行信道估计,计算这个slot上的信道信息,通过码本将PMI反馈给基站,网络侧设备根据终端反馈的码本信息组合出信道信息,并在终端下一次上报CSI之前,网络侧设备以此信道信息进行数据预编码及多用户调度。
为了进一步减少CSI反馈开销,终端可以将每个子带上报PMI改成按照时延域(delay域,即频域)上报PMI,由于delay域的信道更集中,用更少的delay的PMI就可以近似表示全部子带的PMI,其可以视作是将delay域信息压缩之后再上报。
同样,为了减少开销,网络侧设备可以事先对CSI-RS进行预编码,将编码后的CSI-RS发送给终端,终端看到的是经过编码之后的CSI-RS对应的信道,终端只需要在网络侧设备指示的端口中选择若干个强度较大的端口,并上报这些端口对应的系数即可。
在相关技术中,利用AI单元对信道信息进行压缩,能够提升信道特征信息的压缩效果。具体的,终端可以估计CSI参考信号(CSI Reference Signal,CSI-RS)或跟踪参考信号(Tracking Reference Signal,TRS),根据该估计到的信道信息进行计算,得到计算的信道信息,然后,将计算的信道信息或者原始的估计到的信道信息通过编码器进行编码,得到编码结果,最后将编码结果发送给基站。在基站侧,基站可以在接收编码后的结果后,将其输入到解码器中,利用该解码器恢复信道信息。具体的,基于神经网络的CSI压缩反馈方案是,在终端利用编码网络对信道信息进行压缩编码,将压缩后的内容发送给基站,在基站利用解码网络对压缩后的内容进行解码,从而恢复信道信息,此时基站的解码网络和终端的编码网络需要联合训练,达到合理的匹配度。编码网络的输入是信道信息,输出是编码信息,即信道特征信息,解码网络的输入是编码信息,输出是恢复的信道信息。
但是,使用AI单元进行信道信息的压缩-解压缩过程中,鉴于终端位置变化、信道环 境的变化等原因,AI单元的性能会产生波动,甚至无法匹配终端当前的信道状态,导致使用AI单元进行压缩-解压缩后的信道信息不准确。
本申请实施例中,通过在终端侧训练目标AI单元,且该目标AI单元与网络侧使用的第二AI单元(如解码器)相关,这样,可以利用该目标AI单元来检测所述终端向网络侧设备上报目标信道信息的性能,如:利用目标AI单元对第一AI单元输出的第一信道特征信息进行解码或恢复处理,并将解码或恢复后的信道信息与第一信道信息进行比较,两者越相关则表示所述终端向网络侧设备上报目标信道信息的性能越好,从而能够及时发现基于AI的压缩-恢复的信道信息不准确的问题。
值得注意的是,对AI单元的主要评价指标是输入编码器的信道信息和解码器恢复的信道信息的相关性,如果二者完全相同说明AI单元实现了完美压缩和解压缩,通常相关性损失在一定程度内可以认为AI单元是有效的,在相关技术中,为了实现对AI单元的检测,需要在单侧(网络侧或终端侧)同时获取编码器的第一AI单元和解码器的第二AI单元,这样,才能够利用第二AI单元对第一AI单元压缩后的信道特征信息进行恢复,并将恢复的信道信息与输入第一AI单元的信道信息进行比较,并根据两者的相关性来判断第一AI单元和第二AI单元的性能。
例如:假设以终端侧同时获取编码器的第一AI单元和解码器的第二AI单元为例,网络侧将使用的第二AI单元传递至终端,以使终端利用第二AI单元来检测所述终端向网络侧设备上报目标信道信息的性能。但是,该方法受到多方面的限制,如:出于模型私有性、兼容性等方面的考虑,有些模型训练协作方式(type2或者type3)不支持在某一端获取完整模型问题,终端无法同时获取第一AI单元和第二AI单元,从而使得该方案无法实施。此外,即使支持模型在不同节点之间的传递,但是由于可能存在实际使用的模型参数规模较大的情况(尤其是针对需要传递解码器的情况),如果需要向对端传递则会引入不小的开销。
而本申请实施例中,只需要传递用于训练目标AI单元的数据集,其传输开销较小,且满足网络侧对第二AI单元的模型私有化要求。
需要说明的是,本申请实施例中的AI单元也可称为AI模型、AI结构等,或者所述AI单元也可以是指能够实现与AI相关的特定的算法、公式、处理流程、能力等的处理单元,或者所述AI单元可以是针对特定数据集的处理方法、算法、功能、模块或单元,或者所述AI单元可以是运行在图形处理器(Graphics Processing Unit,GPU)、网络处理器(Neural-Network Processing Unit,NPU)、张量处理器(Tensor Processing Unit,TPU)、专用集成电路(Application Specific Integrated Circuit,ASIC)等AI相关硬件上的处理方法、算法、功能、模块或单元,本申请对此不做具体限定。可选地,所述特定数据集包括AI单元的输入和或输出。
可选地,所述AI单元的标识,可以是AI模型标识、AI结构标识、AI算法标识,或者所述AI单元关联的特定数据集的标识,或者所述AI相关的特定场景、环境、信道特 征、设备的标识,或者所述AI相关的功能、特性、能力或模块的标识,本申请对此不做具体限定。
另外,AI功能有多种实现方式,例如神经网络、决策树、支持向量机、贝叶斯分类器等,本申请对此不做具体限定。为了便于说明,本申请实施例中以AI单元为神经网络为例进行说明,但是并不限定AI单元的具体类型。例如:如图2所示,神经网络模型包括输入层、隐层和输出层,其可以根据输入层获取的出入信息(X1~Xn)预测可能的输出结果(Y)。神经网络模型由大量的神经元组成,如图3所示,神经元的参数包括:输入参数a1~aK、权值w、偏置b以及激活函数σ(z),以及与这些参数获取输出值a,其中,常见的激活函数包括S型生长曲线(Sigmoid)函数、双曲正切(tanh)函数、线性整流函数(Rectified Linear Unit,ReLU,其也称之为修正线性单元)函数等等,且上述函数σ(z)中的z可以通过以下公式计算得到:
z=a1w1+…+akwk+aKwK+b
其中,k表示输入参数的总数。
神经网络的参数通过优化算法进行优化。优化算法就是一种能够帮我们最小化或者最大化目标函数(有时候也叫损失函数)的一类算法。而目标函数往往是模型参数和数据的数学组合。例如给定数据X和其对应的标签Y,我们构建一个神经网络模型f(.),有了模神经网络型后,根据输入x就可以得到预测输出f(x),并且可以计算出预测值和真实值之间的差距(f(x)-Y),这个就是损失函数。我们的目的是找到合适的W和b,使上述的损失函数的值达到最小,损失值越小,则说明我们的模型越接近于真实情况。
目前常见的优化算法,基本都是基于误差反向传播(error Back Propagation,BP)算法。BP算法的基本思想是,学习过程由信号的正向传播与误差的反向传播两个过程组成。正向传播时,输入样本从输入层传入,经各隐层逐层处理后,传向输出层。若输出层的实际输出与期望的输出不符,则转入误差的反向传播阶段。误差反传是将输出误差以某种形式通过隐层向输入层逐层反传,并将误差分摊给各层的所有单元,从而获得各层单元的误差信号,此误差信号即作为修正各单元权值的依据。这种信号正向传播与误差反向传播的各层权值调整过程,是周而复始地进行的。权值不断调整的过程,也就是网络的学习训练过程。此过程一直进行到网络输出的误差减少到可接受的程度,或进行到预先设定的学习次数为止。
常见的优化算法有梯度下降(Gradient Descent,GD)、随机梯度下降(Stochastic Gradient Descent,SGD)、小批量梯度下降(mini-batch gradient descent)、动量法(Momentum)、Nesterov(其表示带动量的随机梯度下降)、自适应梯度下降(Adaptive gradient descent,Adagrad)、自适应学习率调整(Adadelta)、均方根误差降速(root mean square prop,RMSprop)、自适应动量估计(Adaptive Moment Estimation,Adam)等。
这些优化算法在误差反向传播时,都是根据损失函数得到的误差/损失,对当前神经元求导数/偏导,加上学习速率、之前的梯度/导数/偏导等影响,得到梯度,将梯度传给上一 层。
本申请实施例中的AI单元包括:
第一AI单元,即编码器对应的AI单元,其部署在终端侧,其输入信息为信道信息,输出信息是信道特征信息;
第二AI单元,即解码器对应的AI单元,其部署在网络侧,其输入信息为信道特征信息,输出信息是恢复的信道信息;
目标AI单元,即代理AI单元,用于模拟第二AI单元的功能。
其中,第一AI单元和第二AI单元可以是独立训练的,如终端训练第一AI单元,网络侧训练第二AI单元。
值得注意的是,目标AI单元需要匹配网络侧实际使用的第二AI单元的性能,则目标AI单元需要在有第二AI单元的一端(即网络侧)进行训练,但是,对于第一AI单元和第二AI单元独立训练的场景,网络侧不会向终端传递AI单元,本申请实施例中是由终端训练目标AI单元。
下面结合附图,通过一些实施例及其应用场景对本申请实施例提供的信息处理方法、信息处理装置以及相关设备进行详细地说明。
请参阅图4,本申请实施例提供的一种信息处理方法,其执行主体可以是终端,其中,终端可以是如图1中列举的各种类型的终端11,或者是除了如图1所示实施例中列举的终端类型之外的其他终端,在此不作具体限定。
如图4所示,该信息处理方法可以包括以下步骤:
步骤401、终端获取第一数据集。
一种实施方式中,第一数据集可以是从网络侧设备接收的,如终端接收来自基站或核心网设备的第一数据集。
另一种实施方式中,第一数据集可以是从第三方节点接收的,如终端接收来自指定服务器的第一数据集。
步骤402、所述终端基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
一种实施方式中,所述目标AI单元与所述第二AI单元相关,可以是:基于至少部分相同的样本数据来训练目标AI单元与第二AI单元,例如:网络侧基于10万个样本数据来联合训练第一AI单元和第二AI单元,终端基于其中的1万个样本数据来训练目标AI单元,这样,基于至少部分相同样本数据所训练的目标AI单元具有与第二AI单元相同或相似的功能。
另一种实施方式中,所述目标AI单元与所述第二AI单元相关,可以是:第一数据集可以包括网络侧设备的第二AI单元的输入信息和输出信息,这样,基于第二AI单元的输入信息和输出信息所训练的目标AI单元具有与第二AI单元相同或相似的功能。
可选地,目标AI单元的模型结构比第二AI单元的模型结构简单,或目标AI单元的参数规模比第二AI单元的参数规模小。
值得提出的是,本申请实施例中的目标AI单元可以模拟第二AI单元的功能,这样,目标AI单元的性能与第二AI单元的性能可以呈一定比例的缩放的关系,从而支持使用目标AI单元恢复的第二信道信息来模拟使用第二AI单元恢复的目标信道信息,其中,对于第二信道信息与输入第一AI单元的第一信道信息之间的相关性,以及目标信道信息与输入第一AI单元的第一信道信息之间的相关性,两者是相关的或可以线性转换的,也就是说,第二信道信息与输入第一AI单元的第一信道信息之间的相关性越高,则可以反映目标信道信息与输入第一AI单元的第一信道信息之间的相关性越高,即所述终端向网络侧设备上报目标信道信息的性能越好。
一种实施方式中,网络侧设备可以从终端或其他节点接收下行信道状态信息,并基于该下行信道状态信息联合训练第一AI单元和第二AI单元,并将该训练过程中,第二AI单元的至少部分输入信息和输出信息作为第一数据集下发给终端,终端则可以基于第一数据集训练目标AI单元。
可选地,可以在协议中约定所述目标AI单元的模型结构,所述终端基于所述第一数据集训练目标AI单元,可以是终端基于所述第一数据集训练目标AI单元的权重参数。
作为一种可选的实施方式,所述方法还包括:
