A model training method, a beam tracking method, an apparatus, a device, and a medium

Abstract

The application discloses a model training method, a beam tracking method, a device, equipment and a medium. The model training method comprises the following steps: generating a millimeter wave beam tracking sample according to a historical millimeter wave beam index sequence; when a current beam tracking model needs to be trained at a sending end or a receiving end, performing beam tracking training on the current beam tracking model according to the millimeter wave beam tracking sample to obtain a target beam tracking model; and the target beam tracking model is used for tracking a millimeter wave beam. The technical scheme of the embodiment of the application reduces the time cost of millimeter wave beam search, keeps stable tracking of the millimeter wave beam sent by a terminal in fast motion, and can effectively adapt to a high dynamic communication environment.

Description

Technical Field

[0001] This invention relates to the field of millimeter-wave communication technology, and in particular to a model training method, beam tracking method, device, equipment, and medium. Background Technology

[0002] Millimeter-wave communication systems possess abundant idle spectrum resources, which can significantly improve data transmission rates, making them a key technology in 5G (5th Generation Mobile Communication Technology) / 6G (6th Generation Mobile Communication Technology) systems. However, millimeter-wave signals experience significant transmission attenuation, requiring beamforming techniques at both the transmitting and receiving ends to enhance signal reception power. Therefore, real-time beam tracking is necessary to ensure the communication link remains stable even as the relative positions of the transmitting and receiving ends change. Currently, commonly used millimeter-wave beam tracking methods include hierarchical search and Kalman filtering, but these methods have high search overhead and algorithm complexity. Summary of the Invention

[0003] This invention provides a model training method, beam tracking method, device, equipment, and medium to solve the problems of high search overhead and high algorithm complexity in existing millimeter-wave beam tracking.

[0004] According to one aspect of the present invention, a model training method is provided, comprising:

[0005] Generate millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences;

[0006] When the current beam tracking model needs to be trained at the transmitting or receiving end, the current beam tracking model is trained based on millimeter-wave beam tracking samples to obtain the target beam tracking model.

[0007] A target beam tracking model for tracking millimeter-wave beams.

[0008] According to another aspect of the present invention, a beam tracking method is provided, comprising:

[0009] Obtain the current millimeter-wave beam differential vector;

[0010] The target beam tracking model is used to analyze the current millimeter-wave beam differential vector to determine the target beam search range that matches the current millimeter-wave beam differential vector.

[0011] The transmitter sends a signal according to the target beam search range and obtains the measured value of the received beam fed back by the receiver.

[0012] The target tracking beam is determined based on the measured values ​​of the received beam fed back by the receiver.

[0013] The target beam tracking model is the model trained by the model training method in any embodiment of the present invention.

[0014] According to another aspect of the present invention, a model training apparatus is provided, comprising:

[0015] The sample generation module is used to generate millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences.

[0016] The target beam tracking model acquisition module is used to train the current beam tracking model based on millimeter-wave beam tracking samples to obtain the target beam tracking model when the current beam tracking model needs to be trained at the transmitting or receiving end.

[0017] A target beam tracking model for tracking millimeter-wave beams.

[0018] According to another aspect of the present invention, a beam tracking device is provided, comprising:

[0019] The differential vector acquisition module is used to acquire the current millimeter-wave beam differential vector;

[0020] The target beam search range determination module is used to analyze the current millimeter-wave beam differential vector through the target beam tracking model and determine the target beam search range that matches the current millimeter-wave beam differential vector.

[0021] The signal transmission and measurement value acquisition module is used to transmit signals through the transmitter according to the target beam search range and acquire the measurement values ​​of the received beam fed back by the receiver.

[0022] The target tracking beam determination module is used to determine the target tracking beam based on the measurement value of the received beam fed back by the receiver.

[0023] The target beam tracking model is a model trained using any of the model training methods of this invention.

[0024] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0025] At least one processor; and

[0026] A memory communicatively connected to the at least one processor; wherein,

[0027] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the model training method according to any embodiment of the present invention, or to perform the waveform tracking method according to any embodiment of the present invention.

[0028] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute either the model training method described in any embodiment of the present invention or the beam tracking method described in any embodiment of the present invention.

[0029] The technical solution of this invention generates millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences. Then, when the transmitting or receiving end needs to train the current beam tracking model, it trains the current beam tracking model based on these samples to obtain the target beam tracking model. In this solution, the millimeter-wave beam index can be used to quickly determine the possible positions of the millimeter-wave beam to be searched. Therefore, when the movement speed of the transmitting or receiving end changes from slow to fast, the millimeter-wave beam index can be used to effectively determine the possible positions of the tracked millimeter-wave beam. Furthermore, when the high-speed movement of the transmitting or receiving end causes beam jumps or even large jumps between adjacent tracking cycles, the difference in the millimeter-wave beam index can be used to determine the possible jump position of the current beam without having to start searching from the neighborhood of the beam position at the previous moment. This solves the problems of high millimeter-wave beam search overhead and high algorithm complexity in existing technologies, reduces the time overhead of millimeter-wave beam search, maintains stable tracking of millimeter-wave beams transmitted by fast-moving terminals, and can effectively adapt to highly dynamic communication environments.

[0030] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of a model training method provided in Embodiment 1 of the present invention;

[0033] Figure 2This is a flowchart of a model training method provided in Embodiment 2 of the present invention;

[0034] Figure 3 This is a flowchart of a beam tracking method provided in Embodiment 3 of the present invention;

[0035] Figure 4 This is a comparison chart of the measured value of a beam and the estimated value of an unmeasured search beam provided in Embodiment 4 of the present invention;

[0036] Figure 5 This is a simulation diagram of the current beam tracking model tracking the transmitted beam of a base station, provided in Embodiment 4 of the present invention.

[0037] Figure 6 This is a convergence diagram of an algorithm for tracking the transmitted beam of a base station using a current beam tracking model, as provided in Embodiment 4 of the present invention.

[0038] Figure 7 This is a schematic diagram of the structure of a model training device provided in Embodiment 5 of the present invention;

[0039] Figure 8 This is a schematic diagram of the structure of a beam tracking device provided in Embodiment Six of the present invention;

[0040] Figure 9 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0042] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0043] Example 1

[0044] Figure 1 This is a flowchart of a model training method provided in Embodiment 1 of the present invention. This embodiment is applicable to the precise beam tracking of millimeter-wave beams. The method can be executed by a model training device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes:

[0045] S110. Generate millimeter-wave beam tracking samples based on the historical millimeter-wave beam index sequence.

[0046] The historical millimeter-wave beam index sequence can be an index sequence composed of the indices of millimeter-wave beams at multiple selected historical moments. Millimeter-wave beam tracking samples can be used to train machine learning models, enabling the trained models to perform beam tracking on millimeter-wave signals. Millimeter-wave beam tracking samples can be sample data generated based on actual beam tracking requirements.

[0047] In this embodiment of the invention, millimeter-wave beams can be transmitted at multiple historical moments in a millimeter-wave communication system to determine a historical millimeter-wave beam index sequence that matches the selected millimeter-wave beam. Then, differential operations are performed on the historical millimeter-wave beam index sequence, and millimeter-wave beam tracking samples are generated based on the differential results.

