WO2025252002A1 - 小区驻留处理方法、装置、终端及网络侧设备 - Google Patents
小区驻留处理方法、装置、终端及网络侧设备Info
- Publication number
- WO2025252002A1 WO2025252002A1 PCT/CN2025/098146 CN2025098146W WO2025252002A1 WO 2025252002 A1 WO2025252002 A1 WO 2025252002A1 CN 2025098146 W CN2025098146 W CN 2025098146W WO 2025252002 A1 WO2025252002 A1 WO 2025252002A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- model
- information
- terminal
- cell
- synchronization signal
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W36/00—Hand-off or reselection arrangements
- H04W36/08—Reselecting an access point
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W48/00—Access restriction; Network selection; Access point selection
- H04W48/16—Discovering, processing access restriction or access information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W48/00—Access restriction; Network selection; Access point selection
- H04W48/20—Selecting an access point
Definitions
- This application belongs to the field of communication technology, specifically relating to a cell dwell processing method, apparatus, terminal, and network-side equipment.
- a terminal when a terminal powers on and performs a cell search, it typically first reads the information from the Subscriber Identity Module (SIM) card. Based on the prior information stored in the SIM card (such as previously stored cell and frequency information), it searches for a suitable cell to camp on. If the terminal does not store prior information, or if it cannot find a suitable cell to camp on based on the stored prior information (e.g., the stored prior information is invalid or inapplicable), the terminal will trigger a frequency scan and network search on the frequency bands supported by the SIM card to perform initial cell selection. This process takes a relatively long time, resulting in a longer delay in the terminal's cell camping process.
- SIM Subscriber Identity Module
- This application provides a cell camping processing method, apparatus, terminal, and network-side equipment, which can solve the problem that the terminal has a long delay in cell camping due to the long time for initial cell search and selection.
- a method for handling community residency issues including:
- the first device obtains the target result based on the first artificial intelligence (AI) model.
- AI artificial intelligence
- the first device is a terminal, a network-side device, or a server
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, wherein the first cell is a cell that can be camped.
- a method for handling community residency issues including:
- the second device performs at least one of the following:
- the second device sends at least a portion of the first information to the first device, the first information being used for inference in the first AI model;
- the second operation includes at least one of the following:
- the second device trains the first AI model, obtains the first AI model, and sends the first AI model to the first device;
- the second device trains the first AI model, obtains a second training result, and sends the second training result to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a terminal, a network-side device, or a server
- the second device is a terminal, a network-side device, or a server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information of the first cell, wherein the first cell is a cell that can be camped.
- a community dwell processing device applied to the first device, comprising:
- the first processing module is used to obtain the target result based on the first artificial intelligence (AI) model.
- AI artificial intelligence
- the first device is a terminal, a network-side device, or a server
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, wherein the first cell is a cell that can be camped.
- a community dwell processing device applied to a second device, comprising:
- An execution module is used to perform at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a terminal, a network-side device, or a server
- the second device is a terminal, a network-side device, or a server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information of the first cell, wherein the first cell is a cell that can be camped.
- an apparatus for cell dwell processing is provided, the apparatus being configured to perform the steps of the method described in the first aspect, or to implement the steps of the method described in the second aspect.
- a terminal including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect.
- a terminal including a processor and a communication interface, wherein,
- the processor is used to obtain the target result based on the first artificial intelligence (AI) model;
- the first device is a terminal, a network-side device, or a server
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, wherein the first cell is a cell that can be camped.
- the communication interface is used to perform at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a network-side device or server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and the information of the first cell, which is a cell that can be camped.
- a network-side device including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect.
- a network-side device including a processor and a communication interface, wherein,
- the processor is used to obtain the target result based on the first artificial intelligence (AI) model;
- the target result is used for cell camping, and the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, which is a cell that can be camped.
- the communication interface is used to perform at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a terminal or a server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information of the first cell, wherein the first cell is a cell that can be camped.
- a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect, or implement the steps of the method described in the second aspect.
- a wireless communication system comprising: a first device and a second device, wherein the first device is configured to perform the steps of the method as described in the first aspect, and the second device is configured to perform the steps of the method as described in the second aspect.
- a chip including a processor and a communication interface coupled to the processor, the processor being configured to run programs or instructions to implement the method as described in the first aspect, or to implement the method as described in the second aspect.
- a computer program/program product is provided, which is stored in a storage medium and is executed by at least one processor to implement the steps of the method as described in the first aspect, or to implement the steps of the method as described in the second aspect.
- This application embodiment obtains a target result through a first device based on a first artificial intelligence (AI) model.
- the first device is a terminal, network-side device, or server.
- the target result is used for cell camping and includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about a first cell, which is a cell suitable for camping.
- the terminal can perform initial cell search and selection based on the time-frequency domain location, beam transmission direction, and first cell information in the target result, thereby avoiding frequency scanning and network searching across the entire frequency band supported by the terminal. Therefore, this application embodiment shortens the initial cell search and selection time, thereby reducing the latency of cell camping and simultaneously reducing the energy consumption of initial cell search and selection.
- Figure 1 is a block diagram of a wireless communication system applicable to an embodiment of this application
- Figures 2a to 2c are examples of reuse styles for SSB and CORESET 0;
- Figure 3 is a schematic diagram of the neuron structure
- Figure 4 is a flowchart illustrating a method for handling cell dwelling provided in an embodiment of this application
- Figure 6 is a schematic diagram of the structure of a community dwell processing device provided in an embodiment of this application.
- FIG. 7 is a schematic diagram of another cell dwell processing device provided in an embodiment of this application.
- Figure 8 is a schematic diagram of the structure of a communication device provided in an embodiment of this application.
- Figure 9 is a schematic diagram of the structure of a terminal provided in an embodiment of this application.
- Figure 10 is a schematic diagram of the structure of a network-side device provided in an embodiment of this application.
- first and second are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by “first” and “second” are generally of the same class, not limited in number; for example, the first object can be one or more.
- “or” in this application indicates at least one of the connected objects.
- the scope of protection for "A or B” covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B.
- the terms “A and/or B,” “at least one of A and B,” and “at least one of A or B” also cover at least the above three scenarios.
- the character “/” generally indicates that the preceding and following objects are in an "or” relationship.
- instruction in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction).
- a direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc., in the instruction sent.
- An indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on 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
- FIG. 1 shows a block diagram of a wireless communication system applicable to an embodiment of this application.
- the wireless communication system includes a terminal 11 and a network-side device 12.
- Terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home (home devices with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game console, personal computer (PC), ATM, or self-service machine, etc.
- PDA personal digital assistant
- UMPC ultra-mobile personal computer
- MID mobile internet device
- AR augmented reality
- VR virtual reality
- robot wearable device
- flight vehicle vehicle user equipment
- VUE shipboard equipment
- pedestrian user equipment PUE
- smart home home devices with wireless communication capabilities, such as refrigerators, televisions, washing machines
- Wearable devices include: smartwatches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc.
- in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment.
- Network-side equipment 12 may include access network equipment or core network equipment, wherein access network equipment may also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit.
- Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (AS), or Wireless Fidelity (WiFi) nodes, etc.
- WLAN Wireless Local Area Network
- WiFi Wireless Fidelity
- a base station may be referred to as a Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit/Receive Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved.
- the base station is not limited to specific technical terms. It should be noted that in this application embodiment, only a base station in an NR system is used as an example for introduction, and the specific type of base station is not limited.
- Cell search is the process by which a terminal obtains time and frequency synchronization information from a cell and decodes the cell's identifier (ID).
- ID the cell's identifier
- PCI Physical Cell Identity
- NR cell search is based on the Primary Synchronization Signal (PSS), the Secondary Synchronization Signal (SSS), and the Demodulation Reference Signal (DMRS) of the Physical Broadcast Channel (PBCH) located on the synchronization grating.
- PSS Primary Synchronization Signal
- SSS Secondary Synchronization Signal
- DMRS Demodulation Reference Signal
- PBCH Physical Broadcast Channel
- the terminal is tuned to a specific frequency and only the Received Signal Strength Indication (RSSI) is measured.
- RSSI Received Signal Strength Indication
- the terminal attempts to detect the SSB and decode the Primary Synchronization Signal (PSS) and Secondary Synchronization Signal (SSS). If the terminal fails at this step, it proceeds to step 1; if it succeeds, it proceeds to the next step.
- PSS Primary Synchronization Signal
- SSS Secondary Synchronization Signal
- MIB Master Information Block
- Control Resource Set 0 i.e., the CORESET of Physical Downlink Control Channel (PDCCH) or Downlink Control Information (DCI) used for SIB1 transmission.
- PDCH Physical Downlink Control Channel
- DCI Downlink Control Information
- PDSCH Physical Downlink Shared Channel
- SIB System Information Block
- SIB1 and other SIBs (if SIB1 carries information about other SIBs).
- the terminal will continuously search for neighboring cells, obtain synchronization, and estimate the reception quality of the cell signal to decide whether to handover (when the UE is in RRC_CONNECTED state) or cell re-selection (when the UE is in RRC_IDLE state).
- the subcarrier spacing (SCS) of the Synchronization Signal and PBCH block (SSB) is determined by the frequency band range: below 6 GHz (i.e., FR1) it supports 15 or 30 kHz SCS, and above 6 GHz (i.e., FR2) it supports 120 or 240 kHz SCS.
- N_ID_(2) has a value range of ⁇ 0,1,2 ⁇
- N_ID_(1) has a value range of ⁇ 0,1,...335 ⁇
- n5/n41/n66/n90 support two SSB SCSs (15kHz and 30kHz). The terminal needs to perform blind detection using the two SCSs separately to determine the SSB SCS of the cell.
- FR2 n257/258/259/260/261 all support 120/240kHz. In this case, it is necessary to try each one to determine the SCS.
- a channel raster can be understood as the selectable location of the center frequency of a carrier wave.
- the protocol first defines a global frequency raster.
- the channel raster is based on the global frequency raster, with limitations on its range and step size according to the operating band.
- the global frequency raster is defined as a set of RF reference frequencies (FREFs), ranging from 0-100 GHz, primarily used to identify the frequency domain locations of RF channels, SSBs, or other resources.
- FREFs RF reference frequencies
- the NR Absolute Radio Frequency Channel Number encodes the frequency domain range of the RF reference frequency.
- the values from 0 to 100 GHz are FR1[0...2016666] and FR2[2016667...3279165].
- the relationship between NR-ARFCN and the RF reference frequency FREF is shown in equation (1).
- the ARFCN frequency point number corresponds to the channel grid.
- the spacing density of the channel grid differs in different NR bands.
- F REF F REF-Offs + ⁇ F Global (N REF –N REF-Offs ) (1)
- the NR-ARFCN parameters of the global frequency grating are shown in Table 1 below.
- NR introduces the concept of a synchronization raster, where synchronization signals are placed according to the synchronization grid.
- GSCN frequency point numbers correspond to synchronization grids.
- GSCNs define the frequency band in the 0-100GHz range, and each GSCN corresponds to a detection frequency point of an SSB.
- the terminal can only blindly detect SSBs at GSCN locations.
- 0-100GHz corresponds to 0-26639 GSCNs.
- GSCNs are used to describe the synchronization channels in various frequency bands. Synchronization grids are subsets of GSCNs, and the frequency spacing of the synchronization grids differs across bands. In band n41, the frequency spacing of the synchronization grids is 3 GSCNs. In band n79, the frequency spacing of the synchronization grids is 16 GSCNs.
- the PSS or SSS and PBCH are always bound, hence also called SSB.
- the time and frequency domain positions of the SSB are no longer fixed but flexible. In the frequency domain, the SSB is no longer fixed in the middle of the frequency band; in the time domain, the position and number of SSBs transmitted can change. Therefore, in NR, complete synchronization of frequency and time domain resources cannot be achieved simply by demodulating the PSS or SSS signal; demodulation of the PBCH must be performed to ultimately achieve synchronization of time and frequency resources.
- a cell-defining SSB (CD-SSB) is defined as an SSB associated with SIB1 (Remaining Minimum System Information (RMSI)).
- SIB1 defines the scheduling information for other SIBs and contains information for initial terminal access.
- the frequency location of the CD-SSB must be on the system synchronization grid.
- NCD-SSB A non-cell-defining SSB (NCD-SSB) is defined accordingly as an SSB not associated with SIB1.
- NCD-SSBs can be used for secondary cell synchronization or as measurement signals configured for terminals.
- An NCD-SSB does not necessarily reside on the system synchronization grid. If an NCD-SSB is located on the system synchronization grid, it can indicate the GSCN of the NCD-SSB through the information it carries.
- a terminal When a terminal detects an SSB during cell search, it first needs to determine whether the SSB is a CD-SSB or an NCD-SSB. This is determined by checking if the subcarrier offset kSSB , provided by the SSB's PBCH and representing the subcarrier offset between the SSB and the common resource block grid, is within the valid subcarrier offset range.
- the valid subcarrier offset range includes 0-23 subcarriers and 0-11 subcarriers, represented by 5 bits and 4 bits respectively, corresponding to frequency ranges FR1 and FR2. If the value of kSSB is within the valid subcarrier offset range, the SSB is a CD-SSB; otherwise, it is an NCD-SSB.
- kSSB > 23, or in FR2 if kSSB > 11, it indicates that the SSB does not exist in the Type 0 Common Search Space (CSS), meaning the current SSB is not associated with SIB1.
- CSS Common Search Space
- these kSSBs can also be used as indexes, combined with the RMSIPDCCH Config (i.e., the MIB's PDCCH Config SIB1), to (indirectly) indicate the GSCN of the next SSB.
- the SSB or synchronization signal in the embodiments of this application may also be called any module that includes at least one of a synchronization signal, a broadcast signal, a broadcast channel (PBCH), a downlink broadcast channel for other system messages, or a control channel thereof.
- PBCH broadcast channel
- Type0-PDCCH Public Search Space Collection (CSS set):
- This search space set is used to listen to SIB1 system messages, corresponding to the DCI scrambled with SI-RNTI in the primary cell of the Master Cell Group (MCG). It is configured in the signaling MIB by IE:pdcch-ConfigSIB1, or in the signaling PDCCH-ConfigCommon by IE:searchSpaceZero, or in the signaling PDCCH-ConfigCommon by IE:searchSpaceZero or searchSpaceSIB1.
- Pattern 1 (as shown in Figure 2a) is time-division multiplexing (the frequency domain range of CORESET 0 includes SSB), and Pattern 2 (as shown in Figure 2b) and Pattern 3 (as shown in Figure 2c) are frequency-division multiplexing (CORESET 0 and SSB are in the same system frame).
- Pattern 2 and Pattern 3 are frequency-division multiplexing (CORESET 0 and SSB are in the same system frame).
- the difference between Pattern 2 and Pattern 3 is that, in the time domain, CORESET 0 in Pattern 2 is positioned slightly earlier than SSB.
- the terminal After the terminal decodes the SSB, it can determine the scheduling information of SIB1 for blind detection at the specific time-frequency resource location.
- the terminal After detecting PBCH, the terminal has completed downlink synchronization. Before performing uplink synchronization, the terminal needs to receive SIB1 to obtain configuration information related to uplink synchronization.
- SIB1 is transmitted in PDSCH and scheduled via PDCCH, and the resource allocation range of PDSCH is within the frequency range of the initial BWP:
- SIB1 PDCCH time-frequency domain resource allocation (the DCI information for scheduling SIB1 is carried in CORESET0).
- the PDCCH of SIB1 is mapped within the common search space (CCS) of type 0-PDCCH;
- the CSS of Type 0-PDCCH is mapped in CORESET 0, and the frequency range of CORESET 0 is exactly the same as that of the initial BWP;
- the lower 4 bits of the signaling 'pdcch-ConfigSIB1' carried in the PBCH indicate the configuration of type 0-PDCCH CSS; the higher 4 bits indicate the configuration of CORESET 0.
- the standard PDSCH uses the Time Domain Resource Assignment (TDRA) table configured by Radio Resource Control (RRC), with the PDCCCH instructing the index in the table to allocate time-domain resources.
- TDRA Time Domain Resource Assignment
- RRC Radio Resource Control
- the three reuse modes of CORESET0 and SSB correspond to three default TDRA tables
- SIB1 allocates frequency domain resources within the initial access bandwidth range, using resource allocation type 1.
- AI can be represented as machine learning (ML).
- ML machine learning
- AI has been widely applied in various fields. Integrating artificial intelligence into wireless communication networks to significantly improve technical indicators such as throughput, latency, and user capacity is an important task for future wireless communication networks.
- AI modules can be implemented in various ways, such as neural networks, decision trees, support vector machines, and Bayesian classifiers.
- the neural network consists of neurons, as shown in Figure 3.
- a1, a2, ..., aK are the inputs
- w is the weight (multiplicative coefficient)
- b is the bias (additive coefficient)
- ⁇ ( .) is the activation function.
- Common activation functions include Sigmoid, tanh, and Rectified Linear Unit (ReLU).
- the expression z a1w1 + ... + akwk + ... + aKwK + b.
- optimization algorithms are a class of algorithms that help us minimize or maximize an objective function, also known as 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 construct a neural network model f(.). With the model, based on the input x, we can obtain the predicted output f(x), and calculate the difference between the predicted value and the true value (f(x) - Y), which is the loss function. Our goal is to find suitable values W and b that minimize the value of the aforementioned loss function. The smaller the loss value, the closer our model is to the reality.
- BP error back propagation
- the learning process consists of two parts: forward propagation of the signal and backward propagation of the error.
- forward propagation the input sample is introduced from the input layer, processed layer by layer by the hidden layers, and then propagated to the output layer. If the actual output of the output layer does not match the expected output, the process transitions to error back propagation.
- Error back propagation involves propagating the output error back to the input layer layer by layer through the hidden layers, distributing the error to all units in each layer, thus obtaining the error signal of each unit.
- This error signal serves as the basis for adjusting the weights of each unit.
- This process of adjusting the weights through forward and backward propagation is cyclical. This 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 predetermined number of learning iterations is reached.
- these optimization algorithms calculate the gradient by taking the derivative or partial derivative of the current neuron with respect to the error or loss obtained from the loss function, and then adding the effects of the learning rate, previous gradients, derivatives or partial derivatives, etc., and then pass the gradient to the previous layer.
- the AI algorithms and models selected vary depending on the type of problem. Based on currently published articles and research findings, the main method for improving 5G network performance using AI is to enhance or replace existing algorithms or processing modules with neural network-based algorithms and AI models. In specific scenarios, neural network-based algorithms and AI models can achieve better performance than deterministic algorithms. Commonly used neural networks include deep neural networks, convolutional neural networks, and recurrent neural networks. Existing AI tools can be used to build, train, and validate neural networks.
- Fine-tuning can be considered a training process that uses the parameters of the pre-trained neural network as initialization.
- parameters of some layers can be frozen; generally, layers closer to the input are frozen, while layers closer to the output are activated. This ensures that the network can still converge.
- the smaller the amount of data during the fine-tuning phase the more layers should be frozen, with only a small number of layers near the output being fine-tuned.
- neural network-based wireless communication systems offer two solutions. The first is to train different neural networks under different transmission conditions, obtaining multiple sets of neural network parameters, and switching the parameters as the actual environment changes. The second is to train a common neural network based on mixed data, where the network parameters do not change with the environment. Both approaches have advantages and disadvantages: the first approach performs excellently under different transmission conditions, but requires storing multiple network parameters and switching them on demand (incurring signaling overhead and frequent switching issues); the second approach only requires storing one set of neural network parameters without switching, but cannot achieve optimal performance under every transmission condition. The construction method of the mixed dataset affects the performance of the second approach.
- a label typically refers to the identifier or annotation of the true category or target value of a data sample. Labels are used to represent the information that the model should learn and predict, for example:
- labels indicate which category a data sample belongs to. For example, in image classification, each image sample has a label that indicates the category of the object or scene contained in the image, such as "dog" or "cat".
- Labels in object detection In object detection tasks, labels typically include the object's location information (bounding box) and category information. Each label identifies a target object in an image, including its location and category. Labels in regression tasks:
- labels typically represent the continuous or real-valued objective to be predicted.
- the label could be the actual selling price of a house.
- Labels in sequence labeling In natural language processing, labels in sequence labeling tasks are often used for tasks such as part-of-speech tagging and named entity recognition. Labels are used to represent the attributes or categories of each word or character in a text sequence.