所述终端基于所述目标AI单元确定第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
可选地,所述终端基于所述目标AI单元确定第一信息,包括:
所述终端基于所述第一AI单元对所述第一信道信息进行所述第一处理,得到所述第一信道特征信息;
所述终端基于所述目标AI单元对所述第一信道特征信息进行第三处理,得到第二信道信息;
所述终端基于所述第二信道信息和所述第一信道信息的相关性,确定第一信息。
其中,第一处理可以是压缩处理、量化处理、归一化处理中的至少一项。第三处理与网络侧设备的第二AI单元的恢复处理相似,例如:第三处理可以包括解压缩处理、解量化处理、解归一化处理中的至少一项。
本实施方式中,利用目标AI单元模拟网络侧设备的第二AI单元对终端上报的第一信道特征信息进行恢复处理,得到的第二信道信息与网络侧设备基于第二AI单元对终端上报的第一信道特征信息进行恢复处理的目标信道信息相同或相似,这样,基于所述第二信道信息和所述第一信道信息的相关性,可以反映所述目标信道信息和所述第一信道信息的 相关性,从而据此确定终端向网络侧设备上报目标信道信息的性能,即所述第二信道信息和所述第一信道信息的相关性越高,则终端向网络侧设备上报目标信道信息的性能越好。
可选地,第一信息包括以下至少一项:
所述目标AI单元的输出信息,即目标AI单元恢复的信道信息,该目标AI单元恢复的信道信息与输入第一AI单元的信道信息的相关性越高,则所述终端向网络侧设备上报目标信道信息的性能越好;
所述目标AI单元的输出信息与所述第一AI单元的输入信息之间的相关性值,例如:信道矩阵的矩阵相似度或相似系数或欧式距离或曼哈顿距离或平方余弦相似度(Square generalized cosine similarity,SGCS)等相关性值;
所述终端向网络侧设备上报目标信道信息的性能值,其中,所述终端向网络侧设备上报目标信道信息的性能值,可以正相关于所述目标AI单元的输出信息与所述第一AI单元的输入信息之间的相关性值,例如:对所述目标AI单元的输出信息与所述第一AI单元的输入信息之间的相关性值进行量化或归一化等处理,得到所述终端向网络侧设备上报目标信道信息的性能值;
目标指示信息,所述目标指示信息用于指示所述第一AI单元或所述第二AI单元失效,其中,所述终端可以在确定所述终端向网络侧设备上报目标信道信息的性能值低于第一阈值的情况下,向网络侧设备上报目标指示信息。
可选地,所述第一阈值可以是协议中约定的或根据业务需求确定的或用户设置的或网络侧设备指示的。
可选地,所述第一阈值可以根据非AI的CSI上报方式确定,例如:将基于码本的CSI上报方式下,终端压缩前的信道信息与网络侧基于码本解压缩后的信道信息之间的相关性值作为第一阈值,若所述目标AI单元的输出信息与所述第一AI单元的输入信息之间的相关性值小于或等于第一阈值,则所述终端向网络侧设备上报目标指示信息。
本实施方式中,终端在训练得到目标AI单元之后,可以利用目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能。
一种实施方式中,终端在确定所述第一信息之后,还可以向网络侧设备发送所述第一信息。
这样,网络侧设备可以根据第一信息及时获知第一AI单元和第二AI单元是否失效,若确定第一AI单元和第二AI单元失效,则网络侧设备可以指示终端采用非AI的方式(如码本)上报信道信息,或者,更新或重新训练第一AI单元和第二AI单元,以提升终端向网络侧设备上报信道信息的性能。
作为一种可选的实施方式,所述方法还包括:
所述终端接收第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
所述终端基于所述目标AI单元确定所述第一信息;
所述第一信息的内容;
所述终端向网络侧设备发送所述第一信息;
所述终端发送所述第一信息的方式。
其中,第四指示信息可以来自网络侧设备。
一种实施方式中,第四指示信息能够指示终端执行“基于所述目标AI单元确定所述第一信息”的操作。
另一种实施方式中,第四指示信息能够指示终端上报的第一信息的内容,如:所述目标AI单元的输出信息、所述目标AI单元的输出信息与所述第一AI单元的输入信息之间的相关性值、所述终端向网络侧设备上报目标信道信息的性能值、目标指示信息等。
又一种实施方式中,第四指示信息能够指示终端是否向网络侧设备上报第一信息,例如:指示终端在确定所述第一AI单元或所述第二AI单元有效的情况下,不向网络侧设备上报第一信息;在确定所述第一AI单元或所述第二AI单元失效的情况下,向网络侧设备上报第一信息。
再一种实施方式中,第四指示信息能够指示终端上报的第一信息的方式,如上报第一信息使用的时域或频域资源、承载第一信息的信令等。
值得提出的是,在终端和网络侧设备独立训练第一AI单元和第二AI单元的情况下,为了使终端训练的第一AI单元与网络侧设备训练的第二AI单元达到合理的匹配度,终端与网络侧设备可以基于相同的样本数据分别训练第一AI单元和第二AI单元,或者,终端可以从网络侧设备或第三方节点接收第二数据集,其中,第二数据集包括:网络侧设备在联合训练第一AI单元和第二AI单元的过程中产生的,所述第一AI单元的输入信息和所述第一AI单元的输出信息。
一种实施方式中,第二数据集与第一数据集可以独立指示。例如:网络侧设备分别向终端发送第一数据集和第二数据集。
可选地,在独立指示第二数据集与第一数据集的情况下,第二数据集与第一数据集所包含的数据个数可以不同。例如:第二数据集中数据的个数大于第一数据集中数据的个数,这样,可以采用较大的第二数据集训练精确度更高的第一AI单元,采用较小的第一数据集训练模型结构较为简单的目标AI单元。
可选地,所述方法还包括:
所述终端获取第二数据集,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息;
所述终端基于所述第二数据集训练所述第一AI单元。
其中,样本数据可以是信道信息。例如:一个样本数据可以是在一个时域或频域采样点获取的信道信息。
与上述第一数据集相似的,第二数据集可以是从网络侧设备接收或者从第三方节点接收,如终端接收来自基站或核心网设备或指定服务器的第二数据集。
例如:网络侧设备基于大量的样本数据联合训练第一AI单元和第二AI单元,该训练过程中,对于每一个样本数据,会有以下训练过程:将样本数据输入第一AI单元,获取第一AI单元输出的信道特征信息,将第一AI单元输出的信道特征信息输入第二AI单元,获取第二AI单元输出的信道信息,此时,网络侧设备将第一AI单元的至少部分输入信息和输出信息作为第二数据集下发给终端,用于终端独立训练第一AI单元。
本实施方式中,终端独立训练第一AI单元,且终端从网络侧设备接收第二数据集,可以使基于该第二数据集训练得到的第一AI单元与网络侧设备训练的第二AI单元匹配。
另一种实施方式中,第二数据集与第一数据集可以联合指示,其中,鉴于第一数据集可以包括网络侧设备在联合训练第一AI单元和第二AI单元的过程中产生的,所述第二AI单元的输入信息和所述第二AI单元的输出信息,第二数据集包括:网络侧设备在联合训练第一AI单元和第二AI单元的过程中产生的,所述第一AI单元的输入信息和所述第一AI单元的输出信息。
可选地,鉴于第一AI单元的输出信息即为第二AI单元的输入信息,可以通过三元组联合指示第二数据集与第一数据集,其中,三元组可以包括:第一AI单元的输入信息、第二AI单元的输入信息、第二AI单元的输出信息;或者,三元组可以包括:第一AI单元的输入信息、第一AI单元的输出信息、第二AI单元的输出信息。
作为一种可选的实施方式,所述第一数据集包括以下至少一项:
与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
一种实施方式中,在第一数据集与第二数据集独立指示的情况下,第一数据集包括网络侧设备在联合训练第一AI单元和第二AI单元过程中产生的所述第二AI单元的输入信息和所述第二AI单元的输出信息。
另一种实施方式中,在第一数据集与第二数据集联合指示的情况下,第一数据集包括与样本数据一一对应的三元组,该三元组的前两位作为终端训练第一AI单元数据集,即所述第一AI单元的输入信息和所述第一AI单元的输出信息,或所述第一AI单元的输入信息和所述第二AI单元的输入信息用于终端训练第一AI单元;该三元组的后两位作为终端训练目标AI单元数据集,即所述第一AI单元的输出信息和所述第二AI单元的输出信息,或所述第二AI单元的输入信息和所述第二AI单元的输出信息用于终端训练目标AI单元。
再一种实施方式中,在第一数据集与第二数据集联合指示的情况下,第一数据集包括 与样本数据一一对应的四元组,其中的前两位用于终端训练第一AI单元,后两位用于终端训练目标AI单元。
可选地,上述四元组中的第一AI单元的输出信息和所述第二AI单元的输入信息可以不相同,例如:终端对于第一AI单元的输出信息使用特定的量化方案,并向网络侧上报的是量化后的信道特征信息,则第一AI单元的输出信息和所述第二AI单元的输入信息可以不相同。
作为一种可选的实施方式,所述终端获取第一数据集,包括:
所述终端接收来自第一设备的第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系。
一种实施方式中,第一设备可以是网络侧设备,该网络侧设备可以用于联合训练第一AI单元和第二AI单元,这样,网络侧设备将联合训练第一AI单元和第二AI单元过程中,第二AI单元的输入信息和输出信息作为训练数据发送给终端,以供终端训练目标AI单元。可选地,第一数据集中还可以包括网络侧设备将联合训练第一AI单元和第二AI单元过程中,第一AI单元的输入信息和输出信息,终端可以基于该信息独立训练第一AI单元。
另一种实施方式中,第一设备可以是第三方节点,例如:服务器。该第三方节点可以从网络侧设备获知第一数据集,或者,该第三方节点可以联合训练第一AI单元和第二AI单元。
一种实施方式中,终端可以从第一设备直接获取第一数据集。
另一种实施方式中,终端可以预先获取第一数据集与第一数据集的标识之间的关联关系,如:协议约定每一个第一数据集及其对应的标识,或者网络侧设备预配置每一个第一数据集及其对应的标识。这样,当第二信息包括所述第一数据集的标识的情况下,终端可以根据第一数据集与第一数据集的标识之间的关联关系,确定第一数据集。
作为一种可选的实施方式,终端可能具有至少两个第一AI单元,此时,所述方法还包括:
所述终端基于第一关联关系,确定所述目标AI单元用于确定所述第一AI单元相关的第一信息,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
其中,目标AI单元用于检测使用关联的第一AI单元处理的信道信息的性能。例如:假设目标AI单元与第一AI单元A关联,第一AI单元A与第二AI单元A匹配,则目标AI单元可以用于检测第一AI单元A的输入信息与第二AI单元A的输出信息之间的相关性,从而确定使用第一AI单元A和第二AI单元A上报给网络侧的信道信息的性能。
可选地,所述第一关联关系包括以下至少一项:
所述目标AI单元与至少一个所述第一AI单元的关联关系;
所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
一种实施方式中,目标AI单元可以与第一AI单元关联,或者,鉴于第一AI单元与对应的第二AI单元匹配,目标AI单元也可以与第二AI单元关联。
另一种实施方式中,目标AI单元可以与第二数据集或第二数据集ID关联,其中,一个第二数据集可以用于训练一个或者至少两个第一AI单元,则隐含了目标AI单元与基于同一个第二数据集训练的一个或至少两个第一AI单元关联。
