Speed adaptive time domain beam prediction method and device
Through the speed adaptive time domain beam prediction method, the time domain beam prediction model and encoding-decoding LSTM network are used to solve the problems of large resource consumption and long scanning time in millimeter wave communication, and efficient and accurate beam prediction and communication quality improvement are achieved.
Patent Information
- Application Number
- CN202411996323.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-11
AI Technical Summary
In millimeter wave communication, traditional beam alignment methods require scanning a large number of beam directions, resulting in large resource consumption and long scanning time, and cannot effectively adapt to the fast mobility of mobile terminals.
A speed adaptive time domain beam prediction method is proposed. By obtaining the RSRP data reported by the mobile user to the base station, inputting the time domain beam prediction model, outputting the RSRP at the future moment, dynamically adjusting the input and output time series length to adapt to the user's movement speed, and beam prediction is performed using the encoding-decoding LSTM network.
It realizes efficient and accurate determination of the optimal beam direction, reduces resource consumption and beam alignment time, and improves communication quality and efficiency.
Smart Images
Figure CN119814103A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of wireless communication technology, and in particular to a speed adaptive time-domain beam prediction method and device. Background Art
[0002] 5G communication systems widely use multi-antenna transmission technology, which is particularly critical when working in high-frequency bands. However, higher frequencies often mean more electromagnetic wave attenuation. This attenuation occurs not only on the propagation path, but also on the terminal antenna (transmitting antenna or receiving antenna). In order to effectively deal with the attenuation problem during millimeter wave propagation, beamforming technology has emerged and has been widely used. Millimeter wave wavelengths are small, and the corresponding large wire sizes are also small. Based on this feature, multiple large wires can be coupled together, and a large number of antennas can be used to achieve beamforming. By aligning multiple antennas to transmit signals in the same direction, the power of the transmitted signal can be significantly increased, effectively combating signal fading, thereby improving the quality of communication.
[0003] However, in millimeter wave communications, how to find a suitable beam management method has become a key issue. Currently, a series of beam management processes have been defined in the 5G standard, including beam scanning, beam measurement, beam selection, and beam alignment. However, these processes face a severe challenge in implementation: due to the continuous movement of mobile terminals, the beam needs to be continuously aligned. In the traditional beam alignment process, the sender and receiver send information in each direction in turn and measure the RSRP in each direction to determine the optimal beam direction. However, this method requires scanning a large number of beam directions, which not only consumes a large amount of communication resources, but also due to the rapid mobility of users, beam scanning usually needs to be performed multiple times, which further increases the communication overhead. In addition, the scanning time must also be controlled within a relatively short time to prevent the beam scanning from failing.
[0004] Therefore, there is an urgent need for an innovative beam management technology that can reduce resource consumption while speeding up beam alignment time to improve communication quality. This technology needs to be able to efficiently and accurately determine the optimal beam direction to adapt to the rapid mobility of mobile terminals and provide strong technical support for the widespread application of millimeter wave communications. Summary of the invention
[0005] The present application aims to solve one of the technical problems in the related art at least to some extent.
[0006] To this end, the first purpose of this application is to propose a speed-adaptive time-domain beam prediction method that can efficiently and accurately determine the optimal beam direction to adapt to the rapid mobility of mobile terminals and provide strong technical support for the widespread application of millimeter wave communications.
[0007] The second objective of the present application is to propose a speed adaptive time domain beam prediction device.
[0008] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a velocity adaptive time-domain beam prediction method, including:
[0009] Obtain RSRP data reported by a mobile user terminal UE to a base station BS;
[0010] Input RSRP data into the time domain beam prediction model, and output RSRP at a future time, wherein the input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE, and the time domain beam prediction model includes a data preprocessing module, a prediction network, and a data post-processing module. The prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result, and the data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result;
[0011] The RSRP at a future time is sent to the UE by the BS, so that the UE can perform beam selection according to the received information.
[0012] Optionally, in one embodiment of the present application, determining the input and output time series length based on the moving speed of the UE includes:
[0013] According to the preset mapping between the UE moving speed and the input time window j and the output time window h, the input time window j and the output time window h corresponding to the UE moving speed are determined.