[0048] S120. When the current beam tracking model needs to be trained at the transmitting or receiving end, the current beam tracking model is trained based on millimeter-wave beam tracking samples to obtain the target beam tracking model.

[0049] The current beam tracking model can be a machine learning model that uses millimeter-wave beam tracking samples as input to update its parameters. The target beam tracking model can be a machine learning model that updates the parameters of the current beam tracking model based on the millimeter-wave beam tracking samples. Optionally, the target beam tracking model can include, but is not limited to, a neural network model. The transmitting end can be a device that transmits the beam. The receiving end can be a device that receives the beam.

[0050] In an embodiment of the present invention, when there is a beam tracking requirement at the sending end or the receiving end, the current beam tracking model needs to be trained to achieve real-time update of the parameters in the model. If the sending end or the receiving end needs to train the current beam tracking model, the current beam tracking model with beam tracking capabilities is obtained, and then the millimeter wave beam tracking samples are input into the current beam tracking model. The current beam tracking model is trained by the millimeter wave beam tracking samples to obtain the target beam tracking model, and then the millimeter wave beam is tracked based on the target beam tracking model, that is, the millimeter wave beam sent by the sending end and / or received by the receiving end is tracked by the target beam tracking model.

[0051] Exemplarily, the current beam tracking model can be trained according to the historical sample data stored and having the same data type as the millimeter wave beam tracking samples. Assume t0 < t1 < t2 < t3 < t4. If the millimeter wave beam tracking samples are generated according to the beam association information in the channels at t2, t3, and t4 moments, the historical sample data required to obtain the current beam tracking model can be the beam association information in the channels at t0, t1, and t2 moments. The embodiments of the present invention do not limit the number of moments and the moment relationship corresponding to the historical sample data and the millimeter wave beam tracking samples.

[0052] When the movement speed of the sending end or the receiving end changes from slow to fast, the millimeter wave beam index can be used to obtain information about the relevant acceleration, so as to effectively judge the possible position of tracking the millimeter wave beam. When the sending end or the receiving end moves at a high speed, resulting in beam jumps or even large jumps within two adjacent tracking periods, the difference of the millimeter wave beam index can be used to effectively identify the speed information of the user.

[0053] The technical solution of the embodiment of the present invention generates millimeter wave beam tracking samples according to the historical millimeter wave beam index sequence, and then when the sending end or the receiving end needs to train the current beam tracking model, the millimeter wave beam tracking samples are input into the current beam tracking model for beam tracking training to obtain the target beam tracking model. In this solution, the millimeter wave beam index can be used to quickly give the possible position of the millimeter wave beam to be searched. Therefore, when the movement speed of the sending end or the receiving end changes from slow to fast, the millimeter wave beam index can be used to effectively judge the possible position of tracking the millimeter wave beam. When the sending end or the receiving end moves at a high speed, resulting in beam jumps or even large jumps within two adjacent tracking periods, the difference of the millimeter wave beam index can be used to determine the possible jump position of the current beam, without starting to search from the neighborhood of the beam position at the previous moment, solving the problems of large millimeter wave beam search overhead and high algorithm complexity in the prior art, reducing the time overhead of millimeter wave beam search, maintaining stable tracking of the millimeter wave beam sent by the fast-moving terminal, and effectively adapting to the high-dynamic communication environment.

[0054] Example 2

[0055] Figure 2 This is a flowchart of a model training method provided in Embodiment 2 of the present invention. This embodiment is based on the above embodiment and provides specific optional implementation methods before generating millimeter-wave beam tracking samples according to the historical millimeter-wave beam index sequence. Figure 2 As shown, the method includes:

[0056] S210. Through the transmitting end or the receiving end, perform differential operations on the historical horizontal beam index sequence and the historical vertical beam index sequence in the historical millimeter wave beam index sequence to obtain the first historical horizontal beam differential vector and the first historical vertical beam differential vector.

[0057] The historical beam horizontal index sequence can be a horizontal index sequence of beams that matches the historical millimeter-wave beam index sequence. The historical beam vertical index sequence can be a vertical index sequence of beams that matches the historical millimeter-wave beam index sequence. The first historical horizontal beam differential vector can be the result of differential calculation of the historical beam horizontal index sequence. The first historical vertical beam differential vector can be the result of differential calculation of the historical beam vertical index sequence.

[0058] In this embodiment of the invention, a historical horizontal beam index sequence and a historical vertical beam index sequence can be determined based on the historical millimeter-wave beam index sequence. Then, a differential operation is performed on the historical horizontal beam index sequence by the transmitting end or the receiving end to obtain a first historical horizontal beam differential vector, and a differential operation is performed on the historical vertical beam index sequence to obtain a first historical vertical beam differential vector.

[0059] S220. Determine the horizontal flip flag value based on the first non-zero element of the first historical horizontal beam differential vector, and determine the vertical flip flag value based on the first non-zero element of the first historical vertical beam differential vector.

[0060] The horizontal flip flag value can be determined based on the first non-zero element of the first historical horizontal beam differential vector, and is used to determine whether the first historical horizontal beam differential vector has been flipped. The vertical flip flag can be determined based on the first non-zero element of the first historical vertical beam differential vector, and is used to determine whether the first historical vertical beam differential vector has been flipped.

[0061] In this embodiment of the invention, the first non-zero element of the first historical horizontal beam differential vector can be obtained, and then the horizontal flip flag value can be determined according to the sign of the first non-zero element of the first historical horizontal beam differential vector, and the vertical flip flag value can be determined according to the sign of the first non-zero element of the first historical vertical beam differential vector.

[0062] Optionally, when the first non-zero element of the first historical horizontal beam differential vector is positive, the horizontal flip flag value can be set to 0; when the first non-zero element of the first historical horizontal beam differential vector is negative, the horizontal flip flag value can be set to 1. When the first non-zero element of the first historical vertical beam differential vector is positive, the vertical flip flag value can be set to 0; when the first non-zero element of the first historical vertical beam differential vector is negative, the vertical flip flag value can be set to 1.

[0063] S230. Update the first historical horizontal beam differential vector according to the horizontal flip flag value to obtain the second historical horizontal beam differential vector, and update the first historical vertical beam differential vector according to the vertical flip flag value to obtain the second historical vertical beam differential vector.

[0064] Specifically, the second historical horizontal beam differential vector can be obtained by updating the first historical horizontal beam differential vector based on the horizontal flip flag value. Similarly, the second historical vertical beam differential vector can be obtained by updating the first historical vertical beam differential vector based on the vertical flip flag value.

[0065] In this embodiment of the invention, the horizontal flip flag value can be used to determine whether the first historical horizontal beam differential vector needs to be flipped. If the first historical horizontal beam differential vector needs to be flipped, the elements in the first historical horizontal beam differential vector are inverted to obtain the second historical horizontal beam differential vector. If the first historical horizontal beam differential vector does not need to be flipped, it is used as the second historical horizontal beam differential vector. It is understood that the vertical flip flag value can also be used to determine whether the first historical vertical beam differential vector needs to be flipped. If the first historical vertical beam differential vector needs to be flipped, the elements in the first historical vertical beam differential vector are inverted to obtain the second historical vertical beam differential vector. If the first historical vertical beam differential vector does not need to be flipped, it is used as the second historical vertical beam differential vector.