可选地,在目标AI单元可以与第二数据集ID关联的情况下,可以采用协议约定或网络侧预配置的方式,使终端获知第二数据集ID对应的第二数据集。
再一种实施方式中,第一数据集与第一AI单元或第一AI单元ID关联,例如:当第一数据集仅包含解码器的输入输出时,需要额外告知第一数据集关联哪一些第一AI单元,即第一数据集与哪一个编码器的输入输出数据集对应。其中,一个第一数据集可以用于训练至少一个目标AI单元,此时,隐含了基于同一个第一数据集训练的一个或至少两个目标AI单元与该第一数据集对应的第一AI单元关联。
又一种实施方式中,第一数据集与第二AI单元或第二AI单元ID关联,其中,鉴于第一AI单元与对应的第二AI单元匹配,一个第一数据集可以用于训练至少一个目标AI单元,此时,隐含了基于同一个第一数据集训练的一个或至少两个目标AI单元与该第一数据集对应的第二AI单元关联。
还有一种实施方式中,第一数据集与所述第二数据集或所述第二数据集ID关联,此时,基于第一数据集训练的目标AI单元,可以用于检测基于关联的第二数据集训练的第一AI单元的性能。
可选地,在所述第一数据集与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
进一步地,N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI 单元中的一个第一AI单元关联;
N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
上述第一关联关系可以是网络侧设备或第三方节点指示的,或者是终端确定的。
可选地,所述方法还包括以下至少一项:
所述终端接收第一指示信息,所述第一指示信息指示所述第一关联关系;
所述终端将基于第一数据训练的目标AI单元与基于第二数据训练的第一AI单元关联,其中,所述第一数据和所述第二数据与相同的样本数据对应。
一种实施方式中,第一指示信息来自以下至少一项:
所述第一数据集的发送节点;
网络侧设备,如基站或核心网设备,例如:第三方节点负责传输第一数据集和第二数据集中的至少一项,基站或核心网设备控制第一关联关系。
可选地,上述第一指示信息可以通过数据集ID进行指示,例如:第一关联关系包括第一数据集ID与第一AI单元ID之间的关联关系。
可选地,上述第一指示信息可以通过下行控制信息(Downlink Control Information,DCI)、媒体接入控制层控制单元(Medium Access Control Control Element,MAC CE)、无线资源控制(Radio Resource Control,RRC)信令中的至少一项指示;或者,
第一指示信息可以携带于AI单元的相关信息中,例如:在传输或配置第二数据集的过程中,将第二数据集与对应的第一数据集关联。
一种实施方式中,终端可以自主决定将与相同的样本数据对应的第一数据和第二数据关联,即将基于相互关联的第一数据训练的目标AI单元和基于第二数据训练的第一AI单元相关联。其中,第一数据表是用于训练目标AI单元的数据,具体可以包括第二AI单元的输入和输出;第二数据表是用于训练第一AI单元的数据,具体可以包括第一AI单元的输入和输出。
例如:当第一数据集为三元组时,终端可以自行关联后两位训练的目标AI单元与前两位训练的第一AI单元。
作为一种可选的实施方式,所述方法还包括:
所述终端发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
所述终端是否支持训练所述目标AI单元。
一种实施方式中,在终端向网络侧设备上报的目标能力信息指示:所述终端具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力的情况下,网络侧设备才会向终端发送第一数据集。
一种实施方式中,在终端向网络侧设备上报的目标能力信息指示:所述终端使用所述 目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能的情况下,网络侧设备可以配置或指示终端上报第一信息。
一种实施方式中,在终端向网络侧设备上报的目标能力信息指示:所述终端支持训练所述目标AI单元的情况下,网络侧设备才会向终端发送第一数据集。
本实施方式中,通过终端能力上报,可以在终端支持的前提下,时终端训练目标AI单元,或基于目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能,避免了因终端不支持而造成不必要的算力和空口资源浪费。
作为一种可选的实施方式,所述方法还包括:
在所述终端支持独立训练第一AI单元的情况下,所述终端确认支持训练所述目标AI单元;
在所述终端仅支持AI单元传递的情况下,所述终端确认不支持训练所述目标AI单元。
本实施方式中,可以在终端支持独立训练第一AI单元的情况下,默认终端支持训练所述目标AI单元;在终端不支持独立训练第一AI单元,仅支持AI单元传递(即网络侧联合训练第一AI单元和第二AI单元,并向终端传递第一AI单元或传递第一AI单元和第二AI单元)的情况下,默认终端不支持训练所述目标AI单元。这样,终端可以直接根据是否支持独立训练的能力来确定终端是否支持训练所述目标AI单元。
作为一种可选的实施方式,所述方法还包括:
所述终端发送所述目标AI单元相关的目标状态信息。
其中,目标状态信息可以用于供网络侧设备判断终端是否需要训练新的目标AI单元,从而据此向终端下发对应的第一数据集。
一种实施方式中,所述目标状态信息指示以下至少一项:
每个第二数据集是否有关联的目标AI单元;
每个第一AI单元是否有关联的目标AI单元;
每个第二数据集是否需要新的目标AI单元;
每个第一AI单元是否需要新的目标AI单元。
一种实施方式中,在每个第二数据集有关联的目标AI单元的情况下,可以采用目标AI单元对基于各自关联的第二数据集训练的第一AI单元进行性能检测。当某些第二数据集没有关联的目标AI单元的情况下,网络侧设备可以据此向终端发送这些第二数据集关联的第一数据集,以使终端据此训练与这些第二数据集关联的目标AI单元。
另一种实施方式中,在每个第一AI单元有关联的目标AI单元的情况下,可以采用目标AI单元对基于各自关联的第一AI单元进行性能检测。当某些第一AI单元没有关联的目标AI单元的情况下,网络侧设备可以据此向终端发送这些第一AI单元关联的第一数据集,以使终端据此训练与这些第一AI单元关联的目标AI单元。
再一种实施方式中,终端可以直接确定哪一些第二数据集需要新的目标AI单元,例 如:某些第二数据集没有关联的目标AI单元,或者由于终端位置移动、通信环境变化等使得某些第二数据集关联的目标AI单元失效。这样,网络侧设备可以据此向终端发送这些第二数据集关联的第一数据集,以使终端据此训练与这些第二数据集关联的目标AI单元。
又一种实施方式中,终端可以直接确定哪一些第一AI单元需要新的目标AI单元,例如:某些第一AI单元没有关联的目标AI单元,或者由于终端位置移动、通信环境变化等使得某些第一AI单元关联的目标AI单元失效。这样,网络侧设备可以据此向终端发送这些第一AI单元关联的第一数据集,以使终端据此训练与这些第一AI单元关联的目标AI单元。
可选地,终端可以周期性地上报目标状态信息。
本实施方式中,终端通过向网络侧设备或第三方节点发送所述目标AI单元相关的目标状态信息,可以使网络侧设备或第三方节点及时获知终端需要训练或更新哪些第二数据集或第一AI单元关联的目标AI单元,从而可以据此向终端下发对应的第一数据集。
作为一种可选的实施方式,所述方法还包括:
所述终端发送第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
一种实施方式中,网络侧设备可以基于第二指示信息确定哪些第二数据集或第一AI单元可以使用目标AI单元,从而下发对应的第一数据集,以使终端训练对应的目标AI单元。
另一种实施方式中,网络侧设备可以基于第二指示信息确定哪些第二数据集或第一AI单元具有关联的目标AI单元,从而可以指示终端使用对应的目标AI单元对终端上报目标信道信息进行性能检测。
值得提出的是,在基于AI单元的CSI反馈中,终端可能会有多个第二数据集,对应每个第二数据集终端会训练一个编码器,在某个环境下,使用对应第一AI单元压缩CSI。但不是所有的第二数据集都可以有目标AI单元,例如:有的信道环境复杂,简单的目标AI单元不足以解码;或者,有的第二数据集可能缺少对应的目标AI单元的数据集,导致无法训练目标AI单元;又或者,有些目标AI单元训练难度大,终端无法实现。因此对于支持目标AI单元的终端,也不是所有第二数据集都可以训练目标AI单元,本实施方式中,终端可以直接上报可以使用目标AI单元检测的第二数据集或第一AI单元,基站指示哪些第二数据集或第一AI单元使用目标AI单元。
作为一种可选的实施方式,所述方法还包括:
所述终端发送目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
所述终端获取第一数据集,包括:
所述终端接收与所述第一AI单元对应的第一数据集。
本实施方式中,由终端请求针对某一组第一AI单元训练目标AI单元,网络侧设备在允许后下发相应的第一数据集。
作为一种可选的实施方式,所述方法还包括:
所述终端接收第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
本实施方式中,由网络侧设备向终端指示训练与那些第一AI单元或第二数据集关联的目标AI单元。
在本申请实施例中,在基于AI的信道信息上报场景下,在终端侧训练目标AI单元,且该目标AI单元与网络侧使用的第二AI单元相关,这样,可以利用该AI单元来检测所述终端向网络侧设备上报目标信道信息的性能,以及时发现基于AI的压缩-恢复的信道信息不准确的问题。
请参阅图5,是本申请实施例提供的另一种信息处理方法,该信息处理方法的执行主体是第一设备,该第一设备可以包括网络侧设备,如基站或核心网设备,或者,该第一设备可以包括第三方节点,如服务器。为了便于说明,本申请实施例中,通常以第一设备是基站为例进行举例说明。
如图5所示,该信息处理方法可以包括以下步骤:
步骤501、第一设备获取下行信道状态信息。
一种实施方式中,第一设备可以从终端接收下行信道状态信息。
另一种实施方式中,第一设备可以获取预先存储或协议约定的下行信道状态信息。
步骤502、所述第一设备基于所述下行信道状态信息训练第一AI单元和第二AI单元。
步骤503、所述第一设备向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系。
其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向所述网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
在实施中,上述第一AI单元、第二AI单元、第二信息、第一数据集、第一数据集的标识、目标信道信息、第一信道信息、第一信道特征信息、第一处理、第二处理分别与如图4所示方法实施例中的第一AI单元、第二AI单元、第二信息、第一数据集、第一数据集的标识、目标信道信息、第一信道信息、第一信道特征信息、第一处理、第二处理具有相同的含义,在此不再赘述。
作为一种可选的实施方式,所述方法还包括:
所述第一设备向所述终端发送第二数据集,其中,所述第二数据集用于所述终端训练所述第一AI单元,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息。
其中,上述第二数据集与如图4所示方法实施例中的第二数据集的含义和作用相同,在此不再赘述。