[0014] Optionally, in one embodiment of the present application, the data preprocessing module is specifically used to:
[0015] The RSRP data is standardized and dimensionally reduced, and the optimal RSRP is screened out. The standardized formula is:
[0016]
[0017] Where μ is RSRP s (t) is the average value, σ is the standard deviation, RSRP t is the RSRP value at time t.
[0018] Optionally, in one embodiment of the present application, obtaining RSRP data reported by a mobile user terminal UE to a base station BS includes:
[0019] The i-th group of RSRP measurements reported by the UE to the BS is denoted as X i , after j beam management cycles, the signal received at the BS is expressed as:
[0020] RBS =(X 1 ,X 2 ,...,X j )
[0021] The prediction process of the prediction network includes:
[0022] Set the objective function of the prediction network to:
[0023]
[0024] in, is the prediction function that maps j historical inputs to future estimates;
[0025] Through the objective function, the time domain beam prediction problem of the prediction network is expressed as an optimization task. The optimization task is to minimize the cumulative prediction error within the specified future time window. The time beam prediction problem is expressed as:
[0026]
[0027] Optionally, in one embodiment of the present application, the prediction module adopts an encoding-decoding LSTM network, which includes an encoder, a decoder and a fully connected FC layer, wherein the encoder processes the input window (X 1 ,X 2 ,...,X j ) is processed, and the temporal features of the sequence are gradually extracted and encoded through a preset number of LSTM layers. The decoder includes a preset number of LSTM layers, and the decoder uses the encoded features to predict the output sequence The fully connected FC layer transforms the LSTM output into the final prediction value.
[0028] To achieve the above-mentioned object, a second aspect of the present invention proposes a speed adaptive time domain beam prediction device, including a base station BS and a mobile user end UE, wherein:
[0029] UE reports RSRP data to BS;
[0030] The BS inputs the RSRP data into the time domain beam prediction model and outputs the RSRP at the future time. The input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE. The time domain beam prediction model includes a data preprocessing module, a prediction network and a data post-processing module. The prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result. The data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result.
[0031] The BS sends the RSRP at the future time to the UE so that the UE can perform beam selection based on the received information.
[0032] Optionally, in one embodiment of the present application, the BS and the UE are in a millimeter wave communication system under the 5G standard, the BS adopts analog beamforming, and the BS and the UE use a uniform planar array UPA.
[0033] Optionally, in one embodiment of the present application, it is set that in each beam measurement scenario, the UE identifies all beam pairs without additional Rx beam scanning attempts, and the RSRP reported by the UE is the instantaneous RSRP calculated assuming the use of the best Rx beam.
[0034] Optionally, in one embodiment of the present application, the UE performs a beam measurement within a time window and reports a group of RSRP.
[0035] Optionally, in one embodiment of the present application, in the communication system, the number of antennas of the BS and the UE are M and N respectively, and the downlink channel matrix is The beamforming weight vector is expressed as:
[0036]
[0037] in, and are the weights of the antenna units in the vertical and horizontal directions of UPA, M beams is the size of the beamforming codebook;
[0038] The beam pair (m,n) signal received by the UE is expressed as:
[0039]
[0040] in,, is the average transmit power, and are the weights of the antenna units in the vertical and horizontal directions of UPA, is the beamforming vector of BS and UE, s is the transmitted pilot signal, n~(0,σ 2 ) is additive noise, and the calculation formula of the RSRP of the beam pair (m,n) is:
[0041]
[0042] Among them, P is the power used by the transmitter for signal transmission, and F is the Frobenius norm.
[0043] The speed-adaptive time-domain beam prediction method and device of the embodiment of the present application dynamically adjusts the input-output time window size according to the moving speed of the user terminal (UE), constructs a mapping relationship between speed and time window through experiments, realizes speed adaptation, and improves prediction efficiency; reduces data complexity and model calculation overhead by performing dimensionality reduction and standardization on data, while ensuring prediction accuracy; learns time-domain prediction through the encoding-decoding LSTM network, flexibly processes input and output sequences of different lengths, and adapts to changing user mobility scenarios.