- Labels are a crucial component in supervised learning tasks, used to train machine learning models. Models learn patterns and regularities by comparing themselves to true labels in order to make predictions or classifications on unseen data. The quality and accuracy of the labels are critical to the model's performance.
- AI model lifecycle management includes multiple AI functional modules: model training module, model management module, model inference module, model monitoring module, and model update module.
- the model training module is used to perform AI model training, validation, and testing, and can generate model performance metrics that can be used as part of the model testing process. If necessary, this function is also responsible for data preparation based on the training data provided by the data collection function (e.g., data preprocessing and cleaning, formatting, and transformation).
- the model management module is used to supervise the operation of AI models or the deployment of AI functions (such as model selection, de)activation, switching, or rollback), and to provide feedback on model monitoring performance.
- This module is also responsible for making decisions based on data received from the data collection module and the model inference module to ensure correct inference operations.
- Management instructions Information input from the model management module to the model inference module. This information may include selecting, (deactivating), activating, or switching models using AI models or AI-based functions, or reverting to non-AI operations (i.e., operations independent of the inference process).
- Model transfer request Used to request a model from the model storage function
- the model training module takes in the required information, such as for model (re)training or updating purposes.
- the model inference module is used to provide the output of the applied AI model using the data (i.e., inference data) provided by the data collection function as input. If necessary, the model inference module is also responsible for data preparation based on the inference data provided by the data collection function (e.g., data preprocessing and cleaning, formatting and transformation).
- inference output data used by the model management module to monitor the performance of AI models or AI functions.
- the AI model may be referred to as an AI unit, machine learning (ML) model, ML unit, AI structure, AI function, AI characteristic, neural network, neural network function, neural network functionality, etc.
- the AI model may refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc. related to AI.
- the AI model may be a processing method, algorithm, function, module, or unit for a specific dataset.
- the AI model may be a processing method, algorithm, function, module, or unit running on AI or ML-related hardware such as a graphics processing unit (GPU), neural processing unit (NPU), tensor processing unit (TPU), or application-specific integrated circuit (ASIC). No further limitations are made here.
- a terminal after a terminal powers on, it performs a cell search, typically by first reading the SIM card information and then using prior information stored in the SIM card (such as previously stored cell and frequency information) to find a suitable cell to camp on.
- prior information stored in the SIM card such as previously stored cell and frequency information
- the prior information stored in the terminal may be outdated. If the terminal cannot find a suitable cell to access using the previously stored prior information after powering on, it will perform a frequency scan on the frequency bands supported by the SIM card to perform an initial cell selection. This process takes a considerable amount of time and consumes a significant amount of power.
- networks deployed by operators in different geographical locations may have different GSCNs.
- This application provides a cell camping processing method to enable rapid initial cell search and selection at any time and location when the terminal powers on, reducing initial search time and saving terminal power.
- an embodiment of this application provides a cell dwell processing method, as shown in Figure 4.
- the cell dwell processing method provided in this embodiment includes:
- Step 401 The first device obtains the target result based on the first artificial intelligence (AI) model;
- AI artificial intelligence
- the first device is a terminal, a network-side device, or a server
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, wherein the first cell is a cell that can be camped.
- the target result can be understood or replaced as the inference result of the first AI model, or the target result can be determined based on the inference result of the first AI model.
- the output of the first AI model includes whether to camp on a certain cell, thereby determining the information of the first cell based on the output result. Since the target result is obtained through the first AI model, and then the cell search and selection are performed based on the first AI result to achieve cell camping of the terminal, cell search in the full frequency band supported by the SIM card can be avoided, thus reducing the latency of the terminal in cell camping.
- the information of the first cell may include PCI.
- the first cell can be understood as a candidate cell.
- cell search can be performed based on the target result, thereby eliminating the need to perform cell search across the entire frequency band supported by the SIM card.
- the terminal when the target result includes information about the first cell, the terminal can directly select the first cell to camp on or perform cell search only in the first cell, thereby saving cell search time.
- the beam transmission direction of the above-mentioned synchronization signal transmission can be understood or replaced with the strongest beam transmission direction of the synchronization signal transmission.
- This application embodiment obtains a target result through a first device based on a first artificial intelligence (AI) model.
- the first device is a terminal, network-side device, or server.
- the target result is used for cell camping and includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about a first cell, which is a cell suitable for camping.
- the terminal can perform initial cell search and selection based on the time-frequency domain location, beam transmission direction, and first cell information in the target result, thereby avoiding frequency scanning and network searching across the entire frequency band supported by the terminal. Therefore, this application embodiment shortens the initial cell search and selection time, thereby reducing the latency of cell camping and simultaneously reducing the energy consumption of initial cell search and selection.
- the first device obtains the target result based on a first artificial intelligence (AI) model, including:
- the first device inputs the first information into the first AI model to obtain the target result
- the first information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the aforementioned time information can be specific to a particular point in time, such as 13:25:38. It can also be a time range, such as 13:00 to 14:00, morning or afternoon, day or night.
- the time information may be obtained through other radio access technologies (RATs), such as Bluetooth, Wi-Fi, 3G, 4G, or LTE.
- RATs radio access technologies
- the aforementioned location information may be specific location coordinates, such as GPS coordinates; or approximate range information of the terminal, such as which street or country it is located in; or location information of the terminal relative to the cell it is staying in or the cell it is accessing, such as being due east of the cell it is staying in.
- the direction of movement of the aforementioned terminal can be an absolute direction, such as 40 degrees east of south; or it can be a relative direction, such as the direction relative to the cell it is staying in or the cell it is accessing.
- the environmental information mentioned above may include weather information, etc.
- the aforementioned network scenario information may include indoor hotspots (inH), urban macrocells (Uma), rural macrocells (RMa), etc., or may include homogeneous or heterogeneous networks, i.e., whether there is overlapping coverage.
- the above frequency domain features may include at least one of the following:
- sync raster or GSCN frequency points where synchronization signals may exist such as frequency domain locations that may lead to synchronization signals as defined in the protocol;
- the number of frequency domain resource blocks (RBs) or subcarriers of the synchronization signal is the number of frequency domain resource blocks (RBs) or subcarriers of the synchronization signal.
- the aforementioned time-domain features may include the number of time-domain symbols of the synchronization signal, or the time window length for time-domain correlation detection of the synchronization signal.
- the above network types may include terrestrial networks (TN), non-terrestrial networks (NTN), and cell-free networks.
- TN terrestrial networks
- NTN non-terrestrial networks
- cell-free networks may include terrestrial networks (TN), non-terrestrial networks (NTN), and cell-free networks.
- the aforementioned emergencies include, but are not limited to, concerts and earthquakes.
- the inference of the first AI model can be performed on a network-side device, on a terminal, or on a server.
- the method further includes:
- the first device sends the target result to the terminal.
- the network-side device or server can perform model inference, thereby reducing the requirements for the terminal and expanding the application scope of using AI models for AI-based cell search and selection.
- the method further includes:
- the first device When the first device is a terminal, the first device activates the first AI model based on a first activation condition, the first activation condition including at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the timer used to activate the first AI model times out; the number of failed synchronization signal block (SSB) detections is greater than or equal to a first threshold; and the first AI model needs to be activated based on at least one of the second pieces of information, where the second pieces of information are at least a portion of the input information of the first AI model.
- SSB synchronization signal block
- the first device activates the first AI model based on a second activation condition, wherein the second activation condition includes at least one of the following:
- the terminal receives target indication information, which indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the number of synchronization signal block (SSB) detection failures is greater than or equal to a first threshold; and it is determined that the first AI model needs to be activated based on at least one of the second pieces of information, wherein the second pieces of information are at least a portion of the input information of the first AI model.
- target indication information which indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on a second preset time period; a cell reselection
- the timer used to activate the first AI model timed out
- the need to activate the first AI model is determined based on at least one of the third pieces of information, wherein the third pieces of information are at least a portion of the input information of the first AI model.
- triggering the activation of the first AI model can be understood or replaced as using the first AI model to perform AI-based cell search and selection.
- the activation of the first AI model can be triggered when the terminal is powered on, that is, the use of the first AI model to perform AI-based cell search and selection is triggered when the terminal is powered on.
- the activation of the first AI model can be triggered when the initial cell search operation is performed, that is, the first AI model is used to perform AI-based cell search and selection when the initial cell search operation is performed.
- the terminal when the terminal is the inference device of the first AI model, the terminal can activate the first AI model if the first activation condition is met.
- the terminal can send the aforementioned target indication information to the network-side device or server to instruct the network-side device or server to activate the first AI model if the first activation condition is met.
- the activation of the first AI model can be triggered when performing the initial cell selection operation.
- an initial search can be performed for a first preset duration by searching the entire frequency band supported by the terminal.
- the first AI model can be activated if the terminal fails to camp on a second cell within a second preset time period.
- the second cell can be understood as or replaced by a new cell or a newly camped cell, i.e., any cell other than the cell where the terminal previously camped.
- the first AI model is activated if the terminal fails to successfully camp on a new cell after the second preset time period.
- the activation of the first AI model can be triggered when performing a cell reselection operation.
- the activation of the first AI model can be triggered when performing a cell handover operation.
- the activation of the first AI model can be triggered when the SSB detection fails more than a certain number of times.
- At least one of the aforementioned second and third information can be the aforementioned first information or a subset of the first information.
- determining that the first AI model needs to be activated based on at least one of the second information can be understood as at least one of the second information indicating that the activation of the first AI model needs to be triggered.
- the eighth information includes the number of synchronization signal detection failures within a certain time period. When this number exceeds a certain threshold, it indicates that the range of synchronization signal detection may be excessively large. In this case, it is necessary to predict the possible locations of synchronization signal resources, triggering the activation of the first AI model for AI-based cell search and selection.
- the method when the first device is a terminal, the method further includes at least one of the following:
- the first device updates the first AI model based on the fourth information, or the first device sends the fourth information to the second device, the fourth information being used to update the first AI model, wherein the fourth information includes at least one of the target result and the terminal's status information;
- the first device performs the first operation
- the first operation includes at least one of the following: reverting to a cell search method that supports all frequency bands on the terminal; triggering a switch of the AI model; triggering retraining of the first AI model; and triggering supervision of the first AI model.
- the first device when the first device supports updating the first AI model (e.g., the first device is a node for training the first AI model), the first device can update the first AI model based on the fourth information. Updating the first AI model can be understood or replaced as fine-tuning the first AI model.
- the first device can send the fourth information to the second device, which will then update the first AI model.
- the updated first AI model can then be obtained from the second device.
- switching AI models may include changing the input information or changing the AI algorithm.
- the first device performing the first operation may be periodically triggered, conditionally triggered, or autonomously triggered by the first device.
- the first device performing the first operation includes:
- the first device performs the first operation
- the first condition includes at least one of the following:
- the time taken to perform inference using the first AI model exceeds the second preset time.
- the first AI model failed to find a suitable cell to reside in
- the inference was not successfully completed using the first AI model.
- the first AI model satisfies at least one of the following:
- the complexity of the AI model is lower than or equal to the second threshold
- the inference latency of the AI model is less than or equal to the third preset duration
- the AI model's reasoning success rate is greater than or equal to the third threshold
- the reliability of the AI model's inference results is greater than or equal to the fourth threshold.
- different AI model complexity index requirements can be defined for base stations or terminals of different types or with different capabilities. For example, for ordinary terminals, the complexity of the AI model used cannot exceed a specific value.
- the inference latency of the AI model being less than or equal to the third preset duration can be understood as the duration for which the first AI model is used to infer the time-frequency position or beam direction of the synchronization signal transmission, or to infer the duration for which the first cell can be camped, cannot exceed the third preset duration.
- the inference success rate of the AI model being greater than or equal to the third threshold can be understood as the success rate of using the first AI model to infer the time-frequency location or beam direction of the synchronization signal transmission, or to infer the first cell that can be camped, not exceeding the third preset duration.
- the reliability of the inference results of the AI model is greater than or equal to a fourth threshold.
- Reliability can be understood as signal strength information such as RSRP, RSRQ, or RSSI of successfully camped cells.
- the RSRP requirement here differs from the RSRP index requirement in the S criterion.
- the measured signal strength information is greater than or equal to or less than or equal to a specific value.
- the first AI model is independently trained by the terminal, network-side device, or server, or the first AI model is jointly trained by at least two of the terminal, network-side device, and server.
- the network-side device may include at least one of a base station and core network equipment (such as core network equipment specifically used for model training).
- the network-side device may be a base station or network-side device associated with the cell where the terminal most recently camped or accessed.
- the network-side device may be a base station or network-side device associated with the cell that sent the RRC release message.
- the first AI model is jointly trained by at least two of the terminal, network-side device, and server.
- at least one of the following may be included:
- the terminal reports the output of the model training to the network-side device or server.
- the network-side device or server uses the information reported by the terminal (i.e., the output of the terminal model training) as one of the inputs for its own model training.
- the network-side device sends the output of the model training to the terminal or server.
- the terminal or server uses the information sent by the network-side device (i.e., the output of the network-side device's model training) as one of the inputs for its own model training.
- At least one of the terminal, network-side device, and server performs offline model training, and then the terminal, network-side device, and server fine-tune it in a real network.
- At least part of the input information for the model training is sent to the terminal by the network-side device.
- the signals or signaling sent by the at least part of the input information include at least one of the following: MAC CE; RRC message; NAS message; user plane data; DCI information; System Information Block (SIB); Layer 1 signaling of Physical Downlink Control Channel (PDCCH); Physical Downlink Shared Channel (PDSCH) information; MSG 2 information; MSG 4 information; MSG B information.
- At least part of the input information for the model training is sent by the terminal to the network-side device.
- the signals or signaling sent by the at least part of the input information include at least one of the following: MAC CE; RRC message; NAS message; user plane data; MSG 1 information; MSG A information; MSG 3 information; information of the Physical Uplink Control Channel (PUCCH); information of the Physical Uplink Shared Channel (PUSCH); information of the Physical Random Access Channel (PRACH); SRS or other uplink reference signals, such as WUS.
- At least part of the input information for the model training is sent to the server by the terminal or network-side device.
- This at least part of the input information can be indicated by Over-The-Top (OTT) messages.
- OTT messages can be provided by third-party service providers, third-party servers, or the Internet.
- the method of obtaining input information for model training may include: triggering at least one reporting (e.g., the terminal sending to the network-side device) or sending (e.g., the network-side device sending to the terminal) of input information for model training after a period of time when the terminal is camped in a cell, or initially selects a cell, or reselects a cell, or enters an RRC connected state or an inactive state.
- triggering at least one reporting e.g., the terminal sending to the network-side device
- sending e.g., the network-side device sending to the terminal
- the reporting or distribution of input information for at least one model training iteration can be periodic or semi-static.
- the aforementioned time period can be a fixed duration configured by the network-side device.
- the network-side device or server can send the trained first AI model to the terminal after training.
- the terminal can perform AI model inference, reducing subsequent signaling interactions and latency for cell search and selection.
- model inference can be performed on the network-side device or server, and the resulting target value can be sent to the terminal, thus reducing the demands on terminal capabilities.
- the method before the first device receives the first AI model from the second device, the method further includes:
- the first device sends fifth information to the second device, the fifth information being used by the second device to train the first AI model;
- the fifth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the first device can be a terminal or a network-side device.
- the terminal sends the fifth information to the network-side device or the server, or the network-side device sends the fifth information to the terminal or the server, in order to train the first AI model.
- the second device After the second device completes the training of the first AI model, it sends the trained first AI model to the first device, and the first device performs inference based on the first AI model to obtain the target result.
- the triggering condition for the first device to send the fifth information to the second device may include at least one of the following:
- Terminal resides in the community
- the terminal selects a cell for the first time
- the terminal enters RRC connection state
- the terminal remains in an inactive state for a preset duration.
- the method further includes:
- the first device receives the first AI model from the second device
- the second device is a terminal, a network-side device, or a server.
- the first device may be only an inference device for the first AI model, or it may be both an inference device for the first AI model and a device for jointly training the first AI model.
- the method before the first device receives the first AI model from the second device, the method further includes:
- the first device inputs the sixth information into the second AI model to obtain the first training result, and the first training result is used by the second device to train the first AI model;
- the first device sends the first training result to the second device
- the sixth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the first device and the second device jointly train the first AI model, and the training of the first AI model is finally completed on the second device.
- the second device can send the trained first AI model to the first device, and the first device can perform inference based on the first AI model to obtain the target result.
- the triggering condition for obtaining the sixth information may include at least one of the following:
- Terminal resides in the community
- the terminal selects a cell for the first time
- the terminal enters RRC connection state
- the terminal remains in an inactive state for a preset duration.
- the method further includes:
- the first device When the first device is the server, the first device obtains the sixth information from at least one of the terminal and the network-side device;
- the first device obtains at least a portion of the sixth information from the terminal;
- the first device obtains at least a portion of the sixth information from the network-side device.
- the method further includes any one of the following:
- the first device trains the first AI model based on the seventh information to obtain the first AI model
- the first device will input the second training result received from the second device into the third AI model for AI training to obtain the first AI model;
- the seventh piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the first device can be both an inference device for the first AI model and a training device for the first AI model. Specifically, when the first device trains the first AI model independently, it can input the seventh information into the AI model to be trained to obtain the first AI model. When the first device and the second device jointly train the first AI model, the first device can receive the second training result from the second device and use the second training result as one of the inputs to the third AI model to train the model and obtain the first AI model.
- the input information used for training the first AI model can be the input information corresponding to the most recent N times the terminal successfully camped in the cell, or the input information corresponding to the most recent N times a synchronization signal (such as CD-SSB) was detected.
- the input information used for inference of the first AI model can be the input information used for training the first AI model, or a subset of the input information used for training the first AI model.
- the input for training the first AI model further includes a target label, which includes at least one of the following:
- the terminal detected a synchronization signal at a preset frequency domain position
- the terminal detected a synchronization signal at a preset time domain location
- the terminal detected a synchronization signal in the preset beam direction
- the terminal successfully camped on the preset cell
- the aforementioned signal quality may include at least one of the following: RSSI, PSS or SSS detected signal strength, channel estimation SNR, SSB-RSRP, L1 RSRP and L3 RSRP.
- the signal quality of the camped cell measured by the terminal may include at least one of the following:
- the terminal measures the signal quality of the cell it is camping at a preset frequency domain location
- the terminal measures the signal quality of the cell it is camping at a preset time domain location
- the terminal measures the signal quality of the cell it is camping in the preset beam direction.
- the method before the first device obtains the target result based on the first artificial intelligence (AI) model, the method further includes:
- the first device periodically triggers the training of the first AI model based on the first configuration information
- the first configuration information includes at least one of the following: the starting point of periodic model training; the interval of periodic model training; the number of model training sessions within one cycle; and the training duration within one cycle.
- the method before the first device obtains the target result based on the first artificial intelligence (AI) model, the method further includes:
- the first device triggers the training of the first AI model using a semi-static triggering method
- the semi-static triggering method includes at least one of the following:
- Semi-static triggering is performed based on the second configuration information
- the second configuration information satisfies at least one of the following: triggering the transmission and activation of the second configuration information based on a target event; configuring the second configuration information through Radio Resource Control (RRC).
- RRC Radio Resource Control
- the method before the first device obtains the target result based on the first artificial intelligence (AI) model, the method further includes:
- the first device triggers the training of the first AI model based on the target event
- the target event includes at least one of the following:
- NCD-SSB non-cell-defined synchronization signal block
- the terminal moves to the edge of the cell or a preset location
- the terminal's moving speed is higher than or equal to a first threshold
- the terminal's moving speed is lower than or equal to the second threshold
- the number of terminals currently connected or registered in the community is higher than or equal to the third threshold
- the number of terminals currently connected to or registered in the community is less than or equal to the fourth threshold
- the number of consecutive inference failures using the first AI model reached the fifth threshold.
- the number of inference failures using the first AI model reached the sixth threshold.
- the terminal was reselected to the second cell
- the tracking area of the terminal changes
- the external environment of the terminal changes
- the terminal moved to the second cell
- the terminal was moved to a new tracking area
- the terminal was moved to a new geographical location
- the change in the terminal's moving speed is greater than or equal to the seventh threshold
- M is an integer greater than 1
- the number of failed detections of the synchronization signal block has reached the eighth threshold.
- the training of the first AI model can be understood as the initial training or retraining of the first AI model.
- the aforementioned target event can be understood or replaced as target condition.
- the aforementioned change in the external environment can be understood as the environment determined by obtaining environmental change information through the terminal's sensors.