本实施方式中,第一设备向终端发送:联合训练第一AI单元和第二AI单元的过程中产生的所述第一AI单元的输入信息和所述第一AI单元的输出信息,可以提升终端据此训练的第一AI单元与第一设备的第二AI单元之间的匹配度。
作为一种可选的实施方式,所述第一数据集包括以下至少一项:
与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入、所述第二AI单元的输入和所述第二AI单元的输出;
与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
作为一种可选的实施方式,所述方法还包括:
所述第一设备向所述终端发送第一指示信息,所述第一指示信息指示第一关联关系,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
作为一种可选的实施方式,所述第一关联关系包括以下至少一项:
所述目标AI单元与至少一个所述第一AI单元的关联关系;
所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
作为一种可选的实施方式,在所述目标AI单元与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和 M分别为正整数。
作为一种可选的实施方式,N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
作为一种可选的实施方式,所述方法还包括:
所述第一设备接收来自所述终端发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
所述终端是否支持训练所述目标AI单元;
所述第一设备向终端发送第二信息,包括:
在所述目标能力信息指示所述终端具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力,或所述终端使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能,或所述终端支持训练所述目标AI单元的情况下,所述第一设备向终端发送第二信息。
作为一种可选的实施方式,所述方法还包括:
所述第一设备接收来自所述终端的与所述目标AI单元相关的目标状态信息。
作为一种可选的实施方式,所述目标状态信息指示以下至少一项:
每个第二数据集是否有关联的目标AI单元;
每个第一AI单元是否有关联的目标AI单元;
每个第二数据集是否需要新的目标AI单元;
每个第一AI单元是否需要新的目标AI单元。
作为一种可选的实施方式,所述方法还包括:
所述第一设备接收来自所述终端的第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
作为一种可选的实施方式,所述方法还包括:
所述第一设备接收来自所述终端的目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
所述第二信息包括与所述第一AI单元对应的第一数据集,或与所述第一AI单元对应的第一数据集的标识。
作为一种可选的实施方式,所述方法还包括:
所述第一设备向所述终端发送第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
作为一种可选的实施方式,所述方法还包括:
所述第一设备接收来自所述终端的第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
作为一种可选的实施方式,所述方法还包括:
所述第一设备向所述终端发送第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
所述终端基于所述目标AI单元确定所述第一信息;
所述第一信息的内容;
所述终端向网络侧设备发送所述第一信息;
所述终端发送所述第一信息的方式。
本申请实施例中,第一设备可以联合训练第一AI单元和第二AI单元,并向终端发送联合训练第一AI单元和第二AI单元的过程中产生的第一数据集,如:第二AI单元的输入和输出,这样,终端便可以据此训练与上述第二AI单元相关的目标AI单元,其具有与如图4所示方法实施例相似的有益效果,为避免重复,在此不再赘述。
为了便于理解本申请实施例提供的信息处理方法,假设第一设备为基站,以基站和终端之间的信息交互过程为例,对本申请实施例提供的信息处理方法进行举例说明:
场景一:独立指示第一数据集,且基于模型ID管理第一关联关系
在基于AI的CSI压缩中,一个典型的AI单元独立训练过程包括如下步骤:
步骤1a.在能力上报阶段,终端告知基站其可以支持进行独立训练(separate training)并且能够在separate training框架下训练代理AI模型(即目标AI单元)。
2a.基站收集下行信道数据。
3a.基站根据所收集的数据训练一组编码器-解码器模型用于CSI压缩,并根据训练得到的编码器模型得到第二数据集(即该编码器的输入输出)。
4a.基站将第二数据集发送至终端。
5a.终端接收第二数据集,并基于第二数据集训练一个或多个终端侧的编码器,再为每一个终端侧的编码器分配一个模型ID,再将模型ID上报至基站。
6a.基站将终端上报的(编码器)模型ID与本地的解码器模型ID进行关联,从而保证终端侧的编码器与基站测的解码器能够配对使用。
7a.终端针对一个或多个本地编码器向基站请求训练相关的代理AI模型。
8a.基站根据终端的代理模型训练请求将一个或多个第一数据集(即解码器的输入输出)以及与每个第一数据集相关联的解码器ID和/或编码器ID发送至终端。
9a.终端根据每个第一数据集训练代理AI模型。
10a.终端按照与相应第一数据集相关联的解码器ID和或编码器ID将每一个代理AI模型与本地的编码器模型相关联。
11a.终端向基站上报在每个第一数据集上是否成功训练了代理AI模型。
12a.基站根据用户上报的代理AI模型状态指示终端使用代理AI模型进行性能检测,指示的内容包括性能监测的上报内容、上报方式等。
13a.用户使用代理AI模型进行模型监测,一种具体的实现方法为:终端将下行信道测量输入编码器-代理AI模型,获取代理AI模型的输出,计算该输出与CSI测量值的平方广义余弦相似度(Square generalized cosine similarity,SGCS)值,并按照基站的指示将SGCS值进行处理后上报。
14a.终端周期性上报针对本地的每个编码器是否存在可用的代理AI模型。
值得说明的是,编码器可以是周期性更新的,在编码器更新后,其对应的代理AI模型可能也需要更新,或者,代理AI模型可能随终端的位置变换、通信环境变换等失效。终端周期性上报针对本地的每个编码器是否存在可用的代理AI模型,可以使网络侧设备获知是否需要更新或训练编码器的代理AI模型。
场景二:联合指示第一数据集和第二数据集,且基于模型ID管理第一关联关系在基于AI的CSI压缩中,一个典型的AI单元独立训练过程包括如下步骤:
1b.在能力上报阶段,终端告知基站其可以支持进行独立训练(separate training)并且能够在separate training框架下训练代理AI模型(即目标AI单元)。
2b.基站收集下行信道数据。
3b..基站根据所收集的数据训练一组编码器-解码器模型用于CSI压缩,并根据训练得到的编码器模型得到第二数据集(即该编码器的输入输出)与第一数据集(即该解码器的输入输出)。
4b.基站将第二数据集与第一数据集联合发送至终端,一种联合发送的方式是使用三元组(编码器输入-编码器输出-解码器输出)发送数据,该方法默认解码器的输入与编码器的输出相同。
5b.终端接收第一数据集与第二数据集,基于编码器输入-输出的数据集(第二数据集)训练一个或多个终端侧的编码器,再基于解码器输入-输出数据集(第一数据集)训练一个或多个代理AI模型。
6b.终端将训练得到的代理AI模型与本地的编码器进行关联,并为每一个编码器分配模型ID,再将编码器模型ID、编码器模型ID对应的第二数据集、以及是否针对该模型训练了可用的代理AI模型的信息上报至基站。
7b.基站将上报的编码器ID与本地的解码器ID进行关联,将编码器与解码器进行配对。
8b.基站根据用户上报的代理AI模型状态指示终端使用代理AI模型进行性能检测,指示的内容包括性能监测的上报内容、上报方式等。
9b.用户使用代理AI模型进行模型监测,一种具体的实现方法为:终端将下行信道测量输入编码器-代理模型,获取代理模型的输出,计算该输出与CSI测量值的平方余弦相似度(SGCS),并按照基站的指示将SGCS值进行处理后上报。
10b.终端周期性上报针对本地的每个编码器是否存在可用的代理AI模型。
场景三:独立指示第一数据集,且基于功能ID(也可称之为数据集ID)管理第一关联关系
在基于AI的CSI压缩中,一个典型的AI单元独立训练过程包括如下步骤:
1c.在能力上报阶段,终端告知基站其可以支持进行separate training并且能够在separate training框架下训练代理AI模型(目标AI单元)。
2c.基站收集下行信道数据。
3c.基站根据所收集的数据训练一组编码器-解码器模型用于CSI压缩,并根据训练得到的编码器模型得到第二数据集(即该编码器的输入输出所构成的数据集)。
4c.基站将第二数据集以及该第二数据集的ID发送至终端。
5c.终端接收第二数据集,并基于第二数据集训练一个或多个终端侧的编码器。
6c.终端针对某一个或多个第二数据集向基站请求训练相关的代理AI模型。
7c.基站根据终端的代理AI模型训练请求将一个或多个第一数据集(即解码器的输入输出)以及与每个第一数据集相关联的第二数据集ID发送至终端。
8c.终端根据每个第一数据集训练代理AI模型,再将每一个训练得到的代理AI模型与对应的第二数据集ID进行关联,并向基站上报在每个第一数据集上是否成功训练了代理AI模型。
9c.基站根据用户上报的代理AI模型状态指示终端使用代理AI模型进行模型性能检测,指示的内容包括性能监测的上报内容、上报方式等。
10c.用户使用代理AI模型进行模型监测,一种具体的实现方法为:终端将下行信道测量输入编码器-代理AI模型,获取代理AI模型的输出,计算该输出与CSI测量值的平方余弦相似度(SGCS),并按照基站的指示将SGCS值进行处理后上报。
11c.终端周期性上报针对本地的每个第二数据集是否存在可用的代理AI模型。
场景四:联合指示第一数据集和第二数据集,且基于功能ID(也可称之为数据集ID)管理第一关联关系
在基于AI的CSI压缩中,一个典型的AI单元独立训练过程包括如下步骤:
1d.在能力上报阶段,终端告知基站其可以支持进行separate training并且能够在separate training框架下训练代理AI模型(即目标AI单元)。
2d.基站收集下行信道数据。
3d.基站根据所收集的数据训练一组编码器-解码器模型用于CSI压缩,并根据训练得到的编码器模型得到第二数据集(即该编码器的输入输出)与第一数据集(即该解码器的输入输出)。
4d.基站将第二数据集与第一数据集联合发送至终端,一种联合发送的方式是使用三元组(编码器输入-编码器输出-解码器输出)发送数据,该方法默认解码器的输入与编码器的输出相同。
5d.终端接收第一数据集与第二数据集,基于编码器输入-输出的数据集(第二数据集)训练一个或多个终端侧的编码器,再基于解码器输入-输出数据集(第一数据集)训练一个或多个代理AI模型。
6d.终端将训练得到的代理AI模型与第二数据集ID进行关联,并将是否针对该第二数据集训练了可用的代理AI模型的信息上报至基站。
7d.基站根据用户上报的代理AI模型状态指示终端使用代理AI模型进行性能检测,指示的内容包括性能监测的上报内容、上报方式等。