[0066] For example, when the horizontal flip flag value is 0, the first historical horizontal beam differential vector can be used directly as the second historical horizontal beam differential vector without flipping it. When the horizontal flip flag value is 1, the first historical horizontal beam differential vector can be flipped, and its elements can be inverted to obtain the second historical horizontal beam differential vector. Similarly, when the vertical flip flag value is 0, the first historical vertical beam differential vector can be used directly as the second historical vertical beam differential vector without flipping it. When the vertical flip flag value is 1, the first historical vertical beam differential vector can be flipped, and its elements can be inverted to obtain the second historical vertical beam differential vector.

[0067] S240. Determine the historical millimeter-wave beam differential vector based on the second historical horizontal beam differential vector and the second historical vertical beam differential vector.

[0068] In this embodiment of the invention, the second historical horizontal beam differential vector and the second historical vertical beam differential vector can be used as historical millimeter-wave beam differential vectors.

[0069] S250. Generate millimeter-wave beam tracking samples based on historical millimeter-wave beam differential vectors and the current beam tracking model.

[0070] In this embodiment of the invention, historical millimeter-wave beam differential vectors can be input into the current beam tracking model, and then millimeter-wave beam tracking samples can be generated based on the output of the current beam tracking model and the historical millimeter-wave beam differential vectors.

[0071] In an optional embodiment of the present invention, generating millimeter-wave beam tracking samples based on historical millimeter-wave beam differential vectors and the current beam tracking model may include: inputting historical millimeter-wave beam differential vectors into the current beam tracking model to obtain the search probability of each search beam; determining the beam search range based on the search probability of each search beam and a search threshold; if there are unmeasured search beams outside the beam search range, obtaining unmeasured associated beams; obtaining the measurement value of the search beams and estimating the unmeasured search beams using the unmeasured associated beams to obtain an estimated value of the unmeasured search beams; and generating millimeter-wave beam tracking samples based on historical millimeter-wave beam differential vectors, the measurement values ​​of the search beams, and the estimated values ​​of the unmeasured search beams.

[0072] The search beam can be a beam associated with the differential vector of a historical millimeter-wave beam. The search probability can be used to describe the search value of the beam. The search threshold can be a pre-set value used to compare with the search probability of the search beam. The beam search range can be the tracking range of the beam, used to identify beams with a search probability greater than the search threshold. The unmeasured search beam can be a beam outside the beam search range that needs to be measured. The unmeasured associated beam can be a successfully measured beam associated with the unmeasured search beam. Optionally, the unmeasured associated beam and the unmeasured search beam are at a certain angle. The unmeasured search beam estimate can be an estimate of the search value of the unmeasured search beam based on the unmeasured associated beam using an estimation function.

[0073] In this embodiment of the invention, historical millimeter-wave beam differential vectors can be input into the current beam tracking model. The current beam tracking model outputs the search probability of each search beam, and then compares the search probability of each search beam with a search threshold. Based on the index of the beam whose search probability is greater than the search threshold, the beam search range is determined. Further, it is determined whether there are unmeasured search beams outside the beam search range. If there are unmeasured search beams outside the beam search range, unmeasured associated beams are selected from the successfully measured search beams. Then, the unmeasured associated beams are used to estimate the unmeasured search beams to obtain the estimated value of the unmeasured search beams. Thus, millimeter-wave beam tracking samples are generated based on the historical millimeter-wave beam differential vectors, the measured values ​​of the search beams, and the estimated values ​​of the unmeasured search beams.

[0074] S260. When the current beam tracking model needs to be trained at the transmitting or receiving end, the current beam tracking model is trained based on millimeter-wave beam tracking samples to obtain the target beam tracking model.

[0075] In an optional embodiment of the present invention, before training the current beam tracking model based on millimeter-wave beam tracking samples, the method further includes: determining the position of the target search beam based on the measured value of the search beam and the estimated value of the unmeasured search beam; determining the target horizontal beam index difference and the target vertical beam index difference that match the target search beam position, and determining the millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample when the target horizontal beam index difference is not equal to the horizontal beam index difference to be compared of the beam with the largest measured value, or the target vertical beam index difference is not equal to the vertical beam index difference to be compared of the beam with the largest measured value; or, calculating the equivalent measured value of the search beam; determining the comparison error based on the search probability of the search beam and the equivalent measured value, and determining the millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample when the comparison error is greater than or equal to the comparison error threshold.

[0076] The target search beam position can be the tracking position of the target search beam. The target search beam can be the beam corresponding to the maximum value between the measured value of the search beam and the estimated value of the unmeasured search beam. The target horizontal beam index difference can be the difference between the horizontal beam index of the target search beam at the current time and the horizontal beam index at the previous time. The target vertical beam index difference is the difference between the vertical beam index of the target search beam at the current time and the vertical beam index at the previous time. The beam with the largest measured value can be the beam with the largest measured value among the search beams within the beam search range. The horizontal beam index difference to be compared can be the difference between the horizontal beam index of the beam with the largest measured value at the current time and the horizontal beam index at the previous time. The vertical beam index difference to be compared can be the difference between the vertical beam index of the beam with the largest measured value at the current time and the vertical beam index at the previous time. Valid millimeter-wave beam tracking samples can be valid samples of millimeter-wave beam tracking samples, used to improve the model accuracy of the current beam tracking model. The equivalent measurement value is equivalent to the measurement value of the search beam, and is determined by the quotient of the measurement value of each search beam and the sum of the measurement values ​​of all search beams. The comparison error can be the mean square error of the equivalent measurement value of each search beam and the search probability of the corresponding search beam. The comparison error threshold can be an error threshold compared with the comparison error.

[0077] In this embodiment of the invention, the measured value of the search beam can be compared with the estimated value of the unmeasured search beam to determine the maximum value among the measured value and the estimated value of the unmeasured search beam. The beam corresponding to this maximum value is then used as the target search beam, and the target search beam position is determined. Furthermore, the difference between the current horizontal beam index and the previous horizontal beam index of the target search beam is used as the target horizontal beam index difference that matches the target search beam position. Similarly, the difference between the current vertical beam index and the previous vertical beam index of the target search beam is used as the target vertical beam index difference that matches the target search beam position. This process yields the horizontal beam index difference and the vertical beam index difference for the beam with the largest measured value. If the target horizontal beam index difference is not equal to the horizontal beam index difference for the beam with the largest measured value, or if the target vertical beam index difference is not equal to the vertical beam index difference for the beam with the largest measured value, the millimeter-wave beam tracking sample is determined to be a valid millimeter-wave beam tracking sample. When the target horizontal beam index difference is equal to the difference between the measured maximum beam's horizontal beam index and the target vertical beam index difference is equal to the difference between the measured maximum beam's vertical beam index and the target vertical beam index difference, the millimeter-wave beam tracking sample is determined to be an invalid millimeter-wave beam tracking sample.