- this may include, but is not limited to, the following environmental changes: being in an urban or rural area, being indoors or outdoors, being in a high-speed or low-speed motion environment, etc.
- the term "the terminal moves to the second cell” can be understood or replaced as: the terminal moves to a new cell, or the terminal moves to a cell other than the serving cell it last accessed.
- the terminal moving to a new tracking area can be understood as the terminal moving to a tracking area other than the tracking area of the serving cell it last accessed.
- terminal moving to a new geographical location can be understood as the terminal moving to a geographical location other than the geographical location of the serving cell it last accessed. For example, if the terminal is powered off in location A and then powered on after moving to location B, it can be considered that the terminal has moved to a new geographical location, with location B being the new geographical location.
- Locations A and B can be different cities, different provinces, or different countries.
- the change in the terminal's moving speed being greater than or equal to the seventh threshold can be understood as: the terminal's moving speed changing significantly in a short period of time, for example, decreasing from 250 km/h to 3 km/h, or increasing from 3 km/h to 250 km/h.
- the method further includes:
- the first device determines that the training of the first AI model is complete based on the eighth piece of information
- the eighth piece of information includes at least one of the following: the received signal strength indication of the synchronization signal; the signal strength of the detected primary or secondary synchronization signal; the channel estimation signal-to-noise ratio of the synchronization signal; the reference signal received power (RSRP) of the synchronization signal block; the RSRP of layer 1; the RSRP of layer 3; the probability that the terminal successfully detects the synchronization signal; the duration of the synchronization signal detected by the terminal; the probability that the terminal successfully camps; and the probability that the terminal successfully synchronizes.
- the received signal strength indication of the synchronization signal the signal strength of the detected primary or secondary synchronization signal
- the channel estimation signal-to-noise ratio of the synchronization signal the reference signal received power (RSRP) of the synchronization signal block
- the RSRP of layer 1 the RSRP of layer 3
- the probability that the terminal successfully detects the synchronization signal the duration of the synchronization signal detected by the terminal
- the probability that the terminal successfully camps and the probability that
- the training of the first AI model can be determined to be complete if some information in the eighth information is greater than or equal to the corresponding preset threshold, or if some information in the eighth information is less than or equal to the corresponding preset threshold.
- the first AI model training is determined to be complete when the probability of the terminal successfully detecting the synchronization signal is greater than or equal to X%, where X is a specific threshold value.
- the probability of the terminal successfully detecting the synchronization signal being greater than or equal to X% can be understood or replaced as the probability of the terminal successfully detecting the synchronization signal being greater than or equal to X% in at least one of a specific frequency domain location, a specific time domain location, and a specific beam direction.
- the terminal determines that the first AI model training is complete when the duration of the synchronization signal detected by the terminal is less than or equal to M, where M is a specific threshold value.
- M is a specific threshold value.
- the duration of the synchronization signal detected by the terminal is less than or equal to M can be understood or replaced as the terminal detecting the synchronization signal for a duration less than or equal to M at least in at least one of a specific frequency domain location, a specific time domain location, and a specific beam direction.
- the eighth information described above further includes at least one of the following:
- the first AI model is determined to be trained successfully when the loss function meets a predefined requirement or value; for example, the training error is less than a predefined threshold value.
- the loss function may include at least one of the following:
- the type of terminal needs to be considered, as different types of terminals may have different AI model training capabilities.
- the input or label information used for model training varies depending on the type of terminal. For example, less input information should be used for model training on terminal devices with weaker capabilities.
- model training can be performed only on the network side, or only a small part of the joint model training can be performed on the terminal side (for example, model training involving user privacy data can be performed on the terminal side).
- Different types of devices require different AI models for training. For example, less capable devices may not be able to apply overly complex AI models.
- the method further includes:
- the first device supervises the first AI model based on the third configuration information
- the third configuration information includes at least one of the following:
- the metrics for model supervision include at least one of the following: error class information or accuracy information between predicted and true values; communication system performance; and model-related information of the first AI model.
- the AI model identifier can be understood or replaced with at least one of the following:
- the detection window information mentioned above includes at least one of the following: the duration of the detection window and the number of samples tested.
- the error information mentioned above may include, but is not limited to, at least one of the following: mean error, mean square error, normalized mean square error, mean absolute error, cross-entropy loss, and root mean square error.
- the aforementioned precision information may include, but is not limited to, at least one of the following: similarity, cosine similarity, correlation, correlation coefficient, and area under the curve (AUC) score.
- the aforementioned communication system performance may include, but is not limited to, at least one of the following: cell search latency, cell dwell success rate, and timing error.
- the statistics for cell search latency, cell dwell success rate, and timing error are obtained within a monitoring window.
- the model-related information of the first AI model may include, but is not limited to, at least one of the following: the runtime of the first AI model, the CPU usage of the first AI model, and the memory space usage of the first AI model.
- the triggering condition for model supervision is determined based on at least one of the following: the metrics of model supervision; the inference results of the first AI model; and the metrics of model inference.
- supervision of the first AI model can be triggered when the model supervision metrics do not meet the requirements, or after the model supervision metrics have not met the requirements for a period of time.
- supervision of the first AI model can be triggered when the inference result label of the first AI model fails to meet the requirements, or after a period of time when the inference result label of the first AI model fails to meet the requirements.
- the inference result label of the first AI model can be considered to fail to meet the requirements if at least one of the following conditions is met:
- the terminal After the terminal used the first AI model for inference for more than M time (i.e. the fifth preset time), it still failed to find a suitable cell to stay in.
- the terminal failed to complete AI inference using the first AI model.
- supervision of the first AI model is triggered when at least one of the metrics for model inference of the first AI model fails to meet the requirements, or after a period of time when at least one of the metrics for model inference of the first AI model fails to meet the requirements.
- supervision of the first AI model is triggered when the RSRP of the terminal successfully camps on the cell fails to meet the requirements.
- the method further includes:
- the first device and the second device transmit target capability information, which includes at least one of the following:
- the network-side device supports the ability to provide auxiliary information for AI inference based on the first AI model
- the method further includes:
- the first device determines the target capability information corresponding to the second device based on at least one of the following:
- the equipment type of the second device is the equipment type of the second device
- the network type of the second device is the network type of the second device
- the reference signal sent by the second device is the reference signal sent by the second device.
- Interface messages between the first device and the second device are Interface messages between the first device and the second device.
- the determination of target capability information can depend on the device type. For example, if the second device is a terminal, different terminal types introduce different target capability information. In this case, the relevant capabilities of the terminal can be implicitly indicated by reporting the terminal type.
- the determination of target capability information can depend on the network type.
- different network types NTN or TN introduce different target capability information.
- the relevant capabilities of the second device can be implicitly indicated by the transmission network type.
- target capability information may be indicated by one or more reference signals, such as by indicating the target capability information of a second device via PRACH resources.
- the terminal can indicate target capability information, such as physical layer control information, like the UCI reported to the network-side device for the terminal, through uplink control information (UCI).
- target capability information such as physical layer control information, like the UCI reported to the network-side device for the terminal, through uplink control information (UCI).
- UCI uplink control information
- the interface messages between the first device and the second device may include at least one of the following:
- the interface messages between the terminal and the server can specifically be specific interface messages between the terminal and the server. These specific interface messages can be related to information about a specific AI model or related to information about all AI models.
- the interface message between the terminal and the network-side device can specifically be a specific interface message between the terminal and the network-side device.
- This specific interface message can be related to a specific AI model or related to all AI models.
- the interface messages between the server and the network-side devices can specifically be interface messages between the server and the network-side devices. These specific interface messages can be related to specific AI models or related to all AI models.
- this application embodiment also provides a cell dwell processing method, as shown in Figure 5, the cell dwell processing method includes:
- Step 501 the second device performs at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a terminal, a network-side device, or a server
- the second device is a terminal, a network-side device, or a server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information of the first cell, wherein the first cell is a cell that can be camped.
- the second device trains the first AI model to obtain the first AI model, including:
- the second device trains the first AI model based on the fifth information to obtain the first AI model
- the fifth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the method before obtaining the first AI model based on the training of the first AI model using the fifth information by the second device, the method further includes:
- the second device When the second device is the server, the second device obtains the fifth information from at least one of the terminal and the network-side device;
- the second device obtains at least a portion of the fifth information from the terminal;
- the second device obtains at least a portion of the fifth information from the network-side device.
- the method further includes:
- the second device receives the fourth information from the first device
- the second device updates the first AI model based on the fourth information
- the second device sends the updated first AI model to the first device.
- the second device trains the first AI model to obtain the first AI model, including:
- the second device receives the first training result from the first device
- the second device inputs the first training result into the fourth AI model to train the first AI model and obtain the first AI model;
- the second device sends the first AI model to the first device.
- the second device trains the first AI model, obtains a second training result, and sends the second training result to the first device, including:
- the second device inputs the ninth information into the fifth AI model to obtain the second training result
- the second device sends the second training result to the first device
- the ninth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the method before the second device inputs the ninth information into the fifth AI model to obtain the second training result, the method further includes:
- the second device When the second device is the server, the second device obtains the ninth information from at least one of the terminal and the network-side device;
- the second device obtains at least a portion of the ninth information from the terminal;
- the second device obtains at least a portion of the ninth information from the network-side device.
- the method further includes:
- the second device receives the target result from the first device.
- the method further includes:
- the second device sends target indication information to a network-side device or server.
- the target indication information is used to activate the first AI model and indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on within a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the number of synchronization signal block (SSB) detection failures is greater than or equal to a first threshold; and the first AI model needs to be activated based on at least one of the second pieces of information, where the second pieces of information are at least a portion of the input information of the first AI model.
- SSB synchronization signal block
- the method further includes:
- the second device periodically triggers the training of the first AI model based on the first configuration information
- the first configuration information includes at least one of the following: the starting point of periodic model training; the interval of periodic model training; the number of model training sessions within one cycle; and the training duration within one cycle.
- the method further includes:
- the second device triggers the training of the first AI model using a semi-static triggering method
- the semi-static triggering method includes at least one of the following:
- Semi-static triggering is performed based on the second configuration information
- the second configuration information satisfies at least one of the following: triggering the transmission and activation of the second configuration information based on a target event; configuring the second configuration information through Radio Resource Control (RRC).
- RRC Radio Resource Control
- the method before the first device obtains the target result based on the first artificial intelligence (AI) model, the method further includes:
- the first device triggers the training of the first AI model based on the target event
- the target event includes at least one of the following:
- NCD-SSB non-cell-defined synchronization signal block
- the terminal moves to the edge of the cell or a preset location
- the terminal's moving speed is higher than or equal to a first threshold
- the terminal's moving speed is lower than or equal to the second threshold
- the number of terminals currently connected or registered in the community is higher than or equal to the third threshold
- the number of terminals currently connected to or registered in the community is less than or equal to the fourth threshold
- the number of consecutive inference failures using the first AI model reached the fifth threshold.
- the number of inference failures using the first AI model reached the sixth threshold.
- the terminal was reselected to the second cell
- the tracking area of the terminal changes
- the external environment of the terminal changes
- the terminal moved to the second cell
- the terminal was moved to a new tracking area
- the terminal was moved to a new geographical location
- the change in the terminal's moving speed is greater than or equal to the seventh threshold
- M is an integer greater than 1
- the number of failed detections of the synchronization signal block has reached the eighth threshold.
- the method further includes:
- the first device determines that the training of the first AI model is complete based on the eighth piece of information
- the eighth piece of information includes at least one of the following: the received signal strength indication of the synchronization signal; the signal strength of the detected primary or secondary synchronization signal; the channel estimation signal-to-noise ratio of the synchronization signal; the reference signal received power (RSRP) of the synchronization signal block; the RSRP of layer 1; the RSRP of layer 3; the probability that the terminal successfully detects the synchronization signal; the duration of the synchronization signal detected by the terminal; the probability that the terminal successfully camps; and the probability that the terminal successfully synchronizes.
- the received signal strength indication of the synchronization signal the signal strength of the detected primary or secondary synchronization signal
- the channel estimation signal-to-noise ratio of the synchronization signal the reference signal received power (RSRP) of the synchronization signal block
- the RSRP of layer 1 the RSRP of layer 3
- the probability that the terminal successfully detects the synchronization signal the duration of the synchronization signal detected by the terminal
- the probability that the terminal successfully camps and the probability that
- the method further includes:
- the second device transmits target capability information to the first device, the target capability information including at least one of the following:
- the network-side device supports the ability to provide auxiliary information for AI inference based on the first AI model
- the method further includes:
- the second device determines the target capability information corresponding to the first device based on at least one of the following:
- the device type of the first device is the device type of the first device
- the network type of the first device is the network type of the first device
- the reference signal sent by the second device is the reference signal sent by the second device.
- the RRC signaling sent by the first device
- Interface messages between the first device and the second device are Interface messages between the first device and the second device.
- the cell dwell processing method provided in this application can be executed by a cell dwell processing device.
- This application uses the cell dwell processing device executing the cell dwell processing method as an example to illustrate the cell dwell processing device provided in this application.
- the cell dwell processing device can be a communication device or a component within a communication device, such as a chip.
- the communication device can be a terminal, a network-side device, or a server, etc.
- the terminal can be, but is not limited to, the type of terminal 11 listed above
- the network-side device can be, but is not limited to, the type of network-side device 12 listed above. This application does not impose specific limitations.
- the cell-based processing device includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware.
- the processing module can be implemented by a processor.
- the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc.
- the receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.
- the cell dwell processing device 600 is applied to the first device, and the cell dwell processing device 600 includes:
- the first processing module 601 is used to obtain the target result based on the first artificial intelligence (AI) model;
- the first device is a terminal, a network-side device, or a server
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, wherein the first cell is a cell that can be camped.
- the first device obtains the target result based on the first artificial intelligence (AI) model, including:
- the first device inputs the first information into the first AI model to obtain the target result
- the first information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the cell dwell processing device 600 further includes: a first sending module, used to send the target result to the terminal when the first device is a network-side device or a server.
- the first processing module 601 is further configured to activate the first AI model based on a first activation condition, the first activation condition including at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by performing a cell search in the full frequency band supported by the terminal; no cell is camped in the second cell within a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the timer used to activate the first AI model times out; the number of failed synchronization signal block (SSB) detections is greater than or equal to a first threshold; and the first AI model needs to be activated based on at least one of the second pieces of information, wherein the second pieces of information are at least a portion of the input information of the first AI model.
- SSB synchronization signal block
- the first processing module 601 is further configured to activate the first AI model based on a second activation condition, wherein the second activation condition includes at least one of the following:
- the terminal receives target indication information, which indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the number of synchronization signal block (SSB) detection failures is greater than or equal to a first threshold; and it is determined that the first AI model needs to be activated based on at least one of the second pieces of information, wherein the second pieces of information are at least a portion of the input information of the first AI model.
- target indication information which indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on a second preset time period; a cell reselection
- the timer used to activate the first AI model timed out
- the need to activate the first AI model is determined based on at least one of the third pieces of information, wherein the third pieces of information are at least a portion of the input information of the first AI model.
- the first processing module 601 is further configured to perform at least one of the following:
- the first AI model is updated based on the fourth information, or the first device sends the fourth information to the second device, the fourth information being used to update the first AI model, wherein the fourth information includes at least one of the target result and the terminal's status information;
- the first operation includes at least one of the following: reverting to a cell search method that supports all frequency bands on the terminal; triggering a switch of the AI model; triggering retraining of the first AI model; and triggering supervision of the first AI model.
- the first processing module 601 is specifically used to perform a first operation when the first condition is met;
- the first condition includes at least one of the following:
- the time taken to perform inference using the first AI model exceeds the second preset time.
- the first AI model failed to find a suitable cell to reside in
- the inference was not successfully completed using the first AI model.
- the first AI model satisfies at least one of the following:
- the complexity of the AI model is lower than or equal to the second threshold
- the inference latency of the AI model is less than or equal to the third preset duration
- the AI model's reasoning success rate is greater than or equal to the third threshold
- the reliability of the AI model's inference results is greater than or equal to the fourth threshold.
- the first AI model is independently trained by the terminal, network-side device, or server, or the first AI model is jointly trained by at least two of the terminal, network-side device, and server.
- the cell dwell processing device 600 further includes: a first receiving module, used to receive the first AI model from a second device;
- the second device is a terminal, a network-side device, or a server.
- the cell dwell processing device 600 further includes: a first sending module, used to send fifth information to the second device, the fifth information being used by the second device to train the first AI model;
- the fifth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the cell dwell processing device 600 further includes: a first sending module, wherein,
- the first processing module 601 is used to input the sixth information into the second AI model to obtain the first training result, and the first training result is used by the second device to train the first AI model;
- the first sending module is used to send the first training result to the second device
- the sixth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the cell dwell processing device 600 further includes: a first receiving module, configured to perform at least one of the following:
- the sixth information is obtained from at least one of the terminal and the network-side device;
- the first device is the network-side device, at least a portion of the sixth information is obtained from the terminal;
- the first device is the terminal
- at least a portion of the sixth information can be obtained from the network-side device.
- the first processing module 601 is further configured to perform at least one of the following:
- the first AI model is trained based on the seventh information to obtain the first AI model
- the second training result received from the second device is input into the third AI model for AI training to obtain the first AI model;
- the seventh piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the input to training the first AI model may further include a target label, which includes at least one of the following:
- the terminal detected a synchronization signal at a preset frequency domain position
- the terminal detected a synchronization signal at a preset time domain location
- the terminal detected a synchronization signal in the preset beam direction
- the terminal successfully camped on the preset cell
- the first processing module 601 is further configured to:
- the training of the first AI model is periodically triggered based on the first configuration information
- the first configuration information includes at least one of the following: the starting point of periodic model training; the interval of periodic model training; the number of model training sessions within one cycle; and the training duration within one cycle.
- the first processing module 601 is further configured to: trigger the training of the first AI model based on a semi-static triggering method;
- the semi-static triggering method includes at least one of the following:
- Semi-static triggering is performed based on the second configuration information
- the second configuration information satisfies at least one of the following: triggering the transmission and activation of the second configuration information based on a target event; configuring the second configuration information through Radio Resource Control (RRC).
- RRC Radio Resource Control
- the first processing module 601 is further configured to: trigger the training of the first AI model based on the target event;
- the target event includes at least one of the following:
- NCD-SSB non-cell-defined synchronization signal block
- the terminal moves to the edge of the cell or a preset location
- the terminal's moving speed is higher than or equal to a first threshold
- the terminal's moving speed is lower than or equal to the second threshold
- the number of terminals currently connected or registered in the community is higher than or equal to the third threshold
- the number of terminals currently connected or registered in the community is less than or equal to the fourth threshold
- the number of consecutive inference failures using the first AI model reached the fifth threshold.
- the number of inference failures using the first AI model reached the sixth threshold.
- the terminal was reselected to the second cell
- the tracking area of the terminal changes
- the external environment of the terminal changes
- the terminal moved to the second cell
- the terminal was moved to a new tracking area
- the terminal was moved to a new geographical location
- the change in the terminal's moving speed is greater than or equal to the seventh threshold
- M is an integer greater than 1
- the number of failed detections of the synchronization signal block has reached the eighth threshold.
- the first processing module 601 is further configured to: determine that the training of the first AI model is complete based on the eighth information;
- the eighth piece of information includes at least one of the following: the received signal strength indication of the synchronization signal; the signal strength of the detected primary or secondary synchronization signal; the channel estimation signal-to-noise ratio of the synchronization signal; the reference signal received power (RSRP) of the synchronization signal block; the RSRP of layer 1; the RSRP of layer 3; the probability that the terminal successfully detects the synchronization signal; the duration of the synchronization signal detected by the terminal; the probability that the terminal successfully camps; and the probability that the terminal successfully synchronizes.
- the received signal strength indication of the synchronization signal the signal strength of the detected primary or secondary synchronization signal
- the channel estimation signal-to-noise ratio of the synchronization signal the reference signal received power (RSRP) of the synchronization signal block
- the RSRP of layer 1 the RSRP of layer 3
- the probability that the terminal successfully detects the synchronization signal the duration of the synchronization signal detected by the terminal
- the probability that the terminal successfully camps and the probability that
- the first processing module 601 is further configured to supervise the first AI model based on third configuration information
- the third configuration information includes at least one of the following:
- the metrics for model supervision include at least one of the following: error class information or accuracy information between predicted and true values; communication system performance; and model-related information of the first AI model.
- the triggering condition for model supervision is determined based on at least one of the following: the metrics of model supervision; the inference results of the first AI model; and the metrics of model inference.