9d.用户使用代理AI模型进行模型监测,一种具体的实现方法为:终端将下行信道测量输入编码器-代理AI模型,获取代理AI模型的输出,计算该输出与CSI测量值的平方余弦相似度(SGCS),并按照基站的指示将SGCS值进行处理后上报。
10d.终端周期性上报针对本地的每个编码器是否存在可用的代理AI模型。
本申请实施例提供的信息处理方法,执行主体可以为信息处理装置。本申请实施例中以信息处理装置执行信息处理方法为例,说明本申请实施例提供的信息处理装置。
请参阅图6,本申请实施例提供的信息处理装置600可以是终端内的装置。如图6所示,该信息处理装置600可以包括以下模块:
第一获取模块601,用于获取第一数据集;
第一训练模块602,用于基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
可选地,信息处理装置600还包括:
第二获取模块,用于获取第二数据集,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息;
所第二训练模块,用于基于所述第二数据集训练所述第一AI单元。
可选地,信息处理装置600还包括:
第一确定模块,用于基于所述目标AI单元确定第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
可选地,所述第一确定模块,具体用:
基于所述第一AI单元对所述第一信道信息进行所述第一处理,得到所述第一信道特 征信息;
基于所述目标AI单元对所述第一信道特征信息进行第三处理,得到第二信道信息;
基于所述第二信道信息和所述第一信道信息的相关性,确定第一信息。
可选地,所述第一数据集包括以下至少一项:
与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
可选地,第一获取模块601,具体用于:
接收来自第一设备的第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系。
可选地,信息处理装置600还包括:
第二确定模块,用于基于第一关联关系,确定所述目标AI单元用于确定所述第一AI单元相关的第一信息,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
可选地,所述第一关联关系包括以下至少一项:
所述目标AI单元与至少一个所述第一AI单元的关联关系;
所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
可选地,信息处理装置600还包括以下至少一项:
第一接收模块,用于接收第一指示信息,所述第一指示信息指示所述第一关联关系;
关联模块,用于将基于第一数据训练的目标AI单元与基于第二数据训练的第一AI单元关联,其中,所述第一数据和所述第二数据与相同的样本数据对应。
可选地,在所述第一数据集与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
可选地,N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
可选地,信息处理装置600还包括:
第三发送模块,用于发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
所述终端是否支持训练所述目标AI单元。
可选地,信息处理装置600还包括:
第三确定模块,用于在所述终端支持独立训练第一AI单元的情况下,确认支持训练所述目标AI单元;
第四确定模块,用于在所述终端仅支持AI单元传递的情况下,确认不支持训练所述目标AI单元。
可选地,信息处理装置600还包括:
第四发送模块,用于发送所述目标AI单元相关的目标状态信息。
可选地,所述目标状态信息指示以下至少一项:
每个第二数据集是否有关联的目标AI单元;
每个第一AI单元是否有关联的目标AI单元;
每个第二数据集是否需要新的目标AI单元;
每个第一AI单元是否需要新的目标AI单元。
可选地,信息处理装置600还包括:
第五发送模块,用于发送第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
可选地,信息处理装置600还包括:
第六发送模块,用于发送目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
第一获取模块601具体用于:
接收与所述第一AI单元对应的第一数据集。
可选地,信息处理装置600还包括:
第二接收模块,用于接收第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
可选地,信息处理装置600还包括:
第五发送模块,用于向网络侧设备发送所述第一信息。
可选地,信息处理装置600还包括:
第三接收模块,用于接收第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
所述终端基于所述目标AI单元确定所述第一信息;
所述第一信息的内容;
所述终端向网络侧设备发送所述第一信息;
所述终端发送所述第一信息的方式。
本申请实施例中的信息处理装置600可以是电子设备,例如具有操作系统的电子设备,也可以是电子设备中的部件,例如集成电路或芯片。该电子设备可以是终端。示例性的,终端可以包括但不限于上述所列举的终端11的类型,本申请实施例不作具体限定。
本申请实施例提供的信息处理装置600能够实现图4所示方法实施例实现的各个过程,并达到相同的技术效果,为避免重复,这里不再赘述。
请参阅图7,本申请实施例提供的另一种信息处理装置,可以是第一设备内的装置,该第一设备可以是网络侧设备,如基站或核心网设备,或者,该第一设备可以是第三方设备,如服务器。
如图7所示,该信息处理装置700可以包括以下模块:
第三获取模块701,用于获取下行信道状态信息;
第三训练模块702,用于基于所述下行信道状态信息训练第一AI单元和第二AI单元;
第一发送模块703,用于向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向所述网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
可选地,信息处理装置700还包括:
第二发送模块,用于向所述终端发送第二数据集,其中,所述第二数据集用于所述终端训练所述第一AI单元,其中,所述第二数据集包括与样本数据一一对应的第一二元组, 所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息。
可选地,所述第一数据集包括以下至少一项:
与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入、所述第二AI单元的输入和所述第二AI单元的输出;
与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
可选地,信息处理装置700还包括:
第七发送模块,用于向所述终端发送第一指示信息,所述第一指示信息指示第一关联关系,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
可选地,所述第一关联关系包括以下至少一项:
所述目标AI单元与至少一个所述第一AI单元的关联关系;
所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
可选地,在所述目标AI单元与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
可选地,N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
可选地,信息处理装置700还包括:
第四接收模块,用于接收来自所述终端发送目标能力信息,其中,所述目标能力信息 指示以下至少一项:
所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
所述终端是否支持训练所述目标AI单元;
第一发送模块703,具体用于:
在所述目标能力信息指示所述终端具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力,或所述终端使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能,或所述终端支持训练所述目标AI单元的情况下,向终端发送第二信息。
可选地,信息处理装置700还包括:
第五接收模块,用于接收来自所述终端的与所述目标AI单元相关的目标状态信息。
可选地,所述目标状态信息指示以下至少一项:
每个第二数据集是否有关联的目标AI单元;
每个第一AI单元是否有关联的目标AI单元;
每个第二数据集是否需要新的目标AI单元;
每个第一AI单元是否需要新的目标AI单元。
可选地,信息处理装置700还包括:
第六接收模块,用于接收来自所述终端的第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
可选地,信息处理装置700还包括:
第七接收模块,用于接收来自所述终端的目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
所述第二信息包括与所述第一AI单元对应的第一数据集,或与所述第一AI单元对应的第一数据集的标识。
可选地,信息处理装置700还包括:
第七发送模块,用于向所述终端发送第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
可选地,信息处理装置700还包括:
第八接收模块,用于接收来自所述终端的第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
可选地,信息处理装置700还包括:
第八发送模块,用于向所述终端发送第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
所述终端基于所述目标AI单元确定所述第一信息;
所述第一信息的内容;
所述终端向网络侧设备发送所述第一信息;
所述终端发送所述第一信息的方式。
本申请实施例提供的信息处理装置700能够实现图5所示方法实施例实现的各个过程,并达到相同的技术效果,为避免重复,这里不再赘述。
如图8所示,本申请实施例还提供一种通信设备800,包括处理器801和存储器802,存储器802上存储有可在所述处理器801上运行的程序或指令,例如,该通信设备800为终端时,该程序或指令被处理器801执行时实现如图4所示信息处理方法实施例的各个步骤,且能达到相同的技术效果。该通信设备800为第一设备时,该程序或指令被处理器801执行时实现如图5所示信息处理方法实施例的各个步骤,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供一种终端,包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如图4所示方法实施例中的步骤。该终端实施例与上述终端侧方法实施例对应,上述方法实施例的各个实施过程和实现方式均可适用于该终端实施例中,且能达到相同的技术效果。具体地,图9为实现本申请实施例的一种终端的硬件结构示意图。
该终端900包括但不限于:射频单元901、网络模块902、音频输出单元903、输入单元904、传感器905、显示单元906、用户输入单元907、接口单元908、存储器909以及处理器910等中的至少部分部件。