[0078] In addition to the methods described above for determining a millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample, the equivalent measurement value of each search beam can be calculated, and the mean square error between the equivalent measurement value of each search beam and the search probability of the corresponding search beam can be calculated to obtain the comparison error. This comparison error is then compared with a comparison error threshold. If the comparison error is greater than or equal to the comparison error threshold, the millimeter-wave beam tracking sample is determined to be a valid millimeter-wave beam tracking sample. If the comparison error is less than the comparison error threshold, the millimeter-wave beam tracking sample is determined to be an invalid millimeter-wave beam tracking sample.

[0079] If there is a sufficient number of valid millimeter-wave beam tracking samples and the current beam tracking model needs to be trained, a portion of the valid millimeter-wave beam tracking samples can be randomly selected to train the current beam tracking model.

[0080] In an optional embodiment of the present invention, the transmitting end or receiving end needs to train the current beam tracking model, which may include: obtaining the target continuous duration during the target time period that matches the historical millimeter-wave beam differential vector, where the measured value of the target search beam is less than the target measurement threshold; if the target continuous duration is greater than the target duration threshold, then determining that the transmitting end or receiving end needs to train the current beam tracking model; or, obtaining the target sample number of valid millimeter-wave beam tracking samples, and if the target sample number is greater than the target number threshold, then determining that the transmitting end or receiving end needs to train the current beam tracking model.

[0081] The target time period can be the period during which historical beams corresponding to historical millimeter-wave beam differential vectors are acquired. The target measurement threshold can be a pre-set lower limit for beam measurement values. The target continuous duration can be the cumulative duration for which the measurement value of the target search beam is continuously less than the target measurement threshold. The target duration threshold can be a threshold compared to the target continuous duration. The target sample number can be the number of effective millimeter-wave beam tracking samples. The target quantity threshold can be a pre-set quantity threshold compared to the target sample number.

[0082] In this embodiment of the invention, it can be determined whether the transmitting end or the receiving end needs to train the beam tracking model using any of the following methods:

[0083] Method 1: First, determine the continuous duration of the target search beam measurement value within the target time period that matches the historical millimeter-wave beam differential vector. Then, compare the continuous duration of the target beam with the target duration threshold. If the continuous duration of the target beam is greater than the target duration threshold, it is determined that the transmitter or receiver needs to train the current beam tracking model. If the continuous duration of the target beam is less than or equal to the target duration threshold, it is determined that the transmitter or receiver does not need to train the current beam tracking model.

[0084] Method 2: Obtain the number of valid millimeter-wave beam tracking samples and use the number of valid millimeter-wave beam tracking samples as the target number of samples. Compare the target number of samples with the target number threshold. If the target number of samples is greater than the target number threshold, it is determined that the transmitter or receiver needs to train the current beam tracking model. If the target number of samples is less than or equal to the target number threshold, it is determined that the transmitter or receiver does not need to train the current beam tracking model.

[0085] Method 3: First, determine the target time period for matching with the historical millimeter-wave beam differential vector. Then, accumulate the number of times the measured value of the target search beam within the target time period is less than the target measurement threshold. Use this accumulated number as the comparison quantity. Compare this comparison quantity with a preset threshold. If the comparison quantity is greater than the preset threshold, it is determined that the transmitter or receiver needs to train the current beam tracking model. If the comparison quantity is less than or equal to the preset threshold, it is determined that the transmitter or receiver does not need to train the current beam tracking model.

[0086] In an optional embodiment of the present invention, training the current beam tracking model based on millimeter-wave beam tracking samples may include: acquiring millimeter-wave beam tracking samples to be trained from the valid millimeter-wave beam tracking samples, and determining the horizontal beam differential vector and the vertical beam differential vector to be trained of the millimeter-wave beam tracking samples to be trained; inputting the horizontal beam differential vector and the vertical beam differential vector to be trained into the current beam tracking model respectively to obtain a first search probability matching the horizontal beam differential vector to be trained and a second search probability matching the vertical beam differential vector to be trained; calculating a first sub-target beam search probability based on the first search probability and the equivalent measurement value matching the horizontal beam differential vector to be trained, and calculating a second sub-target beam search probability based on the second search probability and the equivalent measurement value matching the vertical beam differential vector to be trained; inputting the horizontal beam differential vector to be trained into the current beam tracking model to obtain a first search probability matching the horizontal beam differential vector to be trained and a second search probability matching the vertical beam differential vector to be trained; inputting the horizontal beam differential vector to be trained into the current beam tracking model to obtain a first search probability matching the horizontal beam differential vector to be trained and a second search probability matching the vertical beam differential vector to be trained and a third search probability matching the horizontal beam differential vector to be trained and a fourth search probability matching the horizontal beam differential vector to be trained and a fifth search probability matching the horizontal beam differential vector to be trained and a sixth ... The model takes the horizontal beam difference vector to be trained and the vertical beam difference vector to be trained as input samples of the first model, and the search probabilities of the first sub-target beam and the second sub-target beam as output samples of the first model. Based on the input and output samples of the first model, the current beam tracking model is trained. Alternatively, based on the horizontal and vertical beam difference vectors to be trained, a joint beam difference vector is determined. The joint beam difference vector is input into the current beam tracking model to obtain a third search probability matching the joint beam difference vector. Based on the third search probability and the measured value of the search beam matching the joint beam difference vector, the target beam search probability is calculated. The joint beam difference vector is used as input samples of the second model, and the target beam search probability is used as output samples of the second model. Based on the input and output samples of the second model, the current beam tracking model is trained.

[0087] The millimeter-wave beam tracking samples to be trained can be a subset of the valid millimeter-wave beam tracking samples. The horizontal beam differential vector to be trained can be a second historical horizontal beam differential vector matched with the millimeter-wave beam tracking samples to be trained. The vertical beam differential vector to be trained can be a second historical vertical beam differential vector matched with the millimeter-wave beam tracking samples to be trained. The first search probability can be the search probability obtained by inputting the horizontal beam differential vector to be trained into the current beam tracking model. The second search probability can be the search probability obtained by inputting the vertical beam differential vector to be trained into the current beam tracking model. The first sub-target beam search probability can be the beam search probability determined based on the first search probability, the equivalent measurement value matched with the horizontal beam differential vector to be trained, and a preset weight. The second sub-target beam search probability can be the beam search probability determined based on the second search probability, the equivalent measurement value matched with the vertical beam differential vector to be trained, and a preset weight. For example, assuming the preset weight is 'a', the first search probability is 'value', and the equivalent measurement value matched with the horizontal beam differential vector to be trained is 'value'... Then calculate the product of a and value, and calculate (1-a) and... The product of the two products is used as the search probability of the first subtarget beam. The calculation principle of the search probability of the second subtarget beam is the same as that of the first subtarget beam, and will not be repeated here.

[0088] The input samples for the first model differ from those for the second model. The horizontal beam difference vector and the vertical beam difference vector to be trained, constituting the input samples for the first model, are input separately to the current beam tracking model. The output sample for the first model can be the model output sample corresponding to the input sample for the first model. The joint beam difference vector can be a vector composed of the horizontal beam difference vector and the vertical beam difference vector to be trained. The output sample for the second model can be the model output sample corresponding to the input sample for the second model. The third search probability can be the search probability obtained by inputting the joint beam difference vector into the current beam tracking model. The target beam search probability can be the beam search probability determined based on the third search probability, the measured value of the search beam matched with the joint beam difference vector, and preset weights. The calculation principle for the target beam search probability is the same as that for the first and second sub-target beam search probabilities.