- the cell dwell processing device 600 further includes: a transmission module for transmitting target capability information with the second device, the target capability information including at least one of the following:
- the network-side device supports the ability to provide auxiliary information for AI inference based on the first AI model
- the first processing module 601 is further configured to determine the target capability information corresponding to the second device based on at least one of the following:
- the equipment type of the second device is the equipment type of the second device
- the network type of the second device is the network type of the second device
- the reference signal sent by the second device is the reference signal sent by the second device.
- Interface messages between the first device and the second device are Interface messages between the first device and the second device.
- the cell dwell processing device 700 is applied to the second device, and the cell dwell processing device 700 includes:
- Execution module 701 is configured to perform at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a terminal, a network-side device, or a server
- the second device is a terminal, a network-side device, or a server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information of the first cell, wherein the first cell is a cell that can be camped.
- the execution module 701 is specifically used to: train the first AI model based on the fifth information to obtain the first AI model;
- the fifth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the execution module 701 includes: a second receiving module, configured to perform at least one of the following:
- the fifth information is obtained from at least one of the terminal and the network-side device;
- the second device is the network-side device, at least a portion of the fifth information is obtained from the terminal;
- the second device is the terminal, at least a portion of the fifth information may be obtained from the network-side device.
- the execution module 701 includes:
- the second receiving module is used to receive fourth information from the first device
- the second processing module is used to update the first AI model based on the fourth information
- the second sending module is used to send the updated first AI model to the first device.
- the execution module 701 includes:
- the second receiving module is used to receive the first training result from the first device
- the second processing module is used to input the first training result into the fourth AI model to train the first AI model and obtain the first AI model.
- the second sending module is used to send the first AI model to the first device.
- the execution module 701 includes:
- the second processing module is used to input the ninth information into the fifth AI model to obtain the second training result
- the second sending module is used to send the second training result to the first device
- the ninth piece of information includes at least one of the following:
- the terminal's status information includes at least one of the following: location information, direction of movement, speed of movement, energy consumption status, battery status, environmental information, sensing information, network scene information, operator information, and network type;
- the execution module 701 includes a second receiving module, configured to perform at least one of the following:
- the ninth information is obtained from at least one of the terminal and the network-side device;
- the second device is the network-side device, at least a portion of the ninth information is obtained from the terminal;
- the second device is the terminal, at least a portion of the ninth information is obtained from the network-side device.
- the execution module 701 includes a second receiving module, used to receive the target result from the first device when the second device is a terminal.
- the execution module 701 includes a second sending module, configured to send target indication information to a network-side device or server when the second device is a terminal.
- the target indication information is used to activate the first AI model and indicates at least one of the following: the terminal is powered on; an initial cell search operation is performed; an initial cell selection operation is performed; an initial search is performed for a first preset duration by searching the entire frequency band supported by the terminal; no cell is camped on within a second preset time period; a cell reselection operation is performed; a cell handover operation is performed; the number of synchronization signal block (SSB) detection failures is greater than or equal to a first threshold; and the first AI model needs to be activated based on at least one of the second pieces of information, wherein the second pieces of information are at least a portion of the input information of the first AI model.
- SSB synchronization signal block
- the execution module 701 is specifically used to periodically trigger the training of the first AI model based on the first configuration information
- the first configuration information includes at least one of the following: the starting point of periodic model training; the interval of periodic model training; the number of model training sessions within one cycle; and the training duration within one cycle.
- the execution module 701 is specifically used to trigger the training of the first AI model based on a semi-static triggering method
- the semi-static triggering method includes at least one of the following:
- Semi-static triggering is performed based on the second configuration information
- the second configuration information satisfies at least one of the following: triggering the transmission and activation of the second configuration information based on a target event; configuring the second configuration information through Radio Resource Control (RRC).
- RRC Radio Resource Control
- the execution module 701 is specifically used to trigger the training of the first AI model based on the target event;
- the target event includes at least one of the following:
- NCD-SSB non-cell-defined synchronization signal block
- the terminal moves to the edge of the cell or a preset location
- the terminal's moving speed is higher than or equal to a first threshold
- the terminal's moving speed is lower than or equal to the second threshold
- the number of terminals currently connected or registered in the community is higher than or equal to the third threshold
- the number of terminals currently connected to or registered in the community is less than or equal to the fourth threshold
- the number of consecutive inference failures using the first AI model reached the fifth threshold.
- the number of inference failures using the first AI model reached the sixth threshold.
- the terminal was reselected to the second cell
- the tracking area of the terminal changes
- the external environment of the terminal changes
- the terminal moved to the second cell
- the terminal was moved to a new tracking area
- the terminal was moved to a new geographical location
- the change in the terminal's moving speed is greater than or equal to the seventh threshold
- M is an integer greater than 1
- the number of failed detections of the synchronization signal block has reached the eighth threshold.
- the execution module 701 is specifically used to determine that the training of the first AI model has been completed based on the eighth information
- the eighth information includes at least one of the following: the received signal strength indication of the synchronization signal; the signal strength of the detected primary or secondary synchronization signal; the channel estimation signal-to-noise ratio of the synchronization signal; the reference signal received power (RSRP) of the synchronization signal block; the RSRP of layer 1; the RSRP of layer 3; the probability that the terminal successfully detects the synchronization signal; the duration of the synchronization signal detected by the terminal; the probability that the terminal successfully camps; and the probability that the terminal successfully synchronizes.
- the received signal strength indication of the synchronization signal the signal strength of the detected primary or secondary synchronization signal
- the channel estimation signal-to-noise ratio of the synchronization signal the reference signal received power (RSRP) of the synchronization signal block
- the RSRP of layer 1 the RSRP of layer 3
- the probability that the terminal successfully detects the synchronization signal the duration of the synchronization signal detected by the terminal
- the probability that the terminal successfully camps and the probability that the terminal
- the execution module 701 is further configured to: transmit target capability information with the first device, the target capability information including at least one of the following:
- the network-side device supports the ability to provide auxiliary information for AI inference based on the first AI model
- the execution module 701 further includes:
- the second processing module is configured to determine the target capability information corresponding to the first device based on at least one of the following:
- the device type of the first device is the device type of the first device
- the network type of the first device is the network type of the first device
- the reference signal sent by the second device is the reference signal sent by the second device.
- the RRC signaling sent by the first device
- Interface messages between the first device and the second device are Interface messages between the first device and the second device.
- the cell dwell processing device provided in this application embodiment can implement the various processes implemented in the method embodiments of Figures 4 to 5 and achieve the same technical effect. To avoid repetition, it will not be described again here.
- this application embodiment also provides a communication device 800, including a processor 801 and a memory 802.
- the memory 802 stores a program or instructions that can run on the processor 801.
- the program or instructions are executed by the processor 801, they implement the various steps of the above-described cell camping processing method embodiment and can achieve the same technical effect. To avoid repetition, they will not be described again here.
- This application also provides a terminal, including 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 steps in the method embodiments shown in FIG4 or FIG5.
- This terminal embodiment corresponds to the above-described first device-side or second device-side method embodiments, and all implementation processes and methods of the above-described method embodiments can be applied to this terminal embodiment and can achieve the same technical effect.
- the terminal may be the cell dwell processing device shown in FIG6 or FIG7.
- FIG9 is a schematic diagram of the hardware structure of a terminal implementing an embodiment of this application.
- the terminal 900 includes, but is not limited to, at least some of the following components: radio frequency unit 901, network module 902, audio output unit 903, input unit 904, sensor 905, display unit 906, user input unit 907, interface unit 908, memory 909, and processor 910.
- the terminal 900 may also include a power supply (such as a battery) for powering various components.
- the power supply can be logically connected to the processor 910 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.
- the terminal structure shown in Figure 9 does not constitute a limitation on the terminal.
- the terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.
- the input unit 904 may include a graphics processor 9041 and a microphone 9042.
- the graphics processor 9041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode.
- the display unit 906 may include a display panel 9061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like.
- the user input unit 907 includes at least one of a touch panel 9071 and other input devices 9072.
- the touch panel 9071 is also called a touch screen.
- the touch panel 9071 may include a touch detection device and a touch controller.
- Other input devices 9072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.
- the radio frequency unit 901 can transmit it to the processor 910 for processing; in addition, the radio frequency unit 901 can send uplink data to the network-side device.
- the radio frequency unit 901 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.
- the memory 909 can be used to store software programs or instructions, as well as various data.
- the memory 909 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data.
- the first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.).
- the memory 909 may include volatile memory or non-volatile memory.