本领域技术人员可以理解,终端900还可以包括给各个部件供电的电源(比如电池),电源可以通过电源管理系统与处理器910逻辑相连,从而通过电源管理系统实现管理充电、放电以及功耗管理等功能。图9中示出的终端结构并不构成对终端的限定,终端可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置,在此不再赘述。
应理解的是,本申请实施例中,输入单元904可以包括图形处理器(Graphics Processing Unit,GPU)9041和麦克风9042,图形处理器9041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元906可包括显示面板9061,可以采用液晶显示器、有机发光二极管等形式来配置显示面板9061。用户输入单元907包括触控面板9071以及其他输入设备9072中的至少一种。触控面板9071,也称为触摸屏。触控面板9071可包括触摸检测装置和触摸控制器两个部分。其他输入设备9072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆,在此不再赘述。
本申请实施例中,射频单元901接收来自网络侧设备的下行数据后,可以传输给处理器910进行处理;另外,射频单元901可以向网络侧设备发送上行数据。通常,射频单元901包括但不限于天线、放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器909可用于存储软件程序或指令以及各种数据。存储器909可主要包括存储程序或指令的第一存储区和存储数据的第二存储区,其中,第一存储区可存储操作系统、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播放功能等)等。此外,存储器909可以包括易失性存储器或非易失性存储器。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),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,DDRSDRAM)、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)和直接内存总线随机存取存储器(Direct Rambus RAM,DRRAM)。本申请实施例中的存储器909包括但不限于这些和任意其它适合类型的存储器。
处理器910可包括一个或多个处理单元;可选地,处理器910集成应用处理器和调制解调处理器,其中,应用处理器主要处理涉及操作系统、用户界面和应用程序等的操作,调制解调处理器主要处理无线通信信号,如基带处理器。可以理解的是,上述调制解调处理器也可以不集成到处理器910中。
其中,射频单元901,用于获取第一数据集;
处理器910,用于基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
可选地,射频单元901,还用于获取第二数据集,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息;
处理器910,还用于基于所述第二数据集训练所述第一AI单元。
可选地,处理器910,还用于基于所述目标AI单元确定第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
可选地,处理器910执行的所述基于所述目标AI单元确定第一信息,包括:
基于所述第一AI单元对所述第一信道信息进行所述第一处理,得到所述第一信道特征信息;
基于所述目标AI单元对所述第一信道特征信息进行第三处理,得到第二信道信息;
基于所述第二信道信息和所述第一信道信息的相关性,确定第一信息。
可选地,所述第一数据集包括以下至少一项:
与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息;
与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
可选地,射频单元901执行的所述获取第一数据集,包括:
接收来自第一设备的第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系。
可选地,处理器910,还用于基于第一关联关系,确定所述目标AI单元用于确定所述第一AI单元相关的第一信息,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
可选地,所述第一关联关系包括以下至少一项:
所述目标AI单元与至少一个所述第一AI单元的关联关系;
所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
可选地,射频单元901,还用于接收第一指示信息,所述第一指示信息指示所述第一关联关系;或,
处理器910,还用于将基于第一数据训练的目标AI单元与基于第二数据训练的第一AI单元关联,其中,所述第一数据和所述第二数据与相同的样本数据对应。
可选地,在所述第一数据集与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
可选地,N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
可选地,射频单元901,还用于发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
所述终端是否支持训练所述目标AI单元。
可选地,处理器910,还用于在所述终端支持独立训练第一AI单元的情况下,确认支持训练所述目标AI单元;
处理器910,还用于在所述终端仅支持AI单元传递的情况下,确认不支持训练所述目标AI单元。
可选地,射频单元901,还用于发送所述目标AI单元相关的目标状态信息。
可选地,所述目标状态信息指示以下至少一项:
每个第二数据集是否有关联的目标AI单元;
每个第一AI单元是否有关联的目标AI单元;
每个第二数据集是否需要新的目标AI单元;
每个第一AI单元是否需要新的目标AI单元。
可选地,射频单元901,还用于发送第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
可选地,射频单元901,还用于发送目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
射频单元901执行的所述获取第一数据集,包括:
接收与所述第一AI单元对应的第一数据集。
可选地,射频单元901,还用于接收第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
可选地,射频单元901,还用于向网络侧设备发送所述第一信息。
可选地,射频单元901,还用于接收第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
所述终端基于所述目标AI单元确定所述第一信息;
所述第一信息的内容;
所述终端向网络侧设备发送所述第一信息;
所述终端发送所述第一信息的方式。
可以理解,本实施例中提及的各实现方式的实现过程可以参照如图4所示信息处理方法实施例的相关描述,并达到相同或相应的技术效果,为避免重复,在此不再赘述。
本申请实施例还提供一种网络侧设备,包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如图5所示的方法实施例的步骤。该网络侧设备实施例与上述网络侧设备方法实施例对应,上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中,且能达到相同的技术效果。
具体地,本申请实施例还提供了一种网络侧设备。如图10所示,该网络侧设备1000包括:天线1001、射频装置1002、基带装置1003、处理器1004和存储器1005。天线1001与射频装置1002连接。在上行方向上,射频装置1002通过天线1001接收信息,将接收的信息发送给基带装置1003进行处理。在下行方向上,基带装置1003对要发送的信息进行处理,并发送给射频装置1002,射频装置1002对收到的信息进行处理后经过天线1001发送出去。
以上实施例中网络侧设备执行的方法可以在基带装置1003中实现,该基带装置1003包括基带处理器。
基带装置1003例如可以包括至少一个基带板,该基带板上设置有多个芯片,如图10所示,其中一个芯片例如为基带处理器,通过总线接口与存储器1005连接,以调用存储器1005中的程序,执行以上方法实施例中所示的网络设备操作。
该网络侧设备还可以包括网络接口1006,该接口例如为通用公共无线接口(Common Public Radio Interface,CPRI)。
具体地,本申请实施例的网络侧设备1000还包括:存储在存储器1005上并可在处理器1004上运行的指令或程序,处理器1004调用存储器1005中的指令或程序执行图7所示各模块执行的方法,并达到相同的技术效果,为避免重复,故不在此赘述。
具体地,本申请实施例还提供了一种网络侧设备。如图11所示,该网络侧设备1100包括:处理器1101、网络接口1102和存储器1103。其中,网络接口1102例如为通用公共无线接口(Common Public Radio Interface,CPRI)。
具体地,本申请实施例的网络侧设备1100还包括:存储在存储器1103上并可在处理器1101上运行的指令或程序,处理器1101调用存储器1103中的指令或程序执行如图7所示各模块执行的方法,并达到相同的技术效果,为避免重复,故不在此赘述。
本申请实施例还提供一种可读存储介质,所述可读存储介质上存储有程序或指令,该程序或指令被处理器执行时实现如图4或图5所示方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
其中,所述处理器为上述实施例中所述的终端中的处理器。所述可读存储介质,包括 计算机可读存储介质,如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。在一些示例中,可读存储介质可以是非瞬态的可读存储介质。
本申请实施例另提供了一种芯片,所述芯片包括处理器和通信接口,所述通信接口和所述处理器耦合,所述处理器用于运行程序或指令,实现如图4或图5所示方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
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。
本申请实施例另提供了一种计算机程序/程序产品,所述计算机程序/程序产品被存储在存储介质中,所述计算机程序/程序产品被至少一个处理器执行以实现如图4或图5所示方法实施例的各个过程,且能达到相同的技术效果,为避免重复,这里不再赘述。
本申请实施例还提供了一种通信系统,包括:终端和第一设备,所述终端可用于执行如图4所示信息处理方法的步骤,所述第一设备可用于执行如图5所示信息处理方法的步骤。
需要说明的是,在本文中,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外,需要指出的是,本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能,还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能,例如,可以按不同于所描述的次序来执行所描述的方法,并且还可以添加、省去或组合各种步骤。另外,参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助计算机软件产品加必需的通用硬件平台的方式来实现,当然也可以通过硬件。该计算机软件产品存储在存储介质(如ROM、RAM、磁碟、光盘等)中,包括若干指令,用以使得终端或者网络侧设备执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述,但是本申请并不局限于上述的具体实施方式,上述的具体实施方式仅仅是示意性的,而不是限制性的,本领域的普通技术人员在本申请的启示下,在不脱离本申请宗旨和权利要求所保护的范围情况下,还可做出很多形式的实施方式,这些实施方式均属于本申请的保护之内。