[0089] In this embodiment of the invention, a millimeter-wave beam tracking sample to be trained is randomly selected from the effective millimeter-wave beam tracking samples. Then, the horizontal beam differential vector and the vertical beam differential vector to be trained for each sample are determined. The horizontal beam differential vector is then input into the current beam tracking model to obtain a first search probability matching the horizontal beam differential vector. Similarly, the vertical beam differential vector is input into the current beam tracking model to obtain a second search probability matching the vertical beam differential vector. Further, based on the first search probability and the matching probability of the horizontal beam differential vector... The equivalent measurement value and preset weights are used to calculate the search probability of the first sub-target beam. Based on the second search probability, the equivalent measurement value matching the vertical beam difference vector to be trained, and preset weights, the search probability of the second sub-target beam is calculated. Finally, the horizontal beam difference vector and the vertical beam difference vector to be trained are used as input samples to the first model, and the search probabilities of the first and second sub-target beams are used as output samples to the first model. These input and output samples are then input into the current beam tracking model to train it. Alternatively,

[0090] The horizontal beam differential vector and the vertical beam differential vector to be trained are combined to obtain a joint beam differential vector. This joint beam differential vector is used as the input sample for the second model. The joint beam differential vector is then input into the current beam tracking model to obtain a third search probability that matches the joint beam differential vector. Based on the third search probability, the measured value of the search beam that matches the joint beam differential vector, and the preset weights, the target beam search probability is calculated. The joint beam differential vector is then used as the input sample for the second model, and the target beam search probability is used as the output sample for the second model. Finally, the second model input sample and the second model output sample are input into the current beam tracking model to train the current beam tracking model.

[0091] The technical solution of this invention involves performing differential operations on the historical horizontal beam index sequence and the historical vertical beam index sequence in the historical millimeter-wave beam index sequence at a transmitting or receiving end to obtain a first historical horizontal beam differential vector and a first historical vertical beam differential vector. Then, based on the first non-zero element of the first historical horizontal beam differential vector, a horizontal flip flag value is determined, and based on the first non-zero element of the first historical vertical beam differential vector, a vertical flip flag value is determined. Finally, the first historical horizontal beam differential vector is updated based on the horizontal flip flag values ​​to obtain a second historical... The horizontal beam differential vector is generated, and the first historical vertical beam differential vector is updated according to the vertical flip flag value to obtain the second historical vertical beam differential vector. Based on the second historical horizontal beam differential vector and the second historical vertical beam differential vector, the historical millimeter-wave beam differential vector is determined. Furthermore, based on the historical millimeter-wave beam differential vector and the current beam tracking model, millimeter-wave beam tracking samples are generated. Thus, when the current beam tracking model needs to be trained at the transmitting or receiving end, the current beam tracking model is trained based on the millimeter-wave beam tracking samples to obtain the target beam tracking model. In this scheme, the millimeter-wave beam index can be used to quickly determine the possible location of the millimeter-wave beam to be searched. Therefore, when the speed of the transmitting or receiving end changes from slow to fast, the millimeter-wave beam index can be used to effectively determine the possible location of the tracking millimeter-wave beam. When the high-speed movement of the transmitting or receiving end causes the beam to jump or even jump significantly between two adjacent tracking cycles, the difference of the millimeter-wave beam index can be used to determine the possible jump position of the current beam without having to start searching from the neighborhood of the beam position at the previous moment. This solves the problems of high millimeter-wave beam search overhead and high algorithm complexity in the prior art, reduces the time overhead of millimeter-wave beam search, maintains stable tracking of millimeter-wave beams transmitted by fast-moving terminals, and can effectively adapt to highly dynamic communication environments.

[0092] Example 3

[0093] Figure 3 This is a flowchart of a beam tracking method provided in Embodiment 3 of the present invention. This embodiment is applicable to the precise beam tracking of millimeter waves. The method can be executed by a beam tracking device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 3 As shown, the method includes:

[0094] S310. Obtain the current millimeter-wave beam differential vector.

[0095] The current millimeter-wave beam differential vector can be the millimeter-wave beam differential vector that needs to be input into the target beam tracking model.

[0096] In this embodiment of the invention, the current millimeter-wave beam differential vector can be obtained according to beam tracking requirements. The determination principle of the current millimeter-wave beam differential vector is the same as that of the historical millimeter-wave beam differential vector, only the acquisition time of the beam is not exactly the same. Therefore, the specific generation process of the current millimeter-wave beam differential vector will not be described in detail here.

[0097] S320. Analyze the current millimeter-wave beam differential vector using the target beam tracking model to determine the target beam search range that matches the current millimeter-wave beam differential vector.

[0098] The target beam search range can be the beam range of the desired millimeter wave output by the target beam tracking model. The target beam tracking model can be a model trained by the model training method in any of the embodiments.

[0099] In this embodiment of the invention, after the current millimeter-wave beam differential vector is input into the target beam tracking model, the target beam tracking model analyzes the current millimeter-wave beam differential vector and outputs the target beam search range that matches the current millimeter-wave beam differential vector.

[0100] S330: The transmitting end transmits a signal according to the target beam search range and obtains the measurement value of the received beam fed back by the receiving end.

[0101] In this embodiment of the invention, the transmitting end can transmit at least one signal within the target beam search range, and the receiving end can measure the received signal transmitted by the transmitting end and feed back the measured value of the received beam to the transmitting end.

[0102] S340. Determine the target tracking beam based on the measured value of the received beam fed back by the receiver.

[0103] The target tracking beam can be a millimeter-wave beam corresponding to the maximum measurement value fed back by the receiver, i.e., the required tracking beam.

[0104] In this embodiment of the invention, the target tracking beam can be determined by comparing the measured value of the received beam fed back by the receiving end with the measurement value of the receiving end, thereby transmitting millimeter-wave signals through the target tracking beam.

[0105] The technical solution of this invention obtains the current millimeter-wave beam differential vector, then analyzes the current millimeter-wave beam differential vector using a target beam tracking model to determine the target beam search range matching the current millimeter-wave beam differential vector. The transmitting end then transmits a signal according to the target beam search range and obtains the measured value of the received beam fed back by the receiving end. Based on the measured value of the received beam fed back by the receiving end, the target tracking beam is determined. This solves the problems of high millimeter-wave beam search overhead and high algorithm complexity in existing technologies, reduces the time overhead of millimeter-wave beam search, maintains stable tracking of millimeter-wave beams transmitted by rapidly moving terminals, and can effectively adapt to highly dynamic communication environments.

[0106] Example 4

[0107] Embodiment 4 of the present invention provides an optional embodiment for model training and beam tracking, the specific implementation of which can be found in the following embodiments. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here.

[0108] 1. The transmitting or receiving end inputs the millimeter-wave beam differential vector (historical millimeter-wave beam differential vector) from several past moments into the current beam tracking model, and the current beam tracking model outputs the beam search range.