[0044] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0046] Figure 1 A schematic diagram of a flow chart of a velocity adaptive time-domain beam prediction method provided in Embodiment 1 of the present application;
[0047] Figure 2 This is a diagram of the 5G millimeter wave communication process of an embodiment of the present application;
[0048] Figure 3 A structural diagram of the encoding-decoding LSTM network model designed for an embodiment of the present application;
[0049] Figure 4 A schematic diagram of the Top-k prediction accuracy of an embodiment of the present application;
[0050] Figure 5 A relationship diagram of the input time window length j and the prediction time window length h of an embodiment of the present application;
[0051] Figure 6 A schematic diagram of speed adaptation in an embodiment of the present application;
[0052] Figure 7 This is a structural block diagram of the Tx dimension reduction processing of an embodiment of the present application;
[0053] Figure 8 This is a block diagram of a time domain beam prediction model of an embodiment of the present application;
[0054] Fig. 9 A schematic diagram of the structure of a speed adaptive time-domain beam prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0055] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0056] The speed adaptive time-domain beam prediction method and device of the embodiments of the present application are described below with reference to the accompanying drawings.
[0057] Figure 1 A schematic flow chart of a velocity adaptive time-domain beam prediction method provided in Embodiment 1 of the present application.
[0058] like Figure 1 As shown, the velocity adaptive time-domain beam prediction method comprises the following steps:
[0059] Step 101, obtaining RSRP data reported by a mobile user terminal UE to a base station BS;
[0060] This embodiment is applied to beam prediction in 5G millimeter wave communication systems, especially when the base station and the terminal have a multi-antenna array configuration. Protection is not limited to 5G millimeter wave communication, but also includes beam management in other wireless communication technologies (such as Wi-Fi, satellite communication, etc.). Applicability under different frequency bands or communication standards (such as 6G or future standards).
[0061] In this embodiment, the base station adopts analog beamforming, and the BS and UE use a uniform planar array (UPA).
[0062] In this embodiment, it is assumed that in each beam measurement scenario, the UE does not need to strictly follow the P1 / P2 / P3 procedures. Specifically, it is assumed that in each beam measurement scenario, the UE can perfectly identify all beam pairs without additional Rx beam scanning attempts. The RSRP reported by the UE is only the instantaneous RSRP calculated assuming that the best Rx beam is used. The P1 / P2 / P3 procedure represents the pairing process of the transmit beam (Tx) and the receive beam (Rx). It is divided into three stages, P1 represents the coarse pairing of the downlink Tx beam and the Rx beam; P2 is the adjustment of the downlink Tx beam on the network side; P3 represents the fine-tuning of the downlink Rx beam on the terminal side.
[0063] In this embodiment, in the above communication system, the number of antennas of the BS and the UE are M and N respectively, and the downlink channel matrix is recorded as The beamforming weight vector can be expressed as:
[0064]
[0065] in, and are the weights of the antenna units in the vertical and horizontal directions of the UPA, respectively. beams is the size of the beamforming codebook.
[0066] The beam pair (m,n) signal received by the UE can be expressed as:
[0067]
[0068] in,, is the average transmit power, and are the weights of the antenna units in the vertical and horizontal directions of UPA, is the beamforming vector of BS and UE, s is the transmitted pilot signal, n~(0,σ 2 ) is additive noise, and the calculation formula of the RSRP of the beam pair (m,n) is:
[0069]
[0070] Among them, P is the power used by the transmitter for signal transmission, and F is the Frobenius norm.
[0071] In this embodiment, the time for the UE to measure and report a group of RSRPs in one beam is defined as a time window. In the 5G standard protocol, the time for one beam management is 20ms.
[0072] Figure 2 The 5G millimeter wave communication process diagram applicable to this embodiment is shown in the figure. The number of antennas of the BS and UE are M and N respectively. The BS can send multiple analog beams Tx, and each beam transmits a shaped reference signal (such as CSI-RS) for beam measurement. After the mobile phone scans by receiving the beam Rx, it calculates the RSRP based on the received signal strength and then reports it to the base station. The base station receives the RSRP calculated by the UE at multiple times to obtain the beam set.
[0073] Step 102: input RSRP data into a time domain beam prediction model, and output RSRP at a future time. The input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE. The time domain beam prediction model includes a data preprocessing module, a prediction network, and a data post-processing module. The prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result. The data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result.