Claims (43)

  1. 一种信息处理方法,包括:
    终端获取第一数据集;
    所述终端基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
  2. 根据权利要求1所述的方法,其中,所述方法还包括:
    所述终端获取第二数据集,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息;
    所述终端基于所述第二数据集训练所述第一AI单元。
  3. 根据权利要求1或2所述的方法,其中,所述方法还包括:
    所述终端基于所述目标AI单元确定第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
  4. 根据权利要求3所述的方法,其中,所述终端基于所述目标AI单元确定第一信息,包括:
    所述终端基于所述第一AI单元对所述第一信道信息进行所述第一处理,得到所述第一信道特征信息;
    所述终端基于所述目标AI单元对所述第一信道特征信息进行第三处理,得到第二信道信息;
    所述终端基于所述第二信道信息和所述第一信道信息的相关性,确定第一信息。
  5. 根据权利要求1至4中任一项所述的方法,其中,所述第一数据集包括以下至少一项:
    与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
    与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息;
    与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
  6. 根据权利要求1至5中任一项所述的方法,其中,所述终端获取第一数据集,包括:
    所述终端接收来自第一设备的第二信息,所述第二信息包括第一数据集,或,所述第 二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系。
  7. 根据权利要求1至6中任一项所述的方法,其中,所述方法还包括:
    所述终端基于第一关联关系,确定所述目标AI单元用于确定所述第一AI单元相关的第一信息,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
  8. 根据权利要求7所述的方法,其中,所述第一关联关系包括以下至少一项:
    所述目标AI单元与至少一个所述第一AI单元的关联关系;
    所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
    所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
    所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
    所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
    所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
  9. 根据权利要求8所述的方法,其中,所述方法还包括以下至少一项:
    所述终端接收第一指示信息,所述第一指示信息指示所述第一关联关系;
    所述终端将基于第一数据训练的目标AI单元与基于第二数据训练的第一AI单元关联,其中,所述第一数据和所述第二数据与相同的样本数据对应。
  10. 根据权利要求8或9所述的方法,其中,在所述第一数据集与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
  11. 根据权利要求10所述的方法,其中:
    N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
    N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
    N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
  12. 根据权利要求1至11中任一项所述的方法,其中,所述方法还包括:
    所述终端发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
    所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
    所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
    所述终端是否支持训练所述目标AI单元。
  13. 根据权利要求12所述的方法,其中,所述方法还包括:
    在所述终端支持独立训练第一AI单元的情况下,所述终端确认支持训练所述目标AI单元;
    在所述终端仅支持AI单元传递的情况下,所述终端确认不支持训练所述目标AI单元。
  14. 根据权利要求1至13中任一项所述的方法,其中,所述方法还包括:
    所述终端发送所述目标AI单元相关的目标状态信息。
  15. 根据权利要求14所述的方法,其中,其中,所述目标状态信息指示以下至少一项:
    每个第二数据集是否有关联的目标AI单元;
    每个第一AI单元是否有关联的目标AI单元;
    每个第二数据集是否需要新的目标AI单元;
    每个第一AI单元是否需要新的目标AI单元。
  16. 根据权利要求1至15中任一项所述的方法,其中,所述方法还包括:
    所述终端发送第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单元。
  17. 根据权利要求1至16中任一项所述的方法,其中,所述方法还包括:
    所述终端发送目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
    所述终端获取第一数据集,包括:
    所述终端接收与所述第一AI单元对应的第一数据集。
  18. 根据权利要求1至17中任一项所述的方法,其中,所述方法还包括:
    所述终端接收第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
  19. 根据权利要求3至18中任一项所述的方法,其中,所述方法还包括:
    所述终端向网络侧设备发送所述第一信息。
  20. 根据权利要求3至19中任一项所述的方法,其中,所述方法还包括:
    所述终端接收第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
    所述终端基于所述目标AI单元确定所述第一信息;
    所述第一信息的内容;
    所述终端向网络侧设备发送所述第一信息;
    所述终端发送所述第一信息的方式。
  21. 一种信息处理方法,包括:
    第一设备获取下行信道状态信息;
    所述第一设备基于所述下行信道状态信息训练第一AI单元和第二AI单元;
    所述第一设备向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
    其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
  22. 根据权利要求21所述的方法,其中,所述方法还包括:
    所述第一设备向所述终端发送第二数据集,其中,所述第二数据集用于所述终端训练所述第一AI单元,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息。
  23. 根据权利要求21或22所述的方法,其中,所述第一数据集包括以下至少一项:
    与样本数据一一对应的第二二元组,所述第二二元组包括:所述第二AI单元的输入信息和所述第二AI单元的输出信息;
    与样本数据一一对应的三元组,所述三元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息和所述第二AI单元的输出信息,或者,所述三元组包括:所述第一AI单元的输入、所述第二AI单元的输入和所述第二AI单元的输出;
    与样本数据一一对应的四元组,所述四元组包括:所述第一AI单元的输入信息、所述第一AI单元的输出信息、所述第二AI单元的输入信息和所述第二AI单元的输出信息。
  24. 根据权利要求21至23中任一项所述的方法,其中,所述方法还包括:
    所述第一设备向所述终端发送第一指示信息,所述第一指示信息指示第一关联关系,其中,所述第一关联关系指示所述目标AI单元与所述第一AI单元之间的关联关系。
  25. 根据权利要求24所述的方法,其中,所述第一关联关系包括以下至少一项:
    所述目标AI单元与至少一个所述第一AI单元的关联关系;
    所述目标AI单元与至少一个所述第二AI单元的关联关系,其中,所述第一AI单元与所述第二AI单元对应;
    所述目标AI单元与第二数据集或所述第二数据集的标识的关联关系,其中,所述第二数据集用于训练至少一个所述第一AI单元,所述目标AI单元用于确定基于所述第二数据集训练的第一AI单元相关的第一信息;
    所述第一数据集与至少一个所述第一AI单元或至少一个所述第一AI单元的标识的关联关系;
    所述第一数据集与至少一个所述第二AI单元或至少一个所述第二AI单元的标识的关联关系;
    所述第一数据集与所述第二数据集或所述第二数据集的标识的关联关系。
  26. 根据权利要求25所述的方法,其中,在所述目标AI单元与第二数据集关联的情况下,所述第二数据集用于训练M个第一AI单元,所述第一数据集用于训练N个目标AI单元,所述第一关联关系包括所述N个目标AI单元与所述M个第一AI单元之间的关联关系,N和M分别为正整数。
  27. 根据权利要求26所述的方法,其中:
    N大于M,所述N个目标AI单元中的一个目标AI单元,与所述M个第一AI单元中的至少一个第一AI单元关联;
    N小于M,所述N个目标AI单元中的至少一个目标AI单元,与所述M个第一AI单元中的一个第一AI单元关联;
    N等于M,所述N个目标AI单元与所述M个第一AI单元一一对应关联。
  28. 根据权利要求21至27中任一项所述的方法,其中,所述方法还包括:
    所述第一设备接收来自所述终端发送目标能力信息,其中,所述目标能力信息指示以下至少一项:
    所述终端是否具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力;
    所述终端是否使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能;
    所述终端是否支持训练所述目标AI单元;
    所述第一设备向终端发送第二信息,包括:
    在所述目标能力信息指示所述终端具备在独立训练第一AI单元的框架下训练所述目标AI单元的能力,或所述终端使用所述目标AI单元检测所述终端向网络侧设备上报目标信道信息的性能,或所述终端支持训练所述目标AI单元的情况下,所述第一设备向终端发送第二信息。
  29. 根据权利要求21至28中任一项所述的方法,其中,所述方法还包括:
    所述第一设备接收来自所述终端的与所述目标AI单元相关的目标状态信息。
  30. 根据权利要求29所述的方法,其中,其中,所述目标状态信息指示以下至少一项:
    每个第二数据集是否有关联的目标AI单元;
    每个第一AI单元是否有关联的目标AI单元;
    每个第二数据集是否需要新的目标AI单元;
    每个第一AI单元是否需要新的目标AI单元。
  31. 根据权利要求21至30中任一项所述的方法,其中,所述方法还包括:
    所述第一设备接收来自所述终端的第二指示信息,所述第二指示信息指示所述终端需要的目标AI单元的第一标识信息,所述第一标识信息用于标识第二数据集或第一AI单 元。
  32. 根据权利要求21至31中任一项所述的方法,其中,所述方法还包括:
    所述第一设备接收来自所述终端的目标请求信息,其中,所述目标请求信息用于请求训练与所述第一AI单元关联的目标AI单元;
    所述第二信息包括与所述第一AI单元对应的第一数据集,或与所述第一AI单元对应的第一数据集的标识。
  33. 根据权利要求21至32中任一项所述的方法,其中,所述方法还包括:
    所述第一设备向所述终端发送第三指示信息,其中,所述第三指示信息用于指示所述终端训练与所述第一AI单元关联的目标AI单元,或训练与第二数据集关联的目标AI单元,所述第二数据集用于训练至少一个所述第一AI单元。
  34. 根据权利要求21至33中任一项所述的方法,其中,所述方法还包括:
    所述第一设备接收来自所述终端的第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
  35. 根据权利要求34中任一项所述的方法,其中,所述方法还包括:
    所述第一设备向所述终端发送第四指示信息,其中,所述第四指示信息用于指示以下至少一项:
    所述终端基于所述目标AI单元确定所述第一信息;
    所述第一信息的内容;
    所述终端向网络侧设备发送所述第一信息;
    所述终端发送所述第一信息的方式。
  36. 一种信息处理装置,包括:
    第一获取模块,用于获取第一数据集;
    第一训练模块,用于基于所述第一数据集训练目标AI单元,其中,所述目标AI单元用于检测终端向网络侧设备上报目标信道信息的性能,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于网络侧设备的第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息,所述目标AI单元与所述第二AI单元相关。
  37. 根据权利要求36所述的装置,其中,所述装置还包括:
    第二获取模块,用于获取第二数据集,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息;
    所第二训练模块,用于基于所述第二数据集训练所述第一AI单元。
  38. 根据权利要求36或37所述的装置,其中,所述装置还包括:
    第一确定模块,用于基于所述目标AI单元确定第一信息,其中,所述第一信息与所述终端向网络侧设备上报目标信道信息的性能相关。
  39. 一种信息处理装置,包括:
    第三获取模块,用于获取下行信道状态信息;
    第三训练模块,用于基于所述下行信道状态信息训练第一AI单元和第二AI单元;
    第一发送模块,用于向终端发送第二信息,所述第二信息包括第一数据集,或,所述第二信息包括所述第一数据集的标识,且所述终端获知了所述第一数据集与所述第一数据集的标识之间的关联关系;
    其中,所述第一数据集用于训练目标AI单元,所述目标AI单元用于检测所述终端向网络侧设备上报目标信道信息的性能,其中,所述目标信道信息基于所述终端的第一AI单元对第一信道信息的第一处理,以及基于所述第二AI单元对第一信道特征信息的第二处理得到,所述第一信道特征信息为基于所述终端的第一AI单元对所述第一信道信息进行所述第一处理得到的信息。
  40. 根据权利要求39所述的装置,其中,所述装置还包括:
    第二发送模块,用于向所述终端发送第二数据集,其中,所述第二数据集用于所述终端训练所述第一AI单元,其中,所述第二数据集包括与样本数据一一对应的第一二元组,所述第一二元组包括:所述第一AI单元的输入信息和所述第一AI单元的输出信息。
  41. 一种终端,包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,其中,所述程序或指令被所述处理器执行时实现如权利要求1至20中任一项所述的信息处理方法的步骤。
  42. 一种网络侧设备,包括处理器和存储器,所述存储器存储可在所述处理器上运行的程序或指令,其中,所述程序或指令被所述处理器执行时,实现如权利要求21至35中任一项所述的信息处理方法的步骤。
  43. 一种可读存储介质,所述可读存储介质上存储程序或指令,其中,所述程序或指令被处理器执行时实现如权利要求1至20中任一项所述的信息处理方法的步骤,或者实现如权利要求21至35中任一项所述的信息处理方法的步骤。
PCT/CN2024/088579 2023-04-20 2024-04-18 信息处理方法、信息处理装置、终端及网络侧设备 WO2024217495A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310431601.7A CN118828559A (zh) 2023-04-20 2023-04-20 信息处理方法、信息处理装置、终端及网络侧设备
CN202310431601.7 2023-04-20

Publications (1)

Publication Number Publication Date
WO2024217495A1 true WO2024217495A1 (zh) 2024-10-24

Family

ID=93080559

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/088579 WO2024217495A1 (zh) 2023-04-20 2024-04-18 信息处理方法、信息处理装置、终端及网络侧设备

Country Status (2)

Country Link
CN (1) CN118828559A (zh)
WO (1) WO2024217495A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210065000A1 (en) * 2019-08-30 2021-03-04 Samsung Electronics Co., Ltd. Electronic apparatus for applying personalized artificial intelligence model to another model
WO2022012257A1 (zh) * 2020-07-13 2022-01-20 华为技术有限公司 通信的方法及通信装置
CN114726413A (zh) * 2021-01-04 2022-07-08 中国移动通信有限公司研究院 信道信息获取方法、装置及相关设备
CN115802370A (zh) * 2021-09-10 2023-03-14 华为技术有限公司 一种通信方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210065000A1 (en) * 2019-08-30 2021-03-04 Samsung Electronics Co., Ltd. Electronic apparatus for applying personalized artificial intelligence model to another model
WO2022012257A1 (zh) * 2020-07-13 2022-01-20 华为技术有限公司 通信的方法及通信装置
CN114726413A (zh) * 2021-01-04 2022-07-08 中国移动通信有限公司研究院 信道信息获取方法、装置及相关设备
CN115802370A (zh) * 2021-09-10 2023-03-14 华为技术有限公司 一种通信方法及装置

Also Published As

Publication number Publication date
CN118828559A (zh) 2024-10-22

Similar Documents

Publication Publication Date Title
JP2022553521A (ja) 符号化方法、復号方法、ユーザ機器及びネットワーク機器
US20230412430A1 (en) Inforamtion reporting method and apparatus, first device, and second device
WO2023246618A1 (zh) 信道矩阵处理方法、装置、终端及网络侧设备
WO2023185978A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2024217495A1 (zh) 信息处理方法、信息处理装置、终端及网络侧设备
WO2024222573A1 (zh) 信息处理方法、信息处理装置、终端及网络侧设备
WO2024037380A1 (zh) 信道信息处理方法、装置、通信设备及存储介质
WO2024149156A1 (zh) 信息传输方法、装置、终端及网络侧设备
WO2024032606A1 (zh) 信息传输方法、装置、设备、系统及存储介质
WO2024149157A1 (zh) Csi传输方法、装置、终端及网络侧设备
WO2024140422A1 (zh) Ai单元的性能监测方法、终端及网络侧设备
WO2024055974A1 (zh) Cqi传输方法、装置、终端及网络侧设备
WO2024055993A1 (zh) Cqi传输方法、装置、终端及网络侧设备
WO2024222577A1 (zh) 信息处理方法、信息传输方法、装置、终端及网络侧设备
WO2024222601A1 (zh) Cpu数量上报方法、接收方法、装置、终端及网络侧设备
WO2024140578A1 (zh) 基于ai模型的csi反馈方法、终端及网络侧设备
WO2023179476A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2023179473A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2024007949A1 (zh) Ai模型处理方法、装置、终端及网络侧设备
WO2024104126A1 (zh) 更新ai网络模型的方法、装置和通信设备
WO2024088161A1 (zh) 信息传输方法、信息处理方法、装置和通信设备
WO2023179570A1 (zh) 信道特征信息传输方法、装置、终端及网络侧设备
WO2024164962A1 (zh) 通信处理方法、装置、设备及可读存储介质
WO2024099094A1 (zh) 波束测量方法、装置、终端、网络侧设备及存储介质
WO2024230540A1 (zh) 模型处理方法、装置及通信设备