- the non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
- Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (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 RAM
- Processor 910 may include one or more processing units; optionally, processor 910 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 910.
- the processor 910 is used to obtain the target result based on the first artificial intelligence (AI) model;
- the target result is used for cell camping, and the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and information about the first cell, which is a cell that can be camped.
- the radio frequency unit 901 is used to perform at least one of the following:
- the second operation includes at least one of the following:
- the first AI model is trained to obtain a second training result, which is then sent to the first device.
- the second training result is used by the first device to train the first AI model.
- the first device is a network-side device or server
- the first AI model is used to determine the target result
- the target result is used for cell camping
- the target result includes at least one of the following: the frequency domain location of the synchronization signal transmission; the time domain location of the synchronization signal transmission; the beam transmission direction of the synchronization signal transmission; and the information of the first cell, which is a cell that can be camped.
- This 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 programs or instructions to implement the steps of the method embodiment shown in FIG4 or FIG5.
- This network-side device embodiment corresponds to the above-described first device-side or second device-side method embodiment. All implementation processes and methods of the above-described method embodiments can be applied to this network-side device embodiment and can achieve the same technical effect.
- this application embodiment also provides a network-side device, which may be the cell dwell processing device shown in FIG. 6 or FIG. 7.
- 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 transmitted and sends it to the radio frequency device 1002, which processes the received information and then transmits it through the antenna 1001.
- the method executed by the network-side device in the above embodiments can be implemented in the baseband device 1003, which includes a baseband processor.
- the baseband device 1003 may include at least one baseband board, on which multiple chips are disposed, as shown in FIG10.
- One of the chips is, for example, a baseband processor, which is connected to the memory 1005 via a bus interface to call the program in the memory 1005 to execute the network-side device operation shown in the above method embodiment.
- the network-side device may also include a network interface 1006, such as a Common Public Radio Interface (CPRI).
- CPRI Common Public Radio Interface
- the network-side device 1000 in this application embodiment further includes: instructions or programs stored in memory 1005 and executable on processor 1004.
- Processor 1004 calls the instructions or programs in memory 1005 to execute the methods executed by the modules shown in FIG6 or FIG7 and achieve the same technical effect. To avoid repetition, it will not be described in detail here.
- This application also provides a readable storage medium storing a program or instructions.
- the program or instructions When the program or instructions are executed by a processor, they implement the various processes of the above-described cell dwell processing method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.
- the processor mentioned above is the processor in the terminal described in the above embodiments.
- 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.
- ROM computer read-only memory
- RAM random access memory
- magnetic disk magnetic disk
- optical disk optical disk
- the readable storage medium may be a non-transient readable storage medium.
- This application embodiment also provides a chip, which includes a processor and a communication interface.
- the communication interface and the processor are coupled.
- the processor is used to run programs or instructions to implement the various processes of the above-described cell dwell processing method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.
- chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
- This application also provides a computer program/program product, which includes computer instructions.
- the computer program/program product is executed by at least one processor to implement the various processes of the above-described cell dwell processing method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.
- This application also provides a wireless communication system, including a first device and a second device.
- the first device can be used to perform the steps of the cell camping processing method on the first device side as described above
- the second device can be used to perform the steps of the cell camping processing method on the second device side as described above.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本申请公开了一种小区驻留处理方法、装置、终端及网络侧设备,属于通信技术领域,本申请实施例的小区驻留处理方法包括:第一设备基于第一人工智能AI模型获得目标结果;其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
Description
相关申请的交叉引用
本申请主张在2024年6月4日在中国提交的申请号为202410717888.4的中国专利的优先权,其全部内容通过引用包含于此。
本申请属于通信技术领域,具体涉及一种小区驻留处理方法、装置、终端及网络侧设备。
随着通信技术发展,在通信系统中,终端开机进行小区搜索时,通常会先读取用户识别模块(Subscriber Identity Module,SIM)卡的信息,根据SIM卡中存储的先验信息(如之前存储的小区和频点信息等)寻找合适的小区驻留。如果终端没有存储先验信息,或者基于存储的先验信息没有找到合适的驻留小区(例如存储的先验信息已经失效或者不适用),此时终端将会触发在SIM卡支持的频段上进行扫频搜网,执行初始小区选择,这个过程需要花费较长的时间,从而导致终端进行小区驻留的时延较长。
本申请实施例提供一种小区驻留处理方法、装置、终端及网络侧设备,能够解决由于初始小区搜索和选择的时间较长,导致终端进行小区驻留的时延较长的问题。
第一方面,提供了一种小区驻留处理方法,包括:
第一设备基于第一人工智能AI模型获得目标结果;
其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第二方面,提供了一种小区驻留处理方法,包括:
第二设备执行以下至少一项:
所述第二设备向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
所述第二设备进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
所述第二设备进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第三方面,提供了一种小区驻留处理装置,应用于第一设备,包括:
第一处理模块,用于基于第一人工智能AI模型获得目标结果;
其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第四方面,提供了一种小区驻留处理装置,应用于第二设备,包括:
执行模块,用于执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第五方面,提供了一种小区驻留处理的装置,所述装置被配置为执行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。
第六方面,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。
第七方面,提供了一种终端,包括处理器及通信接口,其中,
在终端为第一设备时,处理器用于基于第一人工智能AI模型获得目标结果;
其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
在终端为第二设备时,通信接口用于执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第八方面,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。
第九方面,提供了一种网络侧设备,包括处理器及通信接口,其中,
在网络侧设备为第一设备时,处理器用于基于第一人工智能AI模型获得目标结果;
其中,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
在网络侧设备为第二设备时,通信接口用于执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为终端或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
第十方面,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。
第十一方面,提供了一种无线通信系统,包括:第一设备及第二设备,所述第一设备可用于执行如第一方面所述的方法的步骤,所述第二设备可用于执行如第二方面所述的方法的步骤。
第十二方面,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。
第十三方面,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,所述计算机程序/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方法的步骤,或实现如第二方面所述的方法的步骤。
本申请实施例通过第一设备基于第一人工智能AI模型获得目标结果;其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。这样,终端可以基于目标结果中的时频域位置、波束发送方向和第一小区的信息进行初始小区搜索和选择的时间,从而避免在终端支持的全频段上进行扫频搜网。因此,本申请实施例缩短了初始小区搜索和选择的时间,从而降低了终端进行小区驻留的时延,与此同时降低了初始小区搜索和选择的能耗。
图1是本申请实施例可应用的一种无线通信系统的框图;
图2a至图2c是SSB和CORESET 0的复用样式示例图;
图3是神经元的结构示意图;
图4是是本申请实施例提供的一种小区驻留处理方法的流程示意图;
图5是是本申请实施例提供的另一种小区驻留处理方法的流程示意图;
图6是是本申请实施例提供的一种小区驻留处理装置的结构示意图;
图7是是本申请实施例提供的另一种小区驻留处理装置的结构示意图;
图8是是本申请实施例提供的一种通信设备的结构示意图;
图9是是本申请实施例提供的一种终端的结构示意图;
图10是是本申请实施例提供的一种网络侧设备的结构示意图。
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”的保护范围至少涵盖三种方案,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。此外,术语“A和/或B”、“A和B中的至少一项”、“A或B中的至少一项”也分别至少涵盖上述三种方案。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)、车载用户设备(Vehicle User Equipment,VUE)、船载设备、行人用户设备(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(Transmit/Receive Point,TRP)或所属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。
为了方便理解,以下对本申请实施例涉及的一些内容进行说明:
一、小区搜索(Cell Search)。
小区搜索是终端获取与小区的时间和频率同步并对该小区的小区标识(ID)进行解码的过程。小区搜索的主要目的包括:
1.与小区取得频率和符号同步(下行同步)。
2.获取系统帧定时timing,即下行帧的起始位置。
3.确定小区的物理小区标识(Physical Cell Identity,PCI)。
NR小区搜索基于主同步信号PSS(Primary Synchronization Signal)和辅同步信号SSS(Secondary Synchronization Signal)以及位于同步光栅上的物理广播信道(Physical broadcast channel,PBCH)解调参考信号(Demodulation Reference Signal,DMRS)。NR小区搜索的总体过程可以分为以下几个步骤:
1.终端调谐到特定频率并仅测量接收信号强度指示(Received Signal Strength Indication,RSSI)。
2.终端尝试检测SSB并解码主同步信号(Primary Synchronisation Signal,PSS)和辅同步信号(Secondary Synchronisation Signal,SSS)。如果终端在这一步失败,则进入步骤1,如果在这一步通过,则进入下一步。
3.一旦终端成功检测到PSS或SSS,终端尝试解码PBCH。
4.一旦终端成功检测到PBCH,它就解码主信息块(Master Information Block,MIB)。
5.基于MIB中的pdcch-ConfigSIB1,查找控制资源集0(ORESET#0,即用于SIB1传输的物理下行控制信道(Physical Downlink Control Channel,PDCCH)或下行控制信息(Downlink Control Information,DCI)的CORESET)的位置和搜索空间(Search Space)信息。
6.在Search Space中盲解码系统信息无线网络临时标识(System Information Radio Network Temporary Identifier,SI-RNTI)加扰的DCI 1_0。
7.基于DCI 1_0的内容,检测并解码携带系统信息块(System Information Block,SIB)1的物理下行共享信道(Physical Downlink Shared Channel,PDSCH)。
8.解码SIB1和其他SIB(如果SIB1携带有关其他SIB的信息)。
终端不仅需要在开机时进行小区搜索,为了支持移动性(mobility),终端会不停地搜索邻居小区、取得同步并估计该小区信号的接收质量,从而决定是切换(handover,当UE处于RRC_CONNECTED态)或小区重选(cell re-selection,当UE处于RRC_IDLE态)。
同步信号块(Synchronization Signal and PBCH block,SSB)子载波间隔(Subcarrier Spacing,SCS)由频段范围决定:6GHz以下(即FR1)支持15或30kHz的SCS,6GHz以上(即FR2)支持120或240khz的SCS。终端从PSS获知N_ID_(2),从SSS获知N_ID_(1),则小区ID即的物理小区标识符(Physical Cell Identifier,PCI)为:
N_cell_ID=3x N_ID_(1)+N_ID_(2);
N_cell_ID=3x N_ID_(1)+N_ID_(2);
其中,N_ID_(2)取值范围为{0,1,2},N_ID_(1)取值范围为{0,1,…335},NR共有336x 3=1008个N_cell_ID,取值范围为{0,1,…,1007}。对于FR1大部分频带只支持一种SSB SCS,终端确定了频带,就同时知道了SSB SCS。但也有部分频带n5/n41/n66/n90支持两种SSB SCS(15kHz和30kHz),终端需要用两种SCS分别进行盲检,以确定小区的SSB SCS。FR2 n257/258/259/260/261都支持120/240khz,这时就只能分别试试一遍确定SCS。
二、同步栅格(Sync raster)。
信道栅格(Channel raster)可以理解为载波的中心频点的可选位置。协议在定义信道栅格时,先定义了全局频率栅格(global frequency raster)。信道栅格就是在全局频率栅格的基础上,根据工作频段(operating band)做了范围及步长的限制。在5G NR中,全局频率栅格定义为频率栅格参考频率(RF reference frequencies,FREF)的集合,频域范围为0-100GHz,主要是为了标识RF信道、SSB或者其他资源的频域位置。
NR绝对射频信道号(NR Absolute Radio Frequency Channel Number,NR-ARFCN)则对RF参考频率的频域范围进行编码,0~100GHz的取值范围为FR1[0…2016666]及FR2[2016667...3279165],NR-ARFCN和RF参考频率FREF的关系如下式(1)所示。ARFCN频点号对应信道栅格。信道栅格在不同NR band的间隔密度不同。
FREF=FREF-Offs+ΔFGlobal(NREF–NREF-Offs)
(1)
FREF=FREF-Offs+ΔFGlobal(NREF–NREF-Offs)
(1)
全球频率光栅的NR-ARFCN参数如下表一所示。
表一:
在NR中,由于信道带宽可能非常大,如果终端按照信道栅格进行同步信号搜索,需要的时间很长,且非常耗电。因此NR引入了Synchronization raster的概念,同步信号按照同步栅格放置。
GSCN频点号对应同步栅格。GSCN对0-100GHz范围的频段做了定义,每个GSCN对应一个SSB的检测频点。终端在进行全频段搜索时,只可能盲检GSCN位置的SSB。相似的,如下表二所示,0~100GHz对应0~26639个GSCN。
表二:
其中,仅支持SCS间隔通道光栅的操作频带的默认值为M=3。
GSCN可用来描述各个频带的同步信道。同步栅格是GSCN的子集,不同band的同步栅格的频率间隔不同。n41 band上,同步栅格的频率间隔为3个GSCN。n79band上,同步栅格的间隔为16个GSCN。
三、同步信号块。
NR中,PSS或SSS和PBCH总是绑定的,因此也称为SSB。一个SSB在时域上一共占用4个符号(time indices l=0~3),在频域上分布在连续的240个子载波(20个RB)。频域由下往上,第121个子载波SC的中心频率,就是SSB的GSCN对应的同步参考频率(SSREF)。NR中SSB的时域位置和频域位置都不再固定,而是灵活可变的。频域上,SSB不再固定于频带中间;时域上,SSB发送的位置和数量都可能变化。所以在NR中,仅通过解调PSS或SSS信号,是无法获得频域和时域资源的完全同步的,必须完成PBCH的解调,才能最终达到时频资源的同步。
小区定义或非小区定义SSB(cell-defining or non cell-defining SSB)。
在NR系统之中,小区定义SSB(cell-defining SSB,CD-SSB)定义为与SIB1(也即剩余最小系统信息(Remaining Minimum SI,RMSI))关联的SSB。SIB1定义了其它SIB的调度信息,并包含用于终端初始接入的信息。CD-SSB的频率位置一定在系统同步栅格上。
非小区定义SSB(non cell-defining SSB,NCD-SSB)相应地定义为不与SIB1关联的SSB。NCD-SSB可以用于辅小区同步,也可以用作给终端配置的测量信号。NCD-SSB不一定位于系统同步栅格上。如果NCD-SSB位于系统同步栅格上,它可以通过其携带的信息指示CD-SSB的GSCN。
当终端在小区搜索期间检测到一个SSB,终端首先需要确定该SSB是CD-SSB还是NCD-SSB。确定的方式是根据该SSB的PBCH所提供的、代表SSB和公共资源块网格之间的子载波偏移kSSB是否在有效的子载波偏移值范围内进行判断。有效的子载波偏移值范围包括0-23个和0-11个子载波,分别使用5bit和4bit表示,分别对应频率范围FR1和FR2。若kSSB的值在有效子载波偏移值范围内,则该SSB是CD-SSB,否则是NCD-SSB。
在FR1中,如果kSSB>23,或在FR2中,如果kSSB>11,都表示SSB不存在类型0(Type 0)公共搜索空间(Common Search Space,CSS),即当前SSB不关联SIB1。不过为了让UE快点找到Cell Defining SSB,这些kSSB还可以作为索引,结合RMSIPDCCH Config(即MIB的PDCCH Config SIB1),(间接的)指示下一个SSB的GSCN。
当取值kSSB=31(FR1)或kSSB=15(FR2)时,终端认为在一定的GSCN范围内不存在CD-SSB。
可选地,本申请实施例中的SSB或者同步信号,也可以叫做任何包含同步信号、广播信号、广播信道(PBCH)、其他系统消息下行广播信道或其控制信道中至少一种的模块。
四、SIB1。
Type0-PDCCH公共搜索空间集合(CSS set):
该搜索空间集合用于监听SIB1系统消息,对应在主小区组(Master Cell Group,MCG)中的原小区(primary cell)中用SI-RNTI加扰的DCI,在信令MIB中由IE:pdcch-ConfigSIB1配置或者在信令PDCCH-ConfigCommon中由IE:serachSpaceZero配置,或者在信令PDCCH-ConfigCommon中由IE:searchSpaceZero或者searchSpaceSIB1配置。
SSB和CORESET 0:
SSB和CORESET 0复用样式(Multiplexing Pattern)分为三种:模式(Pattern)1(如图2a所示)为时分复用(CORESET 0频域范围包含SSB),Pattern 2(如图2b所示)和Pattern 3(如图2c所示)为频分复用(CORESET 0和SSB在同一系统帧)。Pattern 2和Pattern 3差异在于:在时域上,Pattern 2的CORESET 0比SSB位置略微靠前。
终端解完SSB之后,就能知道在具体的时频资源位置去盲检SIB1的调度信息。
在检测到PBCH之后,终端已经完成了下行同步,在进行上行同步之前,终端需要进一步接收SIB1,获得与上行同步相关的配置信息。
SIB1在PDSCH中传输,并通过PDCCH进行调度,且PDSCH的资源分配范围在初始BWP的频率范围内:
(1)SIB1的PDCCH时频域资源分配(调度SIB1的DCI信息就承载在CORESET0里面)。
SIB1的PDCCH映射在type 0-PDCCH的公共搜索空间(CCS)内;
频域上,Type 0-PDCCH的CSS映射在CORESET 0中,且CORESET 0的频率范围与初始BWP完全相同;
PBCH中承载的信令‘pdcch-ConfigSIB1’的低位4bit指示了type 0-PDCCH CSS的配置;高位4bit指示了CORESET 0的配置,
(2)SIB1的PDSCH时频域资源分配。
常规的PDSCH使用无线资源控制(Radio Resource Control,RRC)配置的时域资源分配(Time Domain Resource Assignment,TDRA)表格,由PDCCCH指示表格中的索引进行时域资源分配。但是由于终端在接收SIB1 PDSCH的时候,RRC连接还么有建立,因此需要定义默认的TDRA。