[0074] In this embodiment, when making a prediction, a time domain prediction target optimization function is first constructed, including:
[0075] The set of RSRP measurements reported by the UE is denoted as X. After j beam management periods, the signal received at the BS can be expressed as:
[0076] P BS =(X 1 ,X 2 ,...,X j )
[0077] Based on this, the objective function of constructing the Top-k beams that predict the future h beam measurement moments is:
[0078]
[0079] in represents the prediction function that maps j historical input moments to future estimates.
[0080] The above objective function expresses the time-domain beam prediction problem as an optimization task, seeking to minimize the cumulative prediction error within a specified future time window. Therefore, the time-domain beam prediction problem can be expressed as:
[0081]
[0082] In this embodiment, the time domain beam prediction LSTM network is used as the prediction network. Consider the prediction function It is a prediction function that maps historical inputs to future estimates. Artificial intelligence has a strong ability to process sequence data. The LSTM network uses storage units and gates to save important historical information in sequential data while filtering out irrelevant noise. This enables it to effectively manage long-term dependencies. In addition, LSTM performs well under dynamic and variable-length input and output, and can adapt to changes in different time scales. Therefore, LSTM is used for time-domain beam prediction. The LSTM unit has 4 gate structures, namely input gate i, forget gate f, memory gate g, and output gate o. The calculation formula is as follows:
[0083] i t =σ(W xi x t +W hi h t-1 +b i )
[0084] f t =σ(W xf x t +W hf h t-1 +b f )
[0085] g t =tanh(W xg xt +W hg h t-1 +b g )
[0086] c t =f t ×c t-1 +i t ×g t
[0087] o t =σ(W xo x t +W ho h t-1 +b o )
[0088] h t =o t ×tanh(c t )
[0089] Among them, W i (W xi ,W hi ),W f (W xf ,W hf ),W g (W xg ,W hg ),W o (W xo ,W ho ) are the weight parameter matrices corresponding to each gate. × represents element-by-element multiplication, also known as Hadamard product or point-by-point multiplication. t is the input data of the current time step t, h t-1 Represents the hidden state of the previous time step (t-1). The sigmoid function and tanh function are defined as follows:
[0090]
[0091] The calculation process of LSTM includes four core steps. First, f t Based on the current input and the output of the previous time step, decide which previous information to discard from the cell state. t and g t jointly decide what new information to add to the cell state c t In which i t The sigmoid function is used to control the update ratio, and the tanh function is used to generate candidate values as new candidate vectors. Then, c is updated by combining the forgotten and added information. t , retaining important information and discarding unnecessary content. Finally, ot Determines what information in the cell state will be output, which is scaled by a tanh function before being output as the value of the current time step.
[0092] In this embodiment, in order to better manage the different lengths of the input and output windows, an encoder-decoder LSTM structure is used. The length of the input and output time series is related to the different UE moving speeds and is controlled by the adaptive module. In addition, the network belongs to the category of supervised learning, so labels must be constructed during training and prediction results must be calculated. Specifically, the encoder performs a certain amount of training on the input window (denoted as (X 1 ,X 2 ,...,X j )) is processed, and the temporal features of the sequence are gradually extracted and encoded through four LSTM layers. The decoder also contains four LSTM layers, using the encoded features to predict the output sequence (expressed as ). This allows the model to handle input and output sequences of varying lengths by encoding historical data and decoding it into future predictions through these LSTM layers. Fully connected (FC) layers transform the LSTM outputs into final predicted values.
[0093] Figure 3 Structural diagram of the designed encoding-decoding LSTM network model.
[0094] Evaluation Metrics: Top-k Prediction Accuracy: The accuracy of the actual maximum measured RSRP being one of the predicted Top-k RSRPs.
[0095] Figure 4 This is the prediction accuracy of the time domain beam prediction method proposed in this embodiment, wherein the Top-1 prediction accuracy can reach more than 90%.
[0096] In this embodiment, the length of the input and output time series is controlled by an adaptive module, including:
[0097] Determine the historical observation window j and the predicted future window h. Specifically, associate the speed with the window. Study the UE moving speed v, the historical observation window j, and the predicted future window as variables. Input RSRP data under different UE moving speeds into the designed encoder-decoder LSTM network. Experiment with the historical observation window size from 1 to 6 and the predicted future window size from 2 to 8 to observe and record the best time window.