[0109] 2. The transmitting end transmits a reference signal on a beam within the beam search range, and the receiving end receives the reference signal using the same receiving beam and measures the reference signal. The receiving end then feeds back the measured value to the transmitting end, which selects the target tracking beam for subsequent data transmission based on the measured value. Alternatively, the transmitting end transmits several reference signals on beams within the beam search range, and the receiving end receives the reference signals on the receiving beams within the aforementioned range and measures the reference signals. Based on the measurement results, the receiving end selects the optimal beam for subsequent data transmission.

[0110] 3. The beam differential vector, the measured value of the search beam, and the estimated value of the unmeasured search beam are combined into a millimeter-wave beam tracking sample. The transmitting or receiving end can also determine the validity of the millimeter-wave beam tracking sample and store the valid millimeter-wave beam tracking sample in the memory.

[0111] 4. The transmitting or receiving end determines whether the current beam tracking model needs to be trained. If so, it takes several valid millimeter-wave beam tracking samples to train the current beam tracking model and deletes invalid millimeter-wave beam tracking samples after training.

[0112] Optionally, the historical millimeter-wave beam differential vector can be obtained through the following steps: the transmitting or receiving end performs a differential operation on the horizontal index sequence of the beam at several past moments (historical beam horizontal index sequence) and the vertical index sequence of the beam at several past moments (historical beam vertical index sequence) to obtain the first historical horizontal beam differential vector [SH1,SH2,…,SH]. K ] and the first historical vertical beam differential vector [SV1,SV2,…,SV K ].

[0113] Determine if the first non-zero element of the first historical horizontal beam differential vector is positive. If it is, record the horizontal flip flag (horizontal_reverse_flag) as 0. If it is negative, multiply all elements of the first historical horizontal beam differential vector by -1 to obtain the second historical horizontal beam differential vector [-SH1, -SH2, ..., -SH]. K The horizontal flip flag value is recorded as 1.

[0114] Determine if the first non-zero element of the first historical vertical beam differential vector is positive. If it is, record the vertical flip flag (vertical_reverse_flag) as 0. If it is negative, multiply all elements of the first historical vertical beam differential vector by -1 to obtain the second historical vertical beam differential vector [-SV1, -SV2, ..., -SV]. K ], and record that the vertical flip flag vertical_reverse_flag is 1, that is, the value of the vertical flip flag is 1.

[0115] For example, the index sequence of the beam in the horizontal and vertical directions over the past K+1 times can be differentially analyzed to obtain the historical millimeter-wave beam differential vector.

[0116] Optionally, the beam search range can be obtained in any of the following ways:

[0117] Method 1: Input the second historical horizontal beam differential vector into the current beam tracking model. The current beam tracking model outputs the search probabilities of all horizontal beams. Then, sort the search probabilities of the horizontal beams in descending order and select those greater than value_x. thres N search probabilities [value_x1, value_x2, ..., value_x] N The differences between the horizontal beam index of the corresponding beam and the horizontal beam index of the beam at the previous time step are denoted as H1, H2, ..., H1, respectively. NIf the horizontal flip flag value is 1, the horizontal beam index difference will be flipped to H1 = -H1, H2 = -H2, ..., H... N =-H N If the horizontal flip flag value is 0, it remains unchanged; the second historical vertical beam differential vector is input into the current beam tracking model, which outputs the search probabilities of all vertical beams and sorts them in descending order, selecting those greater than the threshold value_y. thres M search probabilities [value_y1, value_y2, ..., value_y] M The differences between the corresponding vertical beam index and the vertical index of the beam at the previous time step are denoted as V1, V2, ..., V. N If the vertical flip flag value is 1, the vertical beam index difference will be flipped to V1 = -V1, V2 = -V2, ..., V M =-V M If the vertical flip flag value is 0, it remains unchanged. Then, starting from the beam position at the previous moment, it moves horizontally and vertically by (H1, V1), (H2, V1), ..., (H... N (H1,V1), (H1,V2), ..., (H1,V2) M There are N+M-1 beams to be searched.

[0118] For example, if the second historical horizontal beam difference vector is [0, 2, 1] and the horizontal flip flag value is 0, based on the output of the current beam tracking model, relative to the horizontal beam index at the previous moment, it is assumed that the three horizontal beam indices with a difference of 1 (advancing 1 beam), -1 (retreating 1 beam), and 0 (not moving) have a higher search probability; the second historical vertical beam difference vector is [0, 1, -1]. Based on the output of the current beam tracking model, relative to the vertical beam index at the previous moment, the difference values ​​are 0 (not moving), 1 (advancing), -1 (retreating 1 beam), and -1 (reversing). The four vertical beam indices (1 beam), -1 (backward 1 beam), and 2 (forward 2 beams) have a high search probability. If the vertical flip flag value is 1, the difference in the flipped vertical beam index is 0 (no change), -1 (backward 1 beam), 1 (forward 1 beam), and -2 (backward 2 beams). The beam search range can start from the beam position of the previous moment and move horizontally and vertically by (1,0), (1,-1), (1,1), (1,-2), (-1,0), and (0,0) respectively, for a total of 6 search beams.

[0119] Method 2: Combine the second historical horizontal beam differential vector with the second historical vertical beam differential vector, represented as [SH1, SH2, ..., SH K ,SV1,SV2,…,SV KInput the current beam tracking model; the current beam tracking model outputs the search probability corresponding to the search beam; sort the search probabilities corresponding to the search beams in descending order, and select those greater than the threshold value. thres The first N probability values ​​[value1, value2, ..., value N The differences between their corresponding horizontal and vertical beam indices and the beam index at the previous moment are denoted as (H1,V1), (H2,V2), ..., (H2,V2), respectively. N V N If the horizontal flip flag value is 1, the selected N horizontal beam index differences will be flipped to H1 = -H1, H2 = -H2, ..., H... N =-H N If the horizontal flip flag value is 0, it remains unchanged; if the vertical flip flag value is 1, the selected N vertical beam index differences are flipped to V1 = -V1, V2 = -V2, ..., V N =-V N If the vertical flip flag value is 0, it remains unchanged, and then, starting from the beam position of the previous moment, it moves (H1,V1), (H2,V2), ..., (H2,V2) in the horizontal and vertical directions respectively. N V N There are N beams to be searched.

[0120] For example, assuming the second historical horizontal beam difference vector is [0,2,1] and the second historical vertical beam difference vector is [0,1,-1], according to the output of the current beam tracking model, the seven beams with horizontal and vertical beam index differences of (0,1), (0,-1), (0,0), (1,1), (1,0), (1,-1), and (-1,1) relative to the previous moment have a high search probability. When the horizontal flip flag value is 0 and the vertical flip flag value is 1, the flipped beam index difference is (0,-1), (0,1), (0,0), (1,-1), (1,0), (1,1), and (-1,-1). Then the beam search range starts from the beam position of the previous moment and moves horizontally and vertically by (0,-1), (0,1), (0,0), (1,-1), (1,0), (1,1), and (-1,-1) respectively, for a total of seven search beams.