CORESET0与SSB的复用3种模式分别对应3个默认的TDRA表格;
频域上,SIB1在初始接入带宽范围内进行频域资源分配,使用资源分配类型type 1,
五、AI。
AI可以表示为机器学习(machine learning,ML),AI目前在各个领域获得了广泛的应用,将人工智能融入无线通信网络,显著提升吞吐量、时延以及用户容量等技术指标是未来的无线通信网络的重要任务。AI模块有多种实现方式,例如神经网络、决策树、支持向量机、贝叶斯分类器等。
可选地,神经网络由神经元组成,神经元的示意图如图3所示,其中a1,a2,…aK为输入,w为权值,即乘性系数,b为偏置,即加性系数,σ(.)为激活函数。常见的激活函数包括Sigmoid、tanh、线性整流函数(Rectified Linear Unit,ReLU)等。其中,z=a1w1+···+akwk+···+aKwK+b。
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的一类算法,目标函数也可以称之为损失函数。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,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。
常见的优化算法包括:梯度下降(Gradient Descent)、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)、带动量的随机梯度下降(Nesterov)、自适应梯度下降(ADAptive GRADient descent,Adagrad)、Adadelta、均方根误差降速(root mean square prop,RMSprop)和自适应动量估计(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或损失,对当前神经元求导数或偏导,加上学习速率、之前的梯度、导数或偏导等影响,得到梯度,将梯度传给上一层。
一般而言,根据解决类型不同,选取的AI算法和采用的AI模型也有所差别。根据目前发表文章及公开研究成果,借助AI提升5G网络性能的主要方法是通过基于神经网络的算法和AI模型增强或者替代目前已有的算法或处理模块。在特定场景下,基于神经网络的算法和AI模型可以取得比基于确定性算法更好的性能。比较常用的神经网络包括深度神经网络、卷积神经网络和循环神经网络等。借助已有AI工具,可以实现神经网络的搭建、训练与验证工作。
可选地,在实践中,由于实时采集的数据集不够大,直接训练神经网络很难达到收敛。常见的做法是基于离线收集的大量数据预训练网络,使其达到收敛。再用实时采集的数据对预训练的神经网络参数进行微调(fine-tuning),使神经网络适配实际环境。可认为fine-tuning是用预训练的神经网络的参数作为初始化,进行的训练过程。Fine-tuning阶段可以冻结部分层的参数,一般冻结靠近输入端的层,激活靠近输出端的层,这样可以保障网络仍然能收敛。Fine-tuning阶段的数据量越少,建议冻结的层数越多,只微调靠近输出端的少量层。
可选地,神经网络的泛化是指神经网络对未在训练(学习)过程中遇到的数据可以得到合理的输出。面向多变的无线传输环境造成的泛化问题,基于神经网络的无线通信系统有两种解决方案。第一种是在不同的传输条件下分别训练不同的神经网络,获得多组神经网络的参数,随着实际环境的变化切换神经网络的参数。第二种是基于混合的数据训练一个共性的神经网络,神经网络的参数不随环境的变化而切换。这两种模式各有利弊:第一种方案在不同传输条件下性能都很优秀,但是需要存储多个网络参数并按需切换(产生信令开销,频繁切换等问题);第二种方案只需存储一套神经网络参数,无需切换,但无法实现每个传输条件下都达到最优性能。混合数据集的构造方式会影响第二种方案下的性能。
可选地,在机器学习和深度学习中,标签(label)通常是指对数据样本的真实类别或目标值的标识或标注。标签用于表示模型应该学习并预测的信息,例如:
分类任务中的标签:在分类任务中,标签表示数据样本属于哪个类别。例如,对于图像分类,每个图像样本都有一个标签,表示图像中所包含的物体或场景的类别,如“狗”或“猫”。
目标检测中的标签:在目标检测任务中,标签通常包括对象的位置信息(边界框)和类别信息。每个标签标识了图像中的一个目标物体,包括其位置和类别。回归任务中的标签:
在回归任务中,标签通常表示要预测的连续或实数值目标。例如,房价预测任务中的标签可以是房屋的实际销售价格。
序列标注中的标签:在自然语言处理中,序列标注任务中的标签通常用于词性标注、命名实体识别等任务,其中标签用于表示文本序列中每个词或字符的属性或类别。
标签是监督学习任务中的关键组成部分,用于训练机器学习模型。模型通过与真实标签的比较来学习模式和规律,以便在未见过的数据上进行预测或分类。标签的质量和准确性对于模型的性能至关重要。
六、AI的生命周期管理(life cycle management,LCM)。
AI模型的生命周期管理包括多个AI功能模块:模型训练模块、模型管理模块、模型推理模块、模型监控模块、模型更新模块。
其中,模型训练模块用于执行AI模型训练、验证和测试,可以生成可用作模型测试过程一部分的模型性能指标。如果需要,该功能还负责基于数据收集功能提供的训练数据进行数据准备(例如,数据预处理和清理、格式化和转换)。
模型管理模块用于监督AI模型或下发AI功能的操作(例如模型选择、(去)激活、切换或回退),反馈模型监控性能。该模块还负责根据从数据收集模块和模型推理模块接收的数据做出决策,以确保正确的推理操作。
可选地,可以包括以下指令或请求:
管理指令:用于模型管理模块向模型推理模块输入的信息。有关信息可能包括AI模型或基于AI的功能来选择、(停用)激活或切换模型、回退到非AI操作(即不依赖推理过程)等;
模型传输请求:用于向模型存储功能请求模型;
性能反馈或再训练请求:模型训练模块输入所需的信息,例如用于模型(重新)训练或更新目的。
模型推理模块,用于使用数据收集功能提供的数据(即推理数据)作为输入,提供应用AI模型的输出。如果需要,模型推理模块还负责基于数据收集功能提供的推理数据进行数据准备(例如,数据预处理和清理、格式化和转换)。
可选地,推理输出:模型管理模块用于监控AI模型或AI功能性能的数据。
本申请中AI模型可以称之为AI单元、机器学习(machine learning,ML)模型、ML单元、AI结构、AI功能、AI特性、神经网络、神经网络函数、神经网络功能等,或者所述AI模型也可以指能够实现与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模型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模型可以是运行在图形处理单元(Graphics Processing Unit,GPU)、神经处理单元(Neural Processing Unit,NPU)、张量处理单元(Tensor Processing Unit,TPU)、应用集成电路(Application Specific Integrated,ASIC)等AI或ML相关硬件上的处理方法、算法、功能、模块或单元,在此不做进一步的限定。
需要说明的是,终端开机后进行小区搜索,通常会先读取SIM卡信息,根据SIM卡中存储的先验信息(比如之前存储的小区、频点信息等)寻找合适的小区进行驻留。但是在一些情况下,比如你将终端切换到飞行模式后飞往另一个国家,然后在另一个国家切换回正常模式,终端存储的先验信息可能已经过时。终端开机后使用之前存储的先验信息不能找到合适的小区接入,终端会在SIM支持的频段上进行扫频搜网,执行初始小区选择,这个过程需要花费较长的时间,并且能耗也会比较高。另外,运营商在不同的地理位置部署的网络可能会有不同的GSCN。运营商也可能在特殊的场所或者节日,临时部署新的网络,比如在开大型运动会或演唱会的场所,又比如在跨年时人口聚集的场所。这些临时部署的网络可能会有不同的GSCN。如何在任意的时间,任意的地点,当终端开机的时候能够实现快速的初始小区搜索和选择,减少初搜时间,节约终端能耗。为此,本申请实施例提供了一种小区驻留处理方法。
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小区驻留处理方法进行详细地说明。
参照图4,本申请实施例提供了一种小区驻留处理方法,如图4所示。本申请实施例提供的小区驻留处理方法包括:
步骤401,第一设备基于第一人工智能AI模型获得目标结果;
其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
本申请实施例中,上述目标结果可以理解或替换为第一AI模型的推理结果,或者目标结果基于第一AI模型的推理结果确定,例如第一AI模型的输出结果包括是否驻留到某一小区,从而基于输出结果确定第一小区的信息。由于通过第一AI模型获得目标结果,然后基于第一AI结果进行小区的搜索和选择,以实现终端的小区驻留,从而可以避免在SIM卡支持的全频段进行小区搜索,因此降低了终端进行小区驻留的时延。
可选地,第一小区的信息可以包括PCI。其中,第一小区可以理解为候选小区。
可选地,在一些实施例中,当目标结果包括同步信号发送的频域位置、同步信号发送的时域位置和同步信号发送的波束发送方向中的至少一项时,可以基于该目标结果进行小区搜索,从而无需在SIM卡支持的全频段进行小区搜索。
可选地,在一些实施例中,在上述目标结果包括第一小区的信息的情况下,终端可以直接选择该第一小区进行驻留或者只在该第一小区进行小区搜索,从而节省小区搜索的时间。
可选地,上述同步信号发送的波束发送方向可以理解或替换为同步信号发送的最强波束发送方向。
本申请实施例通过第一设备基于第一人工智能AI模型获得目标结果;其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。这样,终端可以基于目标结果中的时频域位置、波束发送方向和第一小区的信息进行初始小区搜索和选择的时间,从而避免在终端支持的全频段上进行扫频搜网。因此,本申请实施例缩短了初始小区搜索和选择的时间,从而降低了终端进行小区驻留的时延,与此同时降低了初始小区搜索和选择的能耗。
可选地,在一些实施例中,所述第一设备基于第一人工智能AI模型获得目标结果包括:
所述第一设备将第一信息输入第一AI模型,获得所述目标结果;
其中,所述第一信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
本申请实施例中,上述时间信息可以具体到特定的时间点,例如,13点25分38秒。也可以是一个时间范围,比如13点到14点,上午或下午,白天或夜晚。
可选地,该时间信息可以是通过其他无线接入技术(Radio Access Technology,RAT)获得,该其他RAT可以是蓝牙、Wi-Fi、3G、4G或LTE等。
可选地,上述位置信息可以是具体的位置坐标,例如GPS坐标;或者是终端的大致范围信息,例如位于哪个街道,哪个国家等范围信息;或者是终端相对于驻留小区或接入小区的位置信息,例如在驻留小区的正东方向。
上述终端的移动方向可以是一个绝对的方向,比如东偏南40度;也可以是一个相对反向,例如相对于驻留小区或接入小区的方向。
上述所处环境信息可以包括天气信息等。
可选地,上述网络场景信息可以包括室内热点(Indoor hotspots,inH)、城市宏蜂窝(Urban Macrocell,Uma)、农村宏蜂窝(Rural Macrocellular,RMa)等,或者包括同构网络或异构网络,即有无重叠覆盖。
可选地,上述频域特征可以包括以下至少一项:
潜在可能存在同步信号的sync raster或GSCN频点位置,例如协议中定义的可能会出向同步信号的频域位置;
同步信号的频域资源块(resource block,RB)或者子载波数目。
可选地,上述时域特征可以包括同步信号的时域符号数目,也可以包括同步信号的时域相关检测的时间窗长。
可选地,上述网络类型可以包括地面网络(Terrestrial Network,TN)、非地面网络(Non Terrestrial Network,NTN)和无小区(cell free)网络等。
可选地,上述突发事件包括但不限于演唱会和地震等。
可选地,可以在网络侧设备进行第一AI模型的推理,也可以在终端进行第一AI模型的推理,还可以在服务器进行第一AI模型的推理。例如,在一些实施例中,在所述第一设备为网络侧设备或者服务器的情况下,所述方法还包括:
所述第一设备向所述终端发送所述目标结果。
本申请实施例中,在终端不具备AI模型推理功能的情况下,可以由网络侧设备或服务器进行模型推理,从而可以降低对终端的要求,提高使用AI模型进行基于AI的小区搜索和选择的应用范围。
可选地,在一些实施例中,所述方法还包括:
在所述第一设备为终端的情况下,所述第一设备基于第一激活条件激活所述第一AI模型,所述第一激活条件包括以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;用于激活所述第一AI模型的定时器超时;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;
或者,在所述第一设备为网络侧设备或服务器的情况下,所述第一设备基于第二激活条件激活所述第一AI模型,所述第二激活条件包括以下至少一项:
从终端接收到目标指示信息,所述目标指示信息用于指示以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;
用于激活所述第一AI模型的定时器超时;
基于第三信息中的至少一项确定需要激活所述第一AI模型,所述第三信息为所述第一AI模型的至少部分输入信息。
本申请实施例中,触发第一AI模型激活可以理解或替换为使用第一AI模型进行基于AI的小区搜索和选择。例如,可以在终端开机就触发第一AI模型激活,即在终端开机就触发使用第一AI模型进行基于AI的小区搜索和选择。又例如,可以在执行初始小区搜索操作的时候触发第一AI模型激活,即在执行初始小区搜索操作的时候使用第一AI模型进行基于AI的小区搜索和选择。
应理解,在终端是第一AI模型的推理设备时,终端可以在满足第一激活条件的情况下,激活第一AI模型,在终端不是第一AI模型的推理设备时,终端可以在满足第一激活条件的情况下,向网络侧设备或服务器发送上述目标指示信息,以指示网络侧设备或服务器激活第一AI模型。
可选地,可以在执行初始小区选择操作的时候触发第一AI模型激活。
可选地,可以在通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长。
可选地,可以在第二预设时间段内没有驻留到第二小区后,触发第一AI模型激活。其中第二小区可以理解或替换为新的小区或新驻留的小区,即除上一次驻留的小区之外的其他小区。也就是说终端超过第二预设时间段没有成功驻留到新的小区后触发第一AI模型激活。
可选地,可以在执行小区重选操作的时候触发第一AI模型激活。
可选地,可以在执行小区切换操作的时候触发第一AI模型激活。
可选地,可以在SSB检测失败超过一定次数的时候触发第一AI模型激活。
可选地,上述第二信息和第三信息中的至少一项可以为上述第一信息或第一信息的子集。其中,基于第二信息中的至少一项确定需要激活所述第一AI模型可以理解为第二信息中的至少一项表明需要触发激活第一AI模型。例如第八信息包括同步信号检测在一定时间的失败次数,当该次数超过一定阈值时,表明同步信号检测的范围可能过分大,这时需要预测可能的同步信号资源位置,促发第一AI模型激活,以使用第一AI模型进行基于AI的小区搜索和选择。
可选地,在一些实施例中,在所述第一设备为终端的情况下,所述方法还包括以下至少一项:
所述第一设备基于第四信息更新所述第一AI模型,或者所述第一设备向第二设备发送第四信息,所述第四信息用于更新所述第一AI模型,其中,所述第四信息包括所述目标结果和所述终端的状态信息中的至少一项;
所述第一设备执行第一操作;
其中,所述第一操作包括以下至少一项:回退到在终端支持的全频段进行小区搜索的方式;触发AI模型的切换;触发所述第一AI模型的重训练;触发所述第一AI模型监督。
本申请实施例中,当第一设备支持对第一AI模型更新(例如第一设备为第一AI模型训练的节点)时,第一设备可以基于第四信息进行第一AI模型的更新,对第一AI模型更新可以理解或替换为对第一AI模型进行微调。当第一设备不支持对第一AI模型更新时,第一设备可以向第二设备发送第四信息,由第二设备对第一AI模型进行更新,然后从第二设备获取更新后的第一AI模型。
可选地,AI模型的切换可以包括更换输入信息或者更换AI算法。
可选地,在一些实施例中,所述第一设备执行第一操作可以是周期性触发,或者条件触发,或者为第一设备自主触发。例如,在一些实施例中,所述第一设备执行第一操作包括:
在满足第一条件的情况下,所述第一设备执行第一操作;
其中,所述第一条件包括以下至少一项:
使用所述第一AI模型进行推理的时长超过第二预设时长;
使用所述第一AI模型未找到合适的小区驻留;
使用所述第一AI模型未成功完成推理。
可选地,在一些实施例中,所述第一AI模型满足以下至少一项:
AI模型的复杂度低于或等于第二阈值;
AI模型的推理时延小于或等于第三预设时长;
AI模型的推理成功率大于或等于第三阈值;
AI模型的推理结果的可靠性大于或等于第四阈值。
本申请实施例中,可以针对不同类型或不同能力的基站或终端定义不同的AI模型复杂度指标要求,比如对于普通的终端,使用的AI模型的复杂度不能超过特定值。
可选地,针对AI模型的推理时延小于或等于第三预设时长可以理解为,使用第一AI模型推理同步信号发送的时频位置或波束方向,或者推理出可以进行驻留的第一小区的时长不能超过第三预设时长。
可选地,针对AI模型的推理成功率大于或等于第三阈值可以理解为,使用第一AI模型推理同步信号发送的时频位置或波束方向,或者推理出可以进行驻留的第一小区的成功率不能超过第三预设时长。
可选地,针对AI模型的推理结果的可靠性大于或等于第四阈值,其中,可靠性可以理解为成功驻留小区的RSRP、RSRQ或RSSI等信号强度信息,这里的RSRP等要求不同于S准则中的RSRP指标要求。例如,测量获得的所述信号强度信息大于等于或小于等于特定值。
可选地,在一些实施例中,所述第一AI模型由所述终端、网络侧设备或服务器独立训练获得,或者所述第一AI模型由所述终端、网络侧设备和服务器中的至少两项联合训练得到。其中,网络侧设备可以包括基站和核心网设备(比如专门用于模型训练的核心网设备)中的至少一项。可选地,该网络侧设备可以为终端最近一次驻留或接入的小区关联的基站或网络侧设备。或者该网络侧设备为发送RRC释放(RRC release)消息的小区关联的基站或网络侧设备。
可选地,针对第一AI模型由所述终端、网络侧设备和服务器中的至少两项联合训练得到的场景可以包括以下至少一项:
终端将模型训练的输出上报给网络侧设备或服务器,网络侧设备或服务器将终端上报信息(即终端模型训练的输出)作为自身模型训练的输入内容之一,进行模型训练。
网络侧设备将模型训练的输出发送给终端或服务器,终端或服务器将网络侧设备发送的信息(即网络侧设备模型训练的输出)作为自身模型训练的输入内容之一,进行模型训练。
终端、网络侧设备和服务器中的至少一项执行线下(offline)模型训练,然后终端、网络侧设备和服务器在实际网络中进行微调。
可选地,终端在进行至少部分模型训练时,模型训练的至少部分输入信息是网络侧设备发送给终端的,该至少部分输入信息发送的信号或信令包括以下至少一项:MAC CE;RRC消息;NAS消息;用户面数据;DCI信息;系统信息块SIB;物理下行控制信道PDCCH的层1信令;物理下行共享信道PDSCH的信息;MSG 2信息;MSG 4信息;MSG B信息。
可选地,网络侧设备在进行至少部分模型训练时,模型训练的至少部分输入信息是终端发送给网络侧设备的,该至少部分输入信息发送的信号或信令包括以下至少一项:MAC CE;RRC消息;NAS消息;用户面数据;MSG 1信息;MSG A信息;MSG 3信息;物理上行控制信道PUCCH的信息;物理上行共享信道PUSCH的信息;物理随机接入信道PRACH的信息;SRS或者其他上行参考信号,比如WUS。
可选地,服务器在进行至少部分模型训练时,模型训练的至少部分输入信息是终端或网络侧设备发送给服务器的,该至少部分输入信息可以通过跨过运营商(Over The Top,OTT)消息来指示。其中,该OTT消息可以为第三方服务商或者第三方服务器或者互联网等提供的消息。
可选地,在一些实施例中,模型训练的输入信息获取方式可以包括:在终端驻留小区、或初次选择小区、或重选小区、或进行RRC连接(connected)态或进入空闲(inactive)态的一段时间后触发至少一次模型训练的输入信息的上报(如终端向网络侧设备发送)或下发(如网络侧设备向终端发送)。
可选地,上述至少一次模型训练的输入信息的上报或下发可以是周期性或半静态的。上述一段时间可以是网络侧设备配置的一个固定时长。
可选地,若第一AI模型最终在网络侧设备或服务器完成训练,则在第一AI模型训练完成后,网络侧设备或服务器可以将训练好的第一AI模型发送给终端,这样由终端完成AI模型的推理,从而可以减少后续信令交互,与此同时可以减少终端进行小区搜索和选择的时延。或者,后续在网络侧设备或服务器执行模型推理,并将推理获得的目标结果发送给终端,这样可以降低对终端能力的需求。
可选地,在一些实施例中,所述第一设备从第二设备接收所述第一AI模型之前,所述方法还包括:
所述第一设备向所述第二设备发送第五信息,所述第五信息用于所述第二设备进行所述第一AI模型的训练;
其中,所述第五信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
本申请实施例中,上述第一设备可以为终端或者网络侧设备,由终端向网络侧设备或服务器发送上述第五信息,或者由网络侧设备向终端或服务器发送第五信息,以进行第一AI模型的训练,第二设备完成第一AI模型训练之后,将训练好的第一AI模型发送给第一设备,由第一设备基于第一AI模型进行推理获得目标结果。
可选地,在一些实施例中,第一设备向第二设备发送第五信息的触发条件可以包括以下至少一项:
终端驻留小区;
终端初次选择小区;
终端重选小区;
终端进入RRC连接态;
终端进入非激活态的时长达到预设时长。
可选地,在一些实施例中,所述方法还包括:
所述第一设备从第二设备接收所述第一AI模型;
其中,所述第二设备为终端、网络侧设备或服务器。
本申请实施例中,第一设备可以仅为第一AI模型的推理设备,进一步也可以为第一AI模型的推理设备和进行第一AI模型联合训练的设备。
可选地,在一些实施例中,所述第一设备从第二设备接收所述第一AI模型之前,所述方法还包括:
所述第一设备将第六信息输入第二AI模型获得第一训练结果,所述第一训练结果用于所述第二设备进行所述第一AI模型的训练;
所述第一设备向第二设备发送所述第一训练结果;
其中,所述第六信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
本申请实施例中,第一设备与第二设备进行第一AI模型联合训练,并最终在第二设备完成第一AI模型的训练,在第二设备完成第一AI模型的训练后,第二设备可以将训练好的第一AI模型发送给第一设备,由第一设备基于第一AI模型进行推理获得目标结果。
可选地,在一些实施例中,第六信息获取的触发条件可以包括以下至少一项:
终端驻留小区;
终端初次选择小区;
终端重选小区;
终端进入RRC连接态;
终端进入非激活态的时长达到预设时长。
可选地,在一些实施例中,所述方法还包括:
在所述第一设备为所述服务器的情况下,所述第一设备从所述终端和网络侧设备中的至少一项获取所述第六信息;
或者,在所述第一设备为所述网络侧设备的情况下,所述第一设备从所述终端获取所述第六信息的至少部分信息;
或者,在所述第一设备为所述终端的情况下,所述第一设备从所述网络侧设备获取所述第六信息的至少部分信息。
可选地,在一些实施例中,所述方法还包括以下任一项:
所述第一设备基于第七信息进行所述第一AI模型的训练,获得所述第一AI模型;
所述第一设备将从第二设备接收到的第二训练结果输入到第三AI模型进行AI训练,获得所述第一AI模型;
其中,所述第七信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
本申请实施例中,第一设备既可以为第一AI模型的推理设备,又可以为第一AI模型的训练设备。具体的,当由第一设备独立训练第一AI模型时,可以将第七信息输入到待训练的AI模型进行训练,获得第一AI模型。当第一设备和第二设备联合训练第一AI模型时,第一设备可以从第二设备接收第二训练结果,并将第二训练结果作为第三AI模型的输入之一进行模型训练,获得第一AI模型。
需要说明的是,用于第一AI模型训练的输入信息(如上述第五信息、第六信息或第七信息)可以为最近N次终端成功进行小区驻留时对应的输入信息,或者是最近N次检测到同步信号(如CD-SSB)时对应的所述输入信息。
进一步地,用于第一AI模型推理的输入信息(如第一信息或第二信息)可以为用于第一AI模型训练的输入信息,或者为用于第一AI模型训练的输入信息的子集。
可选地,在一些实施例中,所述第一AI模型的训练的输入还包括目标标签,所述目标标签包括以下至少一项:
终端是否检测到同步信号;
终端检测到同步信号的时长;
终端检测到的同步信号的频域位置;
终端检测到的同步信号的时域位置;
终端检测到的同步信号的波束方向;
终端在预设频域位置检测到同步信号;
终端在预设时域位置检测到同步信号;
终端在预设波束方向检测到同步信号;
终端测量到的驻留小区的信号质量;
终端成功驻留的小区的标识;
终端成功驻留的小区所在的跟踪区域;
终端成功驻留;
终端成功驻留到预设小区;
第一时刻到第二时刻的时长,所述第一时刻为发起小区搜索的时刻或激活所述第一AI模型的时刻,所述第二时刻为所述终端成功驻留的时刻。
可选地,上述信号质量可以包括以下至少一项:RSSI,PSS或SSS检测信号强度,信道估计SNR、SSB-RSRP、L1 RSRP和L3 RSRP。
可选地,针对终端测量到的驻留小区的信号质量,可以包括以下至少一项:
终端在预设频域位置测量到驻留小区的信号质量;
终端在预设时域位置测量到驻留小区的信号质量;
终端在预设波束方向测量到驻留小区的信号质量。
可选地,在一些实施例中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:
所述第一设备基于第一配置信息进行周期性触发所述第一AI模型的训练;
其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
可选地,在一些实施例中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:
所述第一设备基于半静态触发方式触发所述第一AI模型的训练;
其中,所述半静态触发方式包括以下至少一项:
基于第二配置信息进行半静态触发;
通过物理控制信息激活或去激活模型的半静态训练;
其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
可选地,在一些实施例中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:
所述第一设备基于目标事件触发所述第一AI模型的训练;
其中,所述目标事件包括以下至少一项:
最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;
最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;
所述终端移动到小区边缘或预设位置;
所述终端的移动速度高于或等于第一门限;
所述终端的移动速度低于或等于第二门限;
小区当前接入或驻留的终端的数量高于或等于第三门限;
小区当前接入或驻留的终端的数量低于或等于第四门限;
使用所述第一AI模型推理失败;
使用所述第一AI模型连续推理失败的次数达到第五门限;
使用所述第一AI模型推理失败的次数达到第六门限;
使用所述第一AI模型进行了推理;
所述终端重选到第二小区;
所述终端的跟踪区域发生变化;
所述终端的外部环境发生变化;
所述终端移动到第二小区;
所述终端移动到新的跟踪区域;
所述终端移动到新的地理位置;
所述终端的移动速度变化量大于或等于第七门限;
所述终端的RSRP测量值发生变化;
距离上一次训练的时间达到第四预设时长;
用于触发重新训练的定时器超时;
发生了连续M次模型监督,M为大于1的整数;
发生了L次模型监督,L为正整数;
同步信号块的检测失败次数达到第八门限。
本申请实施例中,所述第一AI模型的训练可以理解为所述第一AI模型的初始训练或重训练。上述目标事件可以理解或替换为目标条件。
可选地,上述外部环境发生变化可以理解为通过终端的传感器获得环境的变化信息所确定的环境。例如可以包括但不限于以下环境变化:处于城市或者农村,处于室内或室外,处于高速运动环境或者低速运动环境等。
可选地,所述终端移动到第二小区可以理解或替换为:终端移动到新的小区,或者终端移动到上次接入的服务小区之外的其他小区。
可选地,所述终端移动到新的跟踪区域可以理解为:终端移动到上次接入的服务小区之所在的跟踪区域之外的其他跟踪区域。
可选地,所述终端移动到新的地理位置可以理解为:终端移动到上次接入的服务小区之所在的地理位置之外的其他地理位置。例如终端在A地关机,移动到B地后进行开机,此时可以认为终端移动到新的地理位置,B地为新的地理位置,A地和B地可以为不同的城市,或者不同的省份,或者不同的国家。
可选地,所述终端的移动速度变化量大于或等于第七门限可以理解为:终端在短时间内移动速度明显变化,例如,从250kM/h下降到3km/H,或者从3km/H上升到250kM/h。
可选地,在一些实施例中,所述方法还包括:
所述第一设备基于第八信息确定所述第一AI模型训练完成;
其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
本申请实施例中,可以在第八信息中的部分信息大于或等于对应的预设门限的情况下,确定第一AI模型训练完成,也可以在第八信息中的部分信息小于或等于对应的预设门限的情况下,确定第一AI模型训练完成。