[0098] Figure 5The relationship diagram between the input time window length j and the prediction time window length h. Specifically, the BS classifies the input data and labels of the set of j beam measurement moments. The input data is sent to the model, and the label is used to calculate the error and prediction accuracy with the prediction beam.
[0099] Figure 6 The speed adaptation module is presented, which stores a set of mapping values of the relationship between different UE moving speeds and the input time window j and the prediction time window h.
[0100] In this embodiment, data preprocessing includes standardization and dimensionality reduction processing. The standardization formula is:
[0101]
[0102] Dimensionality reduction processing such as Figure 7 As shown in the figure, the input Tx dimension is reduced by 1 / 4 to select the optimal RSRP.
[0103] Figure 8 This is the block diagram of the time domain beam prediction model. The designed speed adaptive time domain beam prediction model is mainly divided into four parts, namely, speed adaptive module, data preprocessing module, AI prediction module, and data post-processing module.
[0104] Specifically, when training the model, the data set is divided into training set, validation set and test set according to a certain ratio. Experiments are conducted with different input-output sequence lengths, for example, the input sequence is 4 and the output sequence is 6 as labels. During the training phase, in order to speed up the training of the model, the training set is further divided and smaller subsequences are used as the input of the model.
[0105] Specifically, in the inference phase, the generated data are sequentially fed into the prediction model without further segmentation.
[0106] Step 103: The RSRP at the future time is sent to the UE through the BS, so that the UE performs beam selection according to the received information.
[0107] In this embodiment, the output predicted RSRP and beam index ID for multiple future moments will be further processed by the BS and sent to the UE with other signals. The UE can pre-select a more appropriate beam based on the information, reduce beam scanning and reporting at future moments, greatly improve communication efficiency, and reduce resource consumption.
[0108] In order to implement the above embodiment, the present application also proposes a speed adaptive time domain beam prediction device.
[0109] Fig. 9 A schematic diagram of the structure of a speed adaptive time-domain beam prediction device provided in an embodiment of the present application.
[0110] like Fig. 9 As shown, the speed adaptive time domain beam prediction device includes a base station BS and a mobile user end UE, wherein the UE reports RSRP data to the BS;
[0111] The BS inputs the RSRP data into the time domain beam prediction model and outputs the RSRP at the future time. The input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE. The time domain beam prediction model includes a data preprocessing module, a prediction network and a data post-processing module. The prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result. The data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result.
[0112] The BS sends the RSRP at the future time to the UE so that the UE can perform beam selection based on the received information.
[0113] Optionally, in one embodiment of the present application, the BS and the UE are in a millimeter wave communication system under the 5G standard, the BS adopts analog beamforming, and the BS and the UE use a uniform planar array UPA.
[0114] Optionally, in one embodiment of the present application, it is set that in each beam measurement scenario, the UE identifies all beam pairs without additional Rx beam scanning attempts, and the RSRP reported by the UE is the instantaneous RSRP calculated assuming the use of the best Rx beam.
[0115] Optionally, in one embodiment of the present application, the UE performs a beam measurement within a time window and reports a group of RSRP.
[0116] Optionally, in one embodiment of the present application, in the communication system, the number of antennas of the BS and the UE are M and N respectively, and the downlink channel matrix is The beamforming weight vector is expressed as:
[0117]
[0118] in, and are the weights of the antenna units in the vertical and horizontal directions of UPA, M beams is the size of the beamforming codebook;
[0119] The beam pair (m,n) signal received by the UE is expressed as:
[0120]
[0121] in,, is the average transmit power, and are the weights of the antenna units in the vertical and horizontal directions of UPA, is the beamforming vector of BS and UE, s is the transmitted pilot signal, n~(0,σ 2 ) is additive noise, and the calculation formula of the RSRP of the beam pair (m,n) is:
[0122]
[0123] Among them, P is the power used by the transmitter for signal transmission, and F is the Frobenius norm.
[0124] It should be noted that the above explanation of the speed adaptive time-domain beam prediction method embodiment is also applicable to the speed adaptive time-domain beam prediction device of this embodiment, and will not be repeated here.