[0121] Optionally, the estimated value of the unmeasured search beam can be determined in the following way to determine whether there are 3 beams (H) in the measured beam that satisfy one of the following conditions. a V a ), (H b V b ) and (H c V c): ①H b =H a -1 and V b =V a And H c =H a And V c =V a -1; ②H b =H a +1 and V b =V a And H c =H a And V c =V a -1;③H b =H a -1 and V b =V a And H c =H a And V c =V a +1; ④H b =H a +1 and V b =V a And H c =H a And V c =V a +1. If so, the above three beams can be considered as unmeasured associated beams, and the measurement values ​​corresponding to the above three beams are P respectively. a P b P c If the difference between the horizontal and vertical beam indices is (H b V c The beam was not measured, meaning the difference between the horizontal and vertical beam indices was (H). b V c The beam of the unmeasured search beam can be detected through... Make an estimate, and As an estimate of the unmeasured search beam, the measured value of the search beam is then compared with all the estimates of the unmeasured search beam, and the optimal result is denoted as P(H). opt V opt ), H opt and V opt This represents the difference between the horizontal and vertical indices of the beam corresponding to the optimal result and the beam at the previous time step. The transmitting or receiving end moves the beam index from the previous time step by H in the horizontal and vertical directions respectively. opt and V opt Then, subsequent data transmission will proceed.

[0122] Optionally, the measured values ​​P1, P2, ..., P of the search beam can be...N Convert to equivalent measurement value in, If P i >P j To indicate that the i-th beam is superior to the j-th beam, f() should be a monotonically non-decreasing function if P i <P j This indicates that the i-th beam is superior to the j-th beam, and f() should be a monotonically non-increasing function. For example, or in, This is the floor operator.

[0123] The second historical horizontal beam differential vector, the second historical vertical beam differential vector, the measured values ​​of all search beams, and the estimated values ​​of unmeasured search beams are combined to form a millimeter-wave beam tracking sample.

[0124] After beam search is complete, the validity of the currently generated millimeter-wave beam tracking sample can be determined, as can the validity of already stored millimeter-wave beam tracking samples. The methods for determining the validity of millimeter-wave beam tracking samples are the same in both cases. The difference is that if the currently generated millimeter-wave beam tracking sample is determined to be valid, it is stored. If the already stored millimeter-wave beam tracking sample is determined to be valid, no further processing is required. If the currently generated millimeter-wave beam tracking sample is determined to be invalid, it is deleted. If the already stored millimeter-wave beam tracking sample is determined to be invalid, it is also deleted.

[0125] When determining the validity of the generated millimeter-wave beam tracking samples, the second historical horizontal beam difference vector and the second historical vertical beam difference vector of each sample can be input into the current beam tracking model to obtain the search probability of the beam corresponding to each sample. The maximum value of the search probability of the beam corresponding to each sample is compared with the measured value of the search beam to determine the optimal beam (target search beam). It is then determined whether the optimal beam is consistent with the beam corresponding to the maximum measured value (maximum measured beam). Specifically, it is determined whether the difference in the target horizontal beam index of the target search beam is consistent with the difference in the horizontal beam index of the maximum measured beam, and whether the difference in the target vertical beam index is consistent with the difference in the vertical beam index of the maximum measured beam. If they are consistent, the millimeter-wave beam tracking sample is invalid; otherwise, the generated millimeter-wave beam tracking sample is valid. Alternatively,

[0126] The maximum search probability of each sample's corresponding beam is compared with the equivalent measurement value to determine the optimal beam. Then, it is determined whether the optimal beam matches the beam corresponding to the maximum equivalent measurement value. Specifically, it is checked whether the difference in the target horizontal beam index of the target search beam matches the difference in the horizontal beam index of the beam corresponding to the maximum equivalent measurement value, and whether the difference in the target vertical beam index of the target search beam matches the difference in the vertical beam index of the beam corresponding to the maximum equivalent measurement value. If they match, the millimeter-wave beam tracking sample is considered invalid; otherwise, the generated millimeter-wave beam tracking sample is considered valid. Alternatively,

[0127] Calculate the mean square error between the search probability vector and the equivalent measurement vector in the currently generated millimeter-wave beam tracking samples. If the error to be compared is less than the threshold, the currently generated millimeter-wave beam tracking sample is an invalid millimeter-wave beam tracking sample; if the error to be compared is greater than or equal to the threshold, the valid beam training sample is stored in the memory.

[0128] Optionally, the need for training of the current beam tracking model can be determined based on any of the following criteria: The transmitter or receiver counts the number of comparisons where the measurement value of the optimal beam (target search beam) is inferior to the target measurement threshold within a certain time range (target time period). If the number of comparisons exceeds the target number threshold, the current beam tracking model needs to be trained. The transmitter or receiver counts the continuous duration for which the measurement value corresponding to the optimal beam is continuously inferior to the target measurement threshold. If the continuous duration exceeds the target duration threshold, the current beam tracking model needs to be trained. The transmitter or receiver counts the number of valid millimeter-wave beam tracking samples stored in the memory. If this number exceeds the number threshold, the current beam tracking model needs to be trained.

[0129] Optionally, the current beam tracking model can be trained using one of the following methods: Take several millimeter-wave beam tracking samples to be trained from the effective millimeter-wave beam tracking samples, input the horizontal beam difference vector and the vertical beam difference vector of each sample into the current beam tracking model, and then output the horizontal beam search probability vector [value_x1, value_x2, ..., value_x] respectively. N The search probability vector [value_y1, value_y2, ..., value_y] and the vertical beam. N The search probability in each millimeter-wave beam tracking sample to be trained is calculated using equivalent measurements. Update, for example The horizontal beam difference vector and vertical beam difference vector of each millimeter-wave beam tracking sample to be trained are used as the input samples of the first model. The search probabilities of the horizontal beam (first sub-target beam search probability) and the search probabilities of the vertical beam (second sub-target beam search probability) in the updated millimeter-wave beam tracking samples to be trained are used as the output samples of the first model. Then, the beam tracking model is trained using the first model input samples and the first model output samples. Alternatively,

[0130] Several millimeter-wave beam tracking samples to be trained are extracted from the effective millimeter-wave beam tracking samples. Then, the horizontal beam difference vector and the vertical beam difference vector to be trained of each millimeter-wave beam tracking sample are combined into a joint beam difference vector [SH1,SH2,…,SH]. K ,SV1,SV2,…,SV K Input the current beam tracking model, and output the search probabilities [value1, value2, ..., value] for all beams. N Further utilize equivalent measurement values Update the corresponding search probabilities to obtain the target beam search probabilities. For example, The joint beam differential vector [SH1,SH2,…,SH] K ,SV1,SV2,…,SV K The target beam search probability is used as the second model input sample of the current beam tracking model, and the target beam search probability is used as the second model output sample of the current beam tracking model to train the current beam tracking model.

[0131] Figure 4 This is a comparison chart of the measured value of a beam and the estimated value of the unmeasured search beam provided in Embodiment 4 of the present invention, as shown in the figure. Figure 4 As shown, the measured values ​​of the beam are 0.8, 0.2, 0.6, 0.5, and 0.3, respectively, and the corresponding estimated values ​​of the unmeasured search beam are 0.4, 0.7, 0, and 0.2, respectively.