例如,终端成功检测到同步信号的概率大于或等于X%时确定第一AI模型训练完成,X为一个特定的门限值。进一步地,终端成功检测到同步信号的概率大于或等于X%可以理解或替换为终端在特定的频域位置、特定的时域位置和特定的波束方向中的至少一项中成功检测到同步信号的概率大于或等于X%。
又例如,终端检测到同步信号的时长小于或等于M时确定第一AI模型训练完成,M为一个特定的门限值。进一步地,终端检测到同步信号的时长小于或等于M可以理解或替换为终端在特定的频域位置、特定的时域位置和特定的波束方向中的至少一项中终端检测到同步信号的时长小于或等于M。
可选地,在一些实施例中,上述第八信息还包括以下至少一项:
损失函数;
模型训练的迭代次数;
模型训练微调的迭代次数。
本申请实施例中,在损失函数满足预定义的需求指标或取值时,确定第一AI模型训练完成,例如,训练的误差小于预定义的门限值。其中,损失函数可以包括以下至少一项:
预测值和真实值的均方误差或归一化均方误差;
预测值和真实值的平均绝对误差。
可选地,如果终端执行模型训练,需要考虑终端的类型,不同类型的终端可能具有不同的AI模型训练能力。
针对不同类型的终端,用于模型训练的输入或者标签信息不同。比如,对于能力较弱的终端设备用于模型训练的输入信息应该较少。
针对不同类型的终端,模型训练的执行方式不同。比如,对于能力较弱的终端设备可以考虑仅在网络侧执行模型训练,或者终端侧只执行联合模型训练中的小部分(比如,涉及到用户隐私的数据的模型训练可以在终端侧执行)
针对不同类型的终端,用于模型训练的人工智能模型不同。比如,对于能力较弱的设备可能无法应用过于复杂的人工智能模型。
可选地,在一些实施例中,所述方法还包括:
所述第一设备基于第三配置信息对所述第一AI模型进行监督;
其中,所述第三配置信息包括以下至少一项:
需要进行模型监督的AI模型标识;
模型监督的周期;
模型监督的时长;
模型监督的检测窗口相关信息;
模型监督的触发条件;
模型监督的指标;
其中,所述模型监督的指标包括以下至少一项:预测值和真实值之间的误差类信息或精度信息;通信系统性能;第一AI模型的模型相关信息。
本申请实施例中,当推理环境与训练环境差异较大时,基于AI的小区搜索性能将会变得很差,即会产生失配。因此,需要对基于AI的小区搜索的实际推理性能进行监督,并根据监督结果触发相应的调整操作。
可选地,所述AI模型标识可以理解或替换为以下至少一项:
AI结构标识;
AI算法标识;
AI模型关联的特定数据集的标识;
AI相关的特定场景、环境、信道特征、设备的标识;
AI相关的功能、特性、能力或模块的标识,
可选地,上述检测窗口相关信息包括以下至少一项:检测窗口的时长和进行检测的样本数。
可选地,上述误差类信息可以包括但不限于以下至少一项:平均误差(mean error),均方误差(mean square error),归一化均方误差(normalized mean square error),平均绝对误差(mean absolute error),交叉熵损失(Cross-entropy loss),以及均方根误差(Root Mean Square Error)。
可选地,上述精度信息可以包括但不限于以下至少一项:相似度,余弦相似度,相关性,相关系数,以及曲线下面积(Area Under Curve,AUC)分数(AUC score)。
可选地,上述通信系统性能可以包括但不限于以下至少一项:小区搜索时延、小区驻留成功率和定时(timing)误差。其中,小区搜索时延、小区驻留成功率和定时误差的统计是在监测窗内统计得到的。
可选地,上述第一AI模型的模型相关信息可以包括但不限于以下至少一项:第一AI模型的运行时长、第一AI模型的CPU占用和第一AI模型的内存空间占用。
可选地,在一些实施例中,所述模型监督的触发条件基于以下至少一项确定:模型监督的指标;所述第一AI模型的推理结果;模型推理的指标。
例如,在一些实施例中,可以在模型监督的指标不满足要求时,或者在模型监督的指标不满足要求一段时间后,触发对所述第一AI模型进行监督。
例如,在一些实施例中,可以在所述第一AI模型的推理结果标不满足要求时,或者在所述第一AI模型的推理结果标不满足要求一段时间后,触发对所述第一AI模型进行监督。可选地,可以在满足以下至少一项的情况下,可以认为所述第一AI模型的推理结果标不满足要求:
终端使用第一AI模型推理超过M时长(即第五预设时长)后,仍然没有找到合适的小区驻留;
终端使用第一AI模型没有成功完成AI推理。
例如,在一些实施例中,在第一AI模型的模型推理的指标中的至少一项不满足要求时,或者在第一AI模型的模型推理的指标中的至少一项不满足要求一段时间后,触发对所述第一AI模型进行监督。例如,在终端成功驻留小区的RSRP不满足要求,触发对所述第一AI模型进行监督。
可选地,在一些实施例中,所述方法还包括:
所述第一设备与第二设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:
终端是否支持所述第一AI模型训练的能力;
终端是否支持基于所述第一AI模型进行AI推理的能力;
终端是否支持上报辅助信息用于所述第一AI模型训练的能力;
终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持所述第一AI模型训练的能力;
网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;
网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;
服务器是否支持所述第一AI模型训练的能力;
服务器是否支持基于所述第一AI模型进行AI推理的能力。
可选地,在一些实施例中,所述方法还包括:
所述第一设备基于以下至少一项确定所述第二设备对应的所述目标能力信息:
所述第二设备的设备类型;
第二设备的网络类型;
第二设备发送的参考信号;
所述第二设备发送的控制信息;
所述第二设备发送的RRC信令;
所述第一设备与所述第二设备之间的接口消息。
可选地,目标能力信息的确定可以取决于设备类型,例如,第二设备为终端,不同的终端类型引入不同的目标能力信息。此时可以通过上报终端的类型,即可隐含指示终端的相关能力。
可选地,目标能力信息的确定可以取决于网络类型。例如,不同网络类型(NTN或TN)引入不同的目标能力信息,此时通过传输网络类型,即可隐含指示第二设备(如网络侧设备或服务器)的相关能力。
可选地,在可以在一种或多种参考信号指示目标能力信息,例如通过PRACH资源指示第二设备的目标能力信息。
可选地,在第二设备为终端时,终端可以通过上行控制信息(Uplink Control Information,UCI)指示目标能力信息,比如物理层控制信息,如针对终端的上报到网络侧设备的UCI。
所述第一设备与所述第二设备之间的接口消息可以包括以下至少一项:
终端与服务器的接口消息,具体的,可以为终端与服务器的特定接口消息,该特定接口消息可以为针对特定AI模型相关信息或针对所有AI模型的相关消息;
终端与网络侧设备的接口消息,具体的,可以为终端与网络侧设备的特定接口消息,该特定接口消息可以为针对特定AI模型相关信息或针对所有AI模型的相关消息;
服务器与网络侧设备的接口消息,具体的,可以为服务器与网络侧设备的特定接口消息,该特定接口消息可以为针对特定AI模型相关信息或针对所有AI模型的相关消息。
参照图5,本申请实施例还提供了一种小区驻留处理方法,如图5所示,该小区驻留处理方法包括:
步骤501,第二设备执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
可选地,所述第二设备进行第一AI模型的训练,获得第一AI模型包括:
所述第二设备基于第五信息进行第一AI模型的训练,获得第一AI模型;
其中,所述第五信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述第二设备基于第五信息第一AI模型的训练,获得第一AI模型之前,所述方法还包括:
在所述第二设备为所述服务器的情况下,所述第二设备从所述终端和网络侧设备中的至少一项获取所述第五信息;
或者,在所述第二设备为所述网络侧设备的情况下,所述第二设备从所述终端获取所述第五信息的至少部分信息;
或者,在所述第二设备为所述终端的情况下,所述第二设备从所述网络侧设备获取所述第五信息的至少部分信息。
可选地,所述方法还包括:
所述第二设备从所述第一设备接收第四信息;
所述第二设备基于所述第四信息更新所述第一AI模型;
所述第二设备向所述第一设备发送更新后的第一AI模型。
可选地,所述第二设备进行第一AI模型的训练,获得第一AI模型包括:
所述第二设备从所述第一设备接收第一训练结果;
所述第二设备将所述第一训练结果输入到第四AI模型进行所述第一AI模型的训练,获得所述第一AI模型;
所述第二设备向所述第一设备发送所述第一AI模型。
可选地,所述第二设备进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果包括:
所述第二设备将第九信息输入第五AI模型,获得第二训练结果;
所述第二设备向所述第一设备发送所述第二训练结果;
其中,所述第九信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述第二设备将第九信息输入第五AI模型,获得第二训练结果之前,所述方法还包括:
在所述第二设备为所述服务器的情况下,所述第二设备从所述终端和网络侧设备中的至少一项获取所述第九信息;
或者,在所述第二设备为所述网络侧设备的情况下,所述第二设备从所述终端获取所述第九信息的至少部分信息;
或者,在所述第二设备为所述终端的情况下,所述第二设备从所述网络侧设备获取所述第九信息的至少部分信息。
可选地,在所述第二设备为终端的情况下,所述方法还包括:
所述第二设备从所述第一设备接收所述目标结果。
可选地,在所述第二设备为终端的情况下,所述方法还包括:
所述第二设备向网络侧设备或服务器发送目标指示信息,所述目标指示信息用于激活所述第一AI模型,且所述目标指示信息用于指示以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息。
可选地,所述方法还包括:
所述第二设备基于第一配置信息进行周期性触发所述第一AI模型的训练;
其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
可选地,所述方法还包括:
所述第二设备基于半静态触发方式触发所述第一AI模型的训练;
其中,所述半静态触发方式包括以下至少一项:
基于第二配置信息进行半静态触发;
通过物理控制信息激活或去激活模型的半静态训练;
其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
可选地,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:
所述第一设备基于目标事件触发所述第一AI模型的训练;
其中,所述目标事件包括以下至少一项:
最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;
最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;
所述终端移动到小区边缘或预设位置;
所述终端的移动速度高于或等于第一门限;
所述终端的移动速度低于或等于第二门限;
小区当前接入或驻留的终端的数量高于或等于第三门限;
小区当前接入或驻留的终端的数量低于或等于第四门限;
使用所述第一AI模型推理失败;
使用所述第一AI模型连续推理失败的次数达到第五门限;
使用所述第一AI模型推理失败的次数达到第六门限;
使用所述第一AI模型进行了推理;
所述终端重选到第二小区;
所述终端的跟踪区域发生变化;
所述终端的外部环境发生变化;
所述终端移动到第二小区;
所述终端移动到新的跟踪区域;
所述终端移动到新的地理位置;
所述终端的移动速度变化量大于或等于第七门限;
所述终端的RSRP测量值发生变化;
距离上一次训练的时间达到第四预设时长;
用于触发重新训练的定时器超时;
发生了连续M次模型监督,M为大于1的整数;
发生了L次模型监督,L为正整数;
同步信号块的检测失败次数达到第八门限。
可选地,所述方法还包括:
所述第一设备基于第八信息确定所述第一AI模型训练完成;
其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
可选地,所述方法还包括:
所述第二设备与第一设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:
终端是否支持所述第一AI模型训练的能力;
终端是否支持基于所述第一AI模型进行AI推理的能力;
终端是否支持上报辅助信息用于所述第一AI模型训练的能力;
终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持所述第一AI模型训练的能力;
网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;
网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;
服务器是否支持所述第一AI模型训练的能力;
服务器是否支持基于所述第一AI模型进行AI推理的能力。
可选地,所述方法还包括:
所述第二设备基于以下至少一项确定所述第一设备对应的所述目标能力信息:
所述第一设备的设备类型;
第一设备的网络类型;
第二设备发送的参考信号;
所述第一设备发送的控制信息;
所述第一设备发送的RRC信令;
所述第一设备与所述第二设备之间的接口消息。
本申请实施例提供的小区驻留处理方法,执行主体可以为小区驻留处理装置。本申请实施例中以小区驻留处理装置执行小区驻留处理方法为例,说明本申请实施例提供的小区驻留处理装置。
本申请实施例提供一种小区驻留处理装置,作为一种示例,小区驻留处理装置可以是通信设备或通信设备中的部件,例如芯片。该通信设备可以是终端、网络侧设备或服务器等。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,网络侧设备可以包括但不限于上述所列举的网络侧设备12的类型,本申请实施例不作具体限定。
小区驻留处理装置包括接收模块、发送模块和处理模块。其中,接收模块、发送模块和处理模块可以是通过软件实现,也可以通过硬件实现。当通过硬件实现时,处理模块可以由处理器实现,示例性的,处理器可以包括通用处理器、专用处理器等,例如包括中央处理单元(Central Processing Unit,CPU)、微处理器、数字信号处理器(Digital Signal Processor,DSP)、人工智能(Artificial Intelligent,AI)处理器、图形处理器(Graphics Processing Unit,GPU)、专用集成电路(Application Specific Integrated Circuit,ASIC)、网络处理器(Network Processor,NP)、现场可编程门阵列(Field Programmable Gate Array,FPGA)或者其他可编程逻辑器件、门电路、晶体管、分立硬件组件等。接收模块和发送模块可以由通信接口实现,通信接口可以包括收发器、管脚、电路、总线、射频单元等其中一种或多种。
具体的,参见图6,小区驻留处理装置600应用于第一设备,小区驻留处理装置600包括:
第一处理模块601,用于基于第一人工智能AI模型获得目标结果;
其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
所述第一设备基于第一人工智能AI模型获得目标结果包括:
所述第一设备将第一信息输入第一AI模型,获得所述目标结果;
其中,所述第一信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述小区驻留处理装置600还包括:第一发送模块,用于在所述第一设备为网络侧设备或者服务器的情况下,向所述终端发送所述目标结果。
可选地,在所述第一设备为终端的情况下,所述第一处理模块601还用于基于第一激活条件激活所述第一AI模型,所述第一激活条件包括以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;用于激活所述第一AI模型的定时器超时;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;
或者,在所述第一设备为网络侧设备或服务器的情况下,所述第一处理模块601还用于基于第二激活条件激活所述第一AI模型,所述第二激活条件包括以下至少一项:
从终端接收到目标指示信息,所述目标指示信息用于指示以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;
用于激活所述第一AI模型的定时器超时;
基于第三信息中的至少一项确定需要激活所述第一AI模型,所述第三信息为所述第一AI模型的至少部分输入信息。
可选地,在所述第一设备为终端的情况下,所述第一处理模块601还用于执行以下至少一项:
基于第四信息更新所述第一AI模型,或者所述第一设备向第二设备发送第四信息,所述第四信息用于更新所述第一AI模型,其中,所述第四信息包括所述目标结果和所述终端的状态信息中的至少一项;
执行第一操作;
其中,所述第一操作包括以下至少一项:回退到在终端支持的全频段进行小区搜索的方式;触发AI模型的切换;触发所述第一AI模型的重训练;触发所述第一AI模型监督。
可选地,所述第一处理模块601具体用于在满足第一条件的情况下,执行第一操作;
其中,所述第一条件包括以下至少一项:
使用所述第一AI模型进行推理的时长超过第二预设时长;
使用所述第一AI模型未找到合适的小区驻留;
使用所述第一AI模型未成功完成推理。
可选地,所述第一AI模型满足以下至少一项:
AI模型的复杂度低于或等于第二阈值;
AI模型的推理时延小于或等于第三预设时长;
AI模型的推理成功率大于或等于第三阈值;
AI模型的推理结果的可靠性大于或等于第四阈值。
可选地,所述第一AI模型由所述终端、网络侧设备或服务器独立训练获得,或者所述第一AI模型由所述终端、网络侧设备和服务器中的至少两项联合训练得到。
可选地,所述小区驻留处理装置600还包括:第一接收模块,用于从第二设备接收所述第一AI模型;
其中,所述第二设备为终端、网络侧设备或服务器。
可选地,所述小区驻留处理装置600还包括:第一发送模块,用于向所述第二设备发送第五信息,所述第五信息用于所述第二设备进行所述第一AI模型的训练;
其中,所述第五信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述小区驻留处理装置600还包括:第一发送模块,其中,
所述第一处理模块601用于将第六信息输入第二AI模型获得第一训练结果,所述第一训练结果用于所述第二设备进行所述第一AI模型的训练;
所述第一发送模块,用于向第二设备发送所述第一训练结果;
其中,所述第六信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述小区驻留处理装置600还包括:第一接收模块,用于执行以下至少一项:
在所述第一设备为所述服务器的情况下,从所述终端和网络侧设备中的至少一项获取所述第六信息;
或者,在所述第一设备为所述网络侧设备的情况下,从所述终端获取所述第六信息的至少部分信息;
或者,在所述第一设备为所述终端的情况下,从所述网络侧设备获取所述第六信息的至少部分信息。
可选地,所述第一处理模块601还用于执行以下至少一项:
基于第七信息进行所述第一AI模型的训练,获得所述第一AI模型;
将从第二设备接收到的第二训练结果输入到第三AI模型进行AI训练,获得所述第一AI模型;
其中,所述第七信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述第一AI模型的训练的输入还包括目标标签,所述目标标签包括以下至少一项:
终端是否检测到同步信号;
终端检测到同步信号的时长;
终端检测到的同步信号的频域位置;
终端检测到的同步信号的时域位置;
终端检测到的同步信号的波束方向;
终端在预设频域位置检测到同步信号;
终端在预设时域位置检测到同步信号;
终端在预设波束方向检测到同步信号;
终端测量到的驻留小区的信号质量;
终端成功驻留的小区的标识;
终端成功驻留的小区所在的跟踪区域;
终端成功驻留;
终端成功驻留到预设小区;
第一时刻到第二时刻的时长,所述第一时刻为发起小区搜索的时刻或激活所述第一AI模型的时刻,所述第二时刻为所述终端成功驻留的时刻。
可选地,所述第一处理模块601还用于:
基于第一配置信息进行周期性触发所述第一AI模型的训练;
其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
可选地,所述第一处理模块601还用于:基于半静态触发方式触发所述第一AI模型的训练;
其中,所述半静态触发方式包括以下至少一项:
基于第二配置信息进行半静态触发;
通过物理控制信息激活或去激活模型的半静态训练;
其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
可选地,所述第一处理模块601还用于:基于目标事件触发所述第一AI模型的训练;
其中,所述目标事件包括以下至少一项:
最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;
最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;
所述终端移动到小区边缘或预设位置;
所述终端的移动速度高于或等于第一门限;
所述终端的移动速度低于或等于第二门限;
小区当前接入或驻留的终端的数量高于或等于第三门限;
小区当前接入或驻留的终端的数量低于或等于第四门限;
使用所述第一AI模型推理失败;
使用所述第一AI模型连续推理失败的次数达到第五门限;
使用所述第一AI模型推理失败的次数达到第六门限;
使用所述第一AI模型进行了推理;
所述终端重选到第二小区;
所述终端的跟踪区域发生变化;
所述终端的外部环境发生变化;
所述终端移动到第二小区;
所述终端移动到新的跟踪区域;
所述终端移动到新的地理位置;
所述终端的移动速度变化量大于或等于第七门限;
所述终端的RSRP测量值发生变化;
距离上一次训练的时间达到第四预设时长;
用于触发重新训练的定时器超时;
发生了连续M次模型监督,M为大于1的整数;
发生了L次模型监督,L为正整数;
同步信号块的检测失败次数达到第八门限。
可选地,所述第一处理模块601还用于:基于第八信息确定所述第一AI模型训练完成;
其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
可选地,所述第一处理模块601还用于基于第三配置信息对所述第一AI模型进行监督;
其中,所述第三配置信息包括以下至少一项:
需要进行模型监督的AI模型标识;
模型监督的周期;
模型监督的时长;
模型监督的检测窗口相关信息;
模型监督的触发条件;
模型监督的指标;
其中,所述模型监督的指标包括以下至少一项:预测值和真实值之间的误差类信息或精度信息;通信系统性能;第一AI模型的模型相关信息。
可选地,所述模型监督的触发条件基于以下至少一项确定:模型监督的指标;所述第一AI模型的推理结果;模型推理的指标。
可选地,所述小区驻留处理装置600还包括:传输模块,用于与第二设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:
终端是否支持所述第一AI模型训练的能力;
终端是否支持基于所述第一AI模型进行AI推理的能力;
终端是否支持上报辅助信息用于所述第一AI模型训练的能力;
终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持所述第一AI模型训练的能力;
网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;
网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;
服务器是否支持所述第一AI模型训练的能力;
服务器是否支持基于所述第一AI模型进行AI推理的能力。
可选地,所述第一处理模块601,还用于基于以下至少一项确定所述第二设备对应的所述目标能力信息:
所述第二设备的设备类型;
第二设备的网络类型;
第二设备发送的参考信号;
所述第二设备发送的控制信息;
所述第二设备发送的RRC信令;
所述第一设备与所述第二设备之间的接口消息。
具体的,参见图7,小区驻留处理装置700应用于第二设备,小区驻留处理装置700包括:
执行模块701,用于执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
可选地,所述执行模块701具体用于:基于第五信息进行第一AI模型的训练,获得第一AI模型;
其中,所述第五信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述执行模块701包括:第二接收模块,用于执行以下至少一项:
在所述第二设备为所述服务器的情况下,从所述终端和网络侧设备中的至少一项获取所述第五信息;
或者,在所述第二设备为所述网络侧设备的情况下,从所述终端获取所述第五信息的至少部分信息;
或者,在所述第二设备为所述终端的情况下,从所述网络侧设备获取所述第五信息的至少部分信息。
可选地,所述执行模块701包括:
第二接收模块,用于从所述第一设备接收第四信息;
第二处理模块,用于基于所述第四信息更新所述第一AI模型;
第二发送模块,用于向所述第一设备发送更新后的第一AI模型。
可选地,所述执行模块701包括:
第二接收模块,用于从所述第一设备接收第一训练结果;
第二处理模块,用于将所述第一训练结果输入到第四AI模型进行第一AI模型的训练,获得所述第一AI模型;
第二发送模块,用于向所述第一设备发送所述第一AI模型。
可选地,所述执行模块701包括:
第二处理模块,用于将第九信息输入第五AI模型,获得第二训练结果;
第二发送模块,用于向所述第一设备发送所述第二训练结果;
其中,所述第九信息包括以下至少一项:
时间信息;
终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;
在第一预设时间段内同步信号检测失败的次数;
突发的事件;
同步信号的模式信息;
同步信号的频域特征;
同步信号的时域特征;
同步信号的空域特征。
可选地,所述执行模块701包括第二接收模块,用于执行以下至少一项:
在所述第二设备为所述服务器的情况下,从所述终端和网络侧设备中的至少一项获取所述第九信息;
或者,在所述第二设备为所述网络侧设备的情况下,从所述终端获取所述第九信息的至少部分信息;
或者,在所述第二设备为所述终端的情况下,从所述网络侧设备获取所述第九信息的至少部分信息。
可选地,所述执行模块701包括第二接收模块,用于在所述第二设备为终端的情况下,从所述第一设备接收所述目标结果。
可选地,所述执行模块701包括第二发送模块,用于在所述第二设备为终端的情况下,向网络侧设备或服务器发送目标指示信息,所述目标指示信息用于激活所述第一AI模型,且所述目标指示信息用于指示以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息。
可选地,所述执行模块701具体用于基于第一配置信息进行周期性触发所述第一AI模型的训练;
其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
可选地,所述执行模块701具体用于基于半静态触发方式触发所述第一AI模型的训练;
其中,所述半静态触发方式包括以下至少一项:
基于第二配置信息进行半静态触发;
通过物理控制信息激活或去激活模型的半静态训练;
其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
可选地,所述执行模块701具体用于基于目标事件触发所述第一AI模型的训练;
其中,所述目标事件包括以下至少一项:
最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;
最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;
所述终端移动到小区边缘或预设位置;
所述终端的移动速度高于或等于第一门限;
所述终端的移动速度低于或等于第二门限;
小区当前接入或驻留的终端的数量高于或等于第三门限;
小区当前接入或驻留的终端的数量低于或等于第四门限;
使用所述第一AI模型推理失败;
使用所述第一AI模型连续推理失败的次数达到第五门限;
使用所述第一AI模型推理失败的次数达到第六门限;
使用所述第一AI模型进行了推理;
所述终端重选到第二小区;
所述终端的跟踪区域发生变化;
所述终端的外部环境发生变化;
所述终端移动到第二小区;
所述终端移动到新的跟踪区域;
所述终端移动到新的地理位置;
所述终端的移动速度变化量大于或等于第七门限;
所述终端的RSRP测量值发生变化;
距离上一次训练的时间达到第四预设时长;
用于触发重新训练的定时器超时;
发生了连续M次模型监督,M为大于1的整数;
发生了L次模型监督,L为正整数;
同步信号块的检测失败次数达到第八门限。
可选地,所述执行模块701具体用于基于第八信息确定所述第一AI模型训练完成;
其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
可选地,所述执行模块701还用于:与第一设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:
终端是否支持所述第一AI模型训练的能力;
终端是否支持基于所述第一AI模型进行AI推理的能力;
终端是否支持上报辅助信息用于所述第一AI模型训练的能力;
终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持所述第一AI模型训练的能力;
网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;
网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;
网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;