[0125] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0126] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0127] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0128] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0129] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0130] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0131] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0132] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A velocity adaptive time domain beam prediction method, characterized in that: include: Obtain RSRP data reported by a mobile user terminal UE to a base station BS; Input the RSRP data into a time domain beam prediction model, and output the RSRP at a future time, wherein the input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE, the time domain beam prediction model comprises a data preprocessing module, a prediction network, and a data post-processing module, the prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result, and the data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result; The RSRP at the future time is sent to the UE through the BS, so that the UE performs beam selection according to the received information.
2. The method according to claim 1, characterized in that The length of the input and output time series is determined based on the moving speed of the UE, including: According to the preset mapping between the UE moving speed and the input time window j and the output time window h, the input time window j and the output time window h corresponding to the UE moving speed are determined.
3. The method according to claim 1, characterized in that The data preprocessing module is specifically used for: The RSRP data is standardized and dimensionally reduced, and the optimal RSRP is screened out, wherein the standardized formula is: Where μ is RSRP s (t) is the average value, σ is the standard deviation, RSRP t is the RSRP value at time t.
4. The method according to claim 1, characterized in that The obtaining of RSRP data reported by the mobile user terminal UE to the base station BS includes: The i-th group of RSRP measurements reported by the UE to the BS is denoted as X i , after j beam management cycles, the signal received at the BS is expressed as: R BS =(X 1 ,X 2 ,...,X j ) The prediction process of the prediction network includes: The objective function of the prediction network is set as: in, is the prediction function that maps j historical inputs to future estimates; Through the objective function, the time domain beam prediction problem of the prediction network is expressed as an optimization task, and the optimization task is to seek to minimize the cumulative prediction error in a specified future time window. The time beam prediction problem is expressed as:
5. The method according to claim 4, characterized in that The prediction module adopts an encoding-decoding LSTM network, which includes an encoder, a decoder and a fully connected FC layer, wherein the encoder performs an operation on the input window (X 1 ,X 2 ,...,X j ) is processed, and the temporal features of the sequence are gradually extracted and encoded through a preset number of LSTM layers. The decoder includes a preset number of LSTM layers, and the decoder uses the encoded features to predict the output sequence The fully connected FC layer transforms the LSTM output into the final prediction value.
6. A speed adaptive time domain beam prediction device, characterized in that: It includes a base station BS and a mobile user terminal UE, wherein: UE reports RSRP data to BS; The BS inputs the RSRP data into a time domain beam prediction model, and outputs the RSRP at a future time, wherein the input and output time series lengths of the time domain beam prediction network are determined by the moving speed of the UE, and the time domain beam prediction model includes a data preprocessing module, a prediction network, and a data post-processing module, wherein the prediction network performs beam prediction based on the data preprocessed by the data preprocessing module to obtain a beam prediction result, and the data post-processing module performs data area normalization on the beam prediction result to obtain a final prediction result; The BS sends the RSRP at the future time to the UE, so that the UE performs beam selection according to the received information.
7. The device according to claim 6, characterized in that The BS and UE are in a millimeter wave communication system under the 5G standard. The BS adopts analog beamforming, and the BS and UE use uniform planar array UPA.
8. The device according to claim 6, characterized in that It is assumed that in each beam measurement scenario, the UE identifies all beam pairs without additional Rx beam scanning attempts, and the RSRP reported by the UE is the instantaneous RSRP calculated assuming the best Rx beam is used.
9. The device according to claim 6, characterized in that The UE performs a beam measurement within a time window and reports a set of RSRP.
10. The device according to claim 7, characterized in that In the communication system, the number of antennas of the BS and the UE are M and N respectively, and the downlink channel matrix is The beamforming weight vector is expressed as: in, and are the weights of the antenna units in the vertical and horizontal directions of UPA, M beams is the size of the beamforming codebook; The beam pair (m,n) signal received by the UE is expressed as: in, is the average transmit power, and are the weights of the antenna units in the vertical and horizontal directions of UPA, is the beamforming vector of BS and UE, s is the transmitted pilot signal, n~(0,σ 2 ) is additive noise, and the calculation formula of the RSRP of the beam pair (m,n) is: Among them, P is the power used by the transmitter for signal transmission, and F is the Frobenius norm.