[0132] For example, using equivalent measurements pass When updating the corresponding search probabilities, it is assumed that the search probabilities of beams with horizontal and vertical beam index differences of (1,0) and (0,-1) are 0.5 and 0.7, respectively, and the equivalent measured values ​​are 1 and 0, respectively. That is, the updated target beam search probabilities are 0.5α+(1-α) and 0.7α.

[0133] Figure 5 This is a simulation diagram of the current beam tracking model tracking the transmitted beam of a base station, as provided in Embodiment 4 of the present invention. Figure 5As shown, the beam tracking periods are 1s, 0.5s, and 0.2s, the user's movement speed is 1m / s, and the base station is suspended at a height of 3m. When the user moves on a plane at a height of 0m, the movement range is 0m to 20m in both the x-axis and y-axis directions. The simulation results show that in this communication scenario, when the signal-to-noise ratio is high, the beam tracking error is very small and the algorithm performance is excellent.

[0134] Figure 6 This is a convergence diagram of an algorithm for tracking the transmitted beam of a base station using a current beam tracking model, provided in Embodiment 4 of the present invention. The beam tracking periods are 1s, 0.5s, and 0.2s, the user's movement speed is 1m / s, the base station (BS) is suspended at a height of 3m, and the user moves on a plane at a height of 0m, with a movement range of 0m to 20m in both the x and y axes. The signal-to-noise ratio (SNR) is 15dB. Simulation results show that the algorithm converges quickly with beam tracking periods of 0.5s and 0.2s, reaching convergence in approximately 200-300 seconds. With a beam tracking period of 1s, due to slower sample generation and increased tracking difficulty, the algorithm requires approximately 500s to reach near convergence, achieving a relatively good convergence state in approximately 2000 seconds.

[0135] Example 5

[0136] Figure 7 This is a schematic diagram of a model training device provided in Embodiment 5 of the present invention. Figure 7 As shown, the device includes: a sample generation module 510 and a target beam tracking model acquisition module 520, wherein,

[0137] The sample generation module 510 is used to generate millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences.

[0138] The target beam tracking model acquisition module 520 is used to train the current beam tracking model based on millimeter-wave beam tracking samples to obtain the target beam tracking model when the current beam tracking model needs to be trained at the transmitting end or receiving end.

[0139] A target beam tracking model for tracking millimeter-wave beams.

[0140] The technical solution of this invention generates millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences. Then, when the transmitting or receiving end needs to train the current beam tracking model, it trains the current beam tracking model based on these samples to obtain the target beam tracking model. In this solution, the millimeter-wave beam index can be used to quickly determine the possible positions of the millimeter-wave beam to be searched. Therefore, when the movement speed of the transmitting or receiving end changes from slow to fast, the millimeter-wave beam index can be used to effectively determine the possible positions of the tracked millimeter-wave beam. Furthermore, when the high-speed movement of the transmitting or receiving end causes beam jumps or even large jumps between adjacent tracking cycles, the difference in the millimeter-wave beam index can be used to determine the possible jump position of the current beam without having to start searching from the neighborhood of the beam position at the previous moment. This solves the problems of high millimeter-wave beam search overhead and high algorithm complexity in existing technologies, reduces the time overhead of millimeter-wave beam search, maintains stable tracking of millimeter-wave beams transmitted by fast-moving terminals, and can effectively adapt to highly dynamic communication environments.

[0141] Optionally, the sample generation module 510 includes: a first historical difference vector acquisition unit, a flip flag value determination unit, a second historical difference vector acquisition unit, a historical millimeter-wave beam difference vector determination unit, and a sample generation unit, wherein,

[0142] The first historical differential vector acquisition unit is used to perform differential operations on the historical horizontal beam index sequence and the historical vertical beam index sequence in the historical millimeter-wave beam index sequence through a transmitting end or a receiving end to obtain a first historical horizontal beam differential vector and a first historical vertical beam differential vector; the flip flag value determination unit is used to determine the horizontal flip flag value based on the first non-zero element of the first historical horizontal beam differential vector, and to determine the vertical flip flag value based on the first non-zero element of the first historical vertical beam differential vector; the second historical differential vector acquisition unit is used to determine the horizontal flip flag value based on the first non-zero element of the first historical vertical beam differential vector. The first historical horizontal beam differential vector is updated according to the flip flag value to obtain the second historical horizontal beam differential vector, and the first historical vertical beam differential vector is updated according to the vertical flip flag value to obtain the second historical vertical beam differential vector; the historical millimeter-wave beam differential vector determination unit is used to determine the historical millimeter-wave beam differential vector according to the second historical horizontal beam differential vector and the second historical vertical beam differential vector; the sample generation unit is used to generate the millimeter-wave beam tracking sample according to the historical millimeter-wave beam differential vector and the current beam tracking model.

[0143] Optionally, a sample generation unit is configured to input the historical millimeter-wave beam differential vector into the current beam tracking model to obtain the search probability of each search beam; determine the beam search range based on the search probability of each search beam and the search threshold; if there are unmeasured search beams outside the beam search range, obtain the unmeasured associated beams; obtain the measurement value of the search beams, and estimate the unmeasured search beams using the unmeasured associated beams to obtain the estimated value of the unmeasured search beams; and generate the millimeter-wave beam tracking sample based on the historical millimeter-wave beam differential vector, the measurement value of the search beams, and the estimated value of the unmeasured search beams.

[0144] Optionally, the model training device further includes a valid sample determination module, used to determine the target search beam position based on the measured value of the search beam and the estimated value of the unmeasured search beam; determine the target horizontal beam index difference and the target vertical beam index difference that match the target search beam position, and determine the millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample when the target horizontal beam index difference is not equal to the horizontal beam index difference to be compared of the beam with the largest measured value, or when the target vertical beam index difference is not equal to the vertical beam index difference to be compared of the beam with the largest measured value; or, calculate the equivalent measured value of the search beam; determine the comparison error based on the search probability of the search beam and the equivalent measured value, and determine the millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample when the comparison error is greater than or equal to the comparison error threshold.

[0145] Optionally, the target beam tracking model acquisition module 520 includes a training judgment unit, which is used to acquire the continuous duration of the target search beam measurement value within the target time period that matches the historical millimeter-wave beam differential vector, where the continuous duration of the target beam is less than the target measurement threshold; if the continuous duration of the target beam is greater than the target duration threshold, it is determined that the transmitting end or the receiving end needs to train the current beam tracking model; or, it acquires the target sample number of the effective millimeter-wave beam tracking samples, and if the target sample number is greater than the target number threshold, it is determined that the transmitting end or the receiving end needs to train the current beam tracking model.

[0146] Optionally, the beam tracking training unit is specifically configured to acquire the millimeter-wave beam tracking samples to be trained from the effective millimeter-wave beam tracking samples, and determine the horizontal beam differential vector and the vertical beam differential vector to be trained for the millimeter-wave beam tracking samples to be trained; input the horizontal beam differential vector and the vertical beam differential vector to be trained into the current beam tracking model respectively to obtain a first search probability matching the horizontal beam differential vector and a second search probability matching the vertical beam differential vector; calculate a first sub-target beam search probability based on the first search probability and the equivalent measurement value matching the horizontal beam differential vector, and calculate a second sub-target beam search probability based on the second search probability and the equivalent measurement value matching the vertical beam differential vector