服务器是否支持所述第一AI模型训练的能力;
服务器是否支持基于所述第一AI模型进行AI推理的能力。
可选地,所述执行模块701还包括:
第二处理模块,用于基于以下至少一项确定所述第一设备对应的所述目标能力信息:
所述第一设备的设备类型;
第一设备的网络类型;
第二设备发送的参考信号;
所述第一设备发送的控制信息;
所述第一设备发送的RRC信令;
所述第一设备与所述第二设备之间的接口消息。
本申请实施例提供的小区驻留处理装置能够实现图4至图5的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,该程序或指令被处理器801执行时实现上述小区驻留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或图5所示方法实施例中的步骤。该终端实施例与上述第一设备侧或第二设备侧方法实施例对应,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。该终端可以是图6或图7所示的小区驻留处理装置。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、收发器、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;可选的,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,在终端为第一设备时,处理器910用于基于第一人工智能AI模型获得目标结果;
其中,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
在终端为第二设备时,射频单元901用于执行以下至少一项:
向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;
第二操作;
其中,所述第二操作包括以下至少一项:
进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;
进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;
其中,所述第一设备为网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
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上述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4或图5所示的方法实施例的步骤。该网络侧设备实施例与上述第一设备侧或第二设备侧方法实施例对应,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,该网络侧设备可以是图6或图7所示的小区驻留处理装置。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侧设备操作。
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6或图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上述小区驻留处理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,实现上述小区驻留处理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/程序产品包括计算机指令,所述计算机程序/程序产品被至少一个处理器执行以实现上述小区驻留处理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无线通信系统,包括:第一设备及第二设备,所述第一设备可用于执行如上所述第一设备侧的小区驻留处理方法的步骤,所述第二设备可用于执行如上所述第二设备侧的小区驻留处理方法的步骤。
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 (42)
- 一种小区驻留处理方法,包括:第一设备基于第一人工智能AI模型获得目标结果;其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
- 根据权利要求1所述的方法,其中,所述第一设备基于第一人工智能AI模型获得目标结果包括:所述第一设备将第一信息输入第一AI模型,获得所述目标结果;其中,所述第一信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求1所述的方法,其中,在所述第一设备为网络侧设备或者服务器的情况下,所述方法还包括:所述第一设备向所述终端发送所述目标结果。
- 根据权利要求1所述的方法,其中,所述方法还包括:在所述第一设备为终端的情况下,所述第一设备基于第一激活条件激活所述第一AI模型,所述第一激活条件包括以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;用于激活所述第一AI模型的定时器超时;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;或者,在所述第一设备为网络侧设备或服务器的情况下,所述第一设备基于第二激活条件激活所述第一AI模型,所述第二激活条件包括以下至少一项:从终端接收到目标指示信息,所述目标指示信息用于指示以下至少一项:所述终端开机;执行初始小区搜索操作;执行初始小区选择操作;通过在终端支持的全频段进行小区搜索的方式进行初始搜索达到第一预设时长;在第二预设时间段内没有驻留到第二小区;执行小区重选操作;执行小区切换操作;同步信号块SSB检测失败的次数大于或等于第一阈值;基于第二信息中的至少一项确定需要激活所述第一AI模型,所述第二信息为所述第一AI模型的至少部分输入信息;用于激活所述第一AI模型的定时器超时;基于第三信息中的至少一项确定需要激活所述第一AI模型,所述第三信息为所述第一AI模型的至少部分输入信息。
- 根据权利要求1所述的方法,其中,在所述第一设备为终端的情况下,所述方法还包括以下至少一项:所述第一设备基于第四信息更新所述第一AI模型,或者所述第一设备向第二设备发送第四信息,所述第四信息用于更新所述第一AI模型,其中,所述第四信息包括所述目标结果和所述终端的状态信息中的至少一项;所述第一设备执行第一操作;其中,所述第一操作包括以下至少一项:回退到在终端支持的全频段进行小区搜索的方式;触发AI模型的切换;触发所述第一AI模型的重训练;触发所述第一AI模型监督。
- 根据权利要求5所述的方法,其中,所述第一设备执行第一操作包括:在满足第一条件的情况下,所述第一设备执行第一操作;其中,所述第一条件包括以下至少一项:使用所述第一AI模型进行推理的时长超过第二预设时长;使用所述第一AI模型未找到合适的小区驻留;使用所述第一AI模型未成功完成推理。
- 根据权利要求1所述的方法,其中,所述方法还包括:所述第一设备从第二设备接收所述第一AI模型;其中,所述第二设备为终端、网络侧设备或服务器。
- 根据权利要求7所述的方法,其中,所述第一设备从第二设备接收所述第一AI模型之前,所述方法还包括:所述第一设备向所述第二设备发送第五信息,所述第五信息用于所述第二设备进行所述第一AI模型的训练;其中,所述第五信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求7所述的方法,其中,所述第一设备从第二设备接收所述第一AI模型之前,所述方法还包括:所述第一设备将第六信息输入第二AI模型获得第一训练结果,所述第一训练结果用于所述第二设备进行所述第一AI模型的训练;所述第一设备向第二设备发送所述第一训练结果;其中,所述第六信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求9所述的方法,其中,所述方法还包括:在所述第一设备为所述服务器的情况下,所述第一设备从所述终端和网络侧设备中的至少一项获取所述第六信息;或者,在所述第一设备为所述网络侧设备的情况下,所述第一设备从所述终端获取所述第六信息的至少部分信息;或者,在所述第一设备为所述终端的情况下,所述第一设备从所述网络侧设备获取所述第六信息的至少部分信息。
- 根据权利要求1所述的方法,其中,所述方法还包括以下任一项:所述第一设备基于第七信息进行所述第一AI模型的训练,获得所述第一AI模型;所述第一设备将从第二设备接收到的第二训练结果输入到第三AI模型进行所述第一AI训练,获得所述第一AI模型;其中,所述第七信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求8至11任一项所述的方法,其中,所述第一AI模型的训练的输入还包括目标标签,所述目标标签包括以下至少一项:终端是否检测到同步信号;终端检测到同步信号的时长;终端检测到的同步信号的频域位置;终端检测到的同步信号的时域位置;终端检测到的同步信号的波束方向;终端在预设频域位置检测到同步信号;终端在预设时域位置检测到同步信号;终端在预设波束方向检测到同步信号;终端测量到的驻留小区的信号质量;终端成功驻留的小区的标识;终端成功驻留的小区所在的跟踪区域;终端成功驻留;终端成功驻留到预设小区;第一时刻到第二时刻的时长,所述第一时刻为发起小区搜索的时刻或激活所述第一AI模型的时刻,所述第二时刻为所述终端成功驻留的时刻。
- 根据权利要求1至12任一项所述的方法,其中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:所述第一设备基于第一配置信息进行周期性触发所述第一AI模型的训练;其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
- 根据权利要求1至12任一项所述的方法,其中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:所述第一设备基于半静态触发方式触发所述第一AI模型的训练;其中,所述半静态触发方式包括以下至少一项:基于第二配置信息进行半静态触发;通过物理控制信息激活或去激活模型的半静态训练;其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
- 根据权利要求1至12任一项所述的方法,其中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:所述第一设备基于目标事件触发所述第一AI模型的训练;其中,所述目标事件包括以下至少一项:最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;所述终端移动到小区边缘或预设位置;所述终端的移动速度高于或等于第一门限;所述终端的移动速度低于或等于第二门限;小区当前接入或驻留的终端的数量高于或等于第三门限;小区当前接入或驻留的终端的数量低于或等于第四门限;使用所述第一AI模型推理失败;使用所述第一AI模型连续推理失败的次数达到第五门限;使用所述第一AI模型推理失败的次数达到第六门限;使用所述第一AI模型进行了推理;所述终端重选到第二小区;所述终端的跟踪区域发生变化;所述终端的外部环境发生变化;所述终端移动到第二小区;所述终端移动到新的跟踪区域;所述终端移动到新的地理位置;所述终端的移动速度变化量大于或等于第七门限;所述终端的RSRP测量值发生变化;距离上一次训练的时间达到第四预设时长;用于触发重新训练的定时器超时;发生了连续M次模型监督,M为大于1的整数;发生了L次模型监督,L为正整数;同步信号块的检测失败次数达到第八门限。
- 根据权利要求8至15任一项所述的方法,其中,所述方法还包括:所述第一设备基于第八信息确定所述第一AI模型训练完成;其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
- 根据权利要求1至16任一项所述的方法,其中,所述方法还包括:所述第一设备基于第三配置信息对所述第一AI模型进行监督;其中,所述第三配置信息包括以下至少一项:需要进行模型监督的AI模型标识;模型监督的周期;模型监督的时长;模型监督的检测窗口相关信息;模型监督的触发条件;模型监督的指标;其中,所述模型监督的指标包括以下至少一项:预测值和真实值之间的误差类信息或精度信息;通信系统性能;第一AI模型的模型相关信息。
- 根据权利要求17所述的方法,其中,所述模型监督的触发条件基于以下至少一项确定:模型监督的指标;所述第一AI模型的推理结果;模型推理的指标。
- 根据权利要求1至18任一项所述的方法,其中,所述方法还包括:所述第一设备与第二设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:终端是否支持所述第一AI模型训练的能力;终端是否支持基于所述第一AI模型进行AI推理的能力;终端是否支持上报辅助信息用于所述第一AI模型训练的能力;终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;网络侧设备是否支持所述第一AI模型训练的能力;网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;服务器是否支持所述第一AI模型训练的能力;服务器是否支持基于所述第一AI模型进行AI推理的能力。
- 根据权利要求19所述的方法,其中,所述方法还包括:所述第一设备基于以下至少一项确定所述第二设备对应的所述目标能力信息:所述第二设备的设备类型;第二设备的网络类型;第二设备发送的参考信号;所述第二设备发送的控制信息;所述第二设备发送的RRC信令;所述第一设备与所述第二设备之间的接口消息。
- 一种小区驻留处理方法,包括:第二设备执行以下至少一项:所述第二设备向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;第二操作;其中,所述第二操作包括以下至少一项:所述第二设备进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;所述第二设备进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
- 根据权利要求21所述的方法,其中,所述第二设备进行第一AI模型的训练,获得第一AI模型包括:所述第二设备基于第五信息进行第一AI模型的训练,获得第一AI模型;其中,所述第五信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求22所述的方法,其中,所述第二设备基于第五信息第一AI模型的训练,获得第一AI模型之前,所述方法还包括:在所述第二设备为所述服务器的情况下,所述第二设备从所述终端和网络侧设备中的至少一项获取所述第五信息;或者,在所述第二设备为所述网络侧设备的情况下,所述第二设备从所述终端获取所述第五信息的至少部分信息;或者,在所述第二设备为所述终端的情况下,所述第二设备从所述网络侧设备获取所述第五信息的至少部分信息。
- 根据权利要求23所述的方法,其中,所述方法还包括:所述第二设备从所述第一设备接收第四信息;所述第二设备基于所述第四信息更新所述第一AI模型;所述第二设备向所述第一设备发送更新后的第一AI模型。
- 根据权利要求21所述的方法,其中,所述第二设备进行第一AI模型的训练,获得第一AI模型包括:所述第二设备从所述第一设备接收第一训练结果;所述第二设备将所述第一训练结果输入到第四AI模型进行所述第一AI模型的训练,获得所述第一AI模型;所述第二设备向所述第一设备发送所述第一AI模型。
- 根据权利要求21所述的方法,其中,所述第二设备进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果包括:所述第二设备将第九信息输入第五AI模型,获得第二训练结果;所述第二设备向所述第一设备发送所述第二训练结果;其中,所述第九信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 根据权利要求26所述的方法,其中,所述第二设备将第九信息输入第五AI模型,获得第二训练结果之前,所述方法还包括:在所述第二设备为所述服务器的情况下,所述第二设备从所述终端和网络侧设备中的至少一项获取所述第九信息;或者,在所述第二设备为所述网络侧设备的情况下,所述第二设备从所述终端获取所述第九信息的至少部分信息;或者,在所述第二设备为所述终端的情况下,所述第二设备从所述网络侧设备获取所述第九信息的至少部分信息。
- 根据权利要求21所述的方法,其中,在所述第二设备为终端的情况下,所述方法还包括:所述第二设备从所述第一设备接收所述目标结果。
- 根据权利要求21至28任一项所述的方法,其中,所述方法还包括:所述第二设备基于第一配置信息进行周期性触发所述第一AI模型的训练;其中,所述第一配置信息包括以下至少一项:周期性模型训练的起点;周期性模型训练的间隔;一个周期内的模型训练次数;一个周期内的模型训练时长。
- 根据权利要求21至28任一项所述的方法,其中,所述方法还包括:所述第二设备基于半静态触发方式触发所述第一AI模型的训练;其中,所述半静态触发方式包括以下至少一项:基于第二配置信息进行半静态触发;通过物理控制信息激活或去激活模型的半静态训练;其中,所述第二配置信息满足以下至少一项:基于目标事件触发所述第二配置信息的发送和激活中的至少一项;通过无线资源控制RRC配置所述第二配置信息。
- 根据权利要求21至28任一项所述的方法,其中,所述第一设备基于第一人工智能AI模型获得目标结果之前,所述方法还包括:所述第一设备基于目标事件触发所述第一AI模型的训练;其中,所述目标事件包括以下至少一项:最近一次的小区定义的同步信号块CD-SSB的发送频点发生变化;最近一次的非小区定义的同步信号块NCD-SSB的发送频点发生变化;所述终端移动到小区边缘或预设位置;所述终端的移动速度高于或等于第一门限;所述终端的移动速度低于或等于第二门限;小区当前接入或驻留的终端的数量高于或等于第三门限;小区当前接入或驻留的终端的数量低于或等于第四门限;使用所述第一AI模型推理失败;使用所述第一AI模型连续推理失败的次数达到第五门限;使用所述第一AI模型推理失败的次数达到第六门限;使用所述第一AI模型进行了推理;所述终端重选到第二小区;所述终端的跟踪区域发生变化;所述终端的外部环境发生变化;所述终端移动到第二小区;所述终端移动到新的跟踪区域;所述终端移动到新的地理位置;所述终端的移动速度变化量大于或等于第七门限;所述终端的RSRP测量值发生变化;距离上一次训练的时间达到第四预设时长;用于触发重新训练的定时器超时;发生了连续M次模型监督,M为大于1的整数;发生了L次模型监督,L为正整数;同步信号块的检测失败次数达到第八门限。
- 根据权利要求21至31任一项所述的方法,其中,所述方法还包括:所述第一设备基于第八信息确定所述第一AI模型训练完成;其中,所述第八信息包括以下至少一项:同步信号的接收信号强度指示;检测到的主同步信号或辅同步信号的信号强度;同步信号的信道估计信噪比;同步信号块的参考信号接收功率RSRP;层1的RSRP;层3的RSRP;终端成功检测到同步信号的概率;终端检测到同步信号的时长;终端成功驻留的概率;终端成功同步的概率。
- 根据权利要求21至32任一项所述的方法,其中,所述方法还包括:所述第二设备与第一设备之间传输目标能力信息,所述目标能力信息包括以下至少一项:终端是否支持所述第一AI模型训练的能力;终端是否支持基于所述第一AI模型进行AI推理的能力;终端是否支持上报辅助信息用于所述第一AI模型训练的能力;终端是否支持上报辅助信息用于基于所述第一AI模型进行AI推理的能力;网络侧设备是否支持所述第一AI模型训练的能力;网络侧设备是否支持基于所述第一AI模型进行AI推理的能力;网络侧设备是否支持指示辅助信息用于所述第一AI模型训练的能力;网络侧设备是否支持指示辅助信息用于基于所述第一AI模型进行AI推理的能力;服务器是否支持所述第一AI模型训练的能力;服务器是否支持基于所述第一AI模型进行AI推理的能力。
- 根据权利要求33所述的方法,其中,所述方法还包括:所述第二设备基于以下至少一项确定所述第一设备对应的所述目标能力信息:所述第一设备的设备类型;第一设备的网络类型;第二设备发送的参考信号;所述第一设备发送的控制信息;所述第一设备发送的RRC信令;所述第一设备与所述第二设备之间的接口消息。
- 一种小区驻留处理装置,应用于第一设备,包括:第一处理模块,用于基于第一人工智能AI模型获得目标结果;其中,所述第一设备为终端、网络侧设备或服务器,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
- 根据权利要求35所述的装置,其中,所述第一处理模块具体用于将第一信息输入第一AI模型,获得所述目标结果;其中,所述第一信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 一种小区驻留处理装置,应用于第二设备,其中,包括:执行模块,用于执行以下至少一项:向第一设备发送第一信息中的至少部分信息,所述第一信息用于第一AI模型的推理;第二操作;其中,所述第二操作包括以下至少一项:进行第一AI模型的训练,获得第一AI模型,并向第一设备发送第一AI模型;进行第一AI模型的训练,获得第二训练结果,并向第一设备发送所述第二训练结果,所述第二训练结果用于所述第一设备进行第一AI模型的训练;其中,所述第一设备为终端、网络侧设备或服务器,所述第二设备为终端、网络侧设备或服务器,所述第一AI模型用于确定目标结果,所述目标结果用于进行小区驻留,且所述目标结果包括以下至少一项:同步信号发送的频域位置;同步信号发送的时域位置;同步信号发送的波束发送方向;第一小区的信息,所述第一小区为可进行驻留的小区。
- 根据权利要求37所述的装置,其中,所述执行模块具体用于基于第五信息进行第一AI模型的训练,获得第一AI模型;其中,所述第五信息包括以下至少一项:时间信息;终端的状态信息,所述状态信息包括以下至少一项:位置信息、移动方向、移动速度、能耗状况、电量状况、所处环境信息、感知信息、网络场景信息、运营商信息和网络类型;在第一预设时间段内同步信号检测失败的次数;突发的事件;同步信号的模式信息;同步信号的频域特征;同步信号的时域特征;同步信号的空域特征。
- 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,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34任一项所述的小区驻留处理方法的步骤。
- 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,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34任一项所述的小区驻留处理方法的步骤。
- 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权利要求1至34任一项所述的小区驻留处理方法的步骤。
- 一种计算机程序产品,包括计算机指令,所述计算机指令被处理器执行时实现如权利要求1至34中任一项所述的小区驻留处理方法的步骤。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410717888.4A CN121078505A (zh) | 2024-06-04 | 2024-06-04 | 小区驻留处理方法、装置、终端及网络侧设备 |
| CN202410717888.4 | 2024-06-04 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2025252002A1 true WO2025252002A1 (zh) | 2025-12-11 |
Family
ID=97835204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2025/098146 Pending WO2025252002A1 (zh) | 2024-06-04 | 2025-05-29 | 小区驻留处理方法、装置、终端及网络侧设备 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN121078505A (zh) |
| WO (1) | WO2025252002A1 (zh) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114071484A (zh) * | 2020-07-30 | 2022-02-18 | 华为技术有限公司 | 基于人工智能的通信方法和通信装置 |
| WO2022261834A1 (zh) * | 2021-06-15 | 2022-12-22 | Oppo广东移动通信有限公司 | 小区切换方法、装置、设备及存储介质 |
| CN116017616A (zh) * | 2021-10-21 | 2023-04-25 | 维沃移动通信有限公司 | 小区重选方法、装置及相关设备 |
| CN116419354A (zh) * | 2021-12-30 | 2023-07-11 | 华为技术有限公司 | 通信方法及装置 |
-
2024
- 2024-06-04 CN CN202410717888.4A patent/CN121078505A/zh active Pending
-
2025
- 2025-05-29 WO PCT/CN2025/098146 patent/WO2025252002A1/zh active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114071484A (zh) * | 2020-07-30 | 2022-02-18 | 华为技术有限公司 | 基于人工智能的通信方法和通信装置 |
| WO2022261834A1 (zh) * | 2021-06-15 | 2022-12-22 | Oppo广东移动通信有限公司 | 小区切换方法、装置、设备及存储介质 |
| CN116017616A (zh) * | 2021-10-21 | 2023-04-25 | 维沃移动通信有限公司 | 小区重选方法、装置及相关设备 |
| CN116419354A (zh) * | 2021-12-30 | 2023-07-11 | 华为技术有限公司 | 通信方法及装置 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN121078505A (zh) | 2025-12-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20240267761A1 (en) | Model request method, model request processing method, and related device | |
| CN116458194A (zh) | 无线节点之间的机器学习模型共享 | |
| CN117440449A (zh) | Ai模型信息传输方法、装置及设备 | |
| US12402022B2 (en) | Performance-driven network parameter changes in a communication network | |
| WO2024022316A1 (zh) | 下行参考信号发送方法、装置、终端及网络侧设备 | |
| CN115329954A (zh) | 训练数据集获取方法、无线传输方法、装置及通信设备 | |
| GB2625108A (en) | Wake-up signal based cell re-selection | |
| CN118785234A (zh) | 基于ai的csi预测的数据采集方法及装置 | |
| WO2025252002A1 (zh) | 小区驻留处理方法、装置、终端及网络侧设备 | |
| CN120128943A (zh) | 无线通信方法、装置及设备 | |
| WO2025252057A1 (zh) | 传输控制方法、装置、第一设备及第二设备 | |
| CN121078474A (zh) | 传输控制方法、装置及通信设备 | |
| WO2026012157A1 (zh) | 信息传输方法、装置及通信设备 | |
| US20260040360A1 (en) | Adaptive carrier selection for random access channel procedures using machine learning | |
| US20260129510A1 (en) | System and method for boosting power of synchronization signals for aerial cells | |
| WO2026067402A1 (zh) | 传输方法、装置、第一设备及第二设备 | |
| CN120166480A (zh) | 通信方法、装置、终端、网络侧设备及存储介质 | |
| CN121334777A (zh) | 信息获取方法、信息传输方法、装置及通信设备 | |
| WO2026021600A1 (zh) | 通信方法、ai模型的训练方法、装置、设备和存储介质 | |
| WO2026040901A1 (zh) | 随机接入传输方法、终端及网络侧设备 | |
| WO2025228057A1 (zh) | 通信方法和通信装置 | |
| WO2026026850A1 (zh) | 小区选择或重选方法、装置及相关设备 | |
| CN122002266A (zh) | 操作执行方法、关联建立方法、装置、终端及网络侧设备 | |
| CN120075906A (zh) | 通信方法、终端及网络侧设备 | |
| WO2025082325A1 (zh) | 行为执行方法、装置、终端、网络侧设备及介质 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 25819055 Country of ref document: EP Kind code of ref document: A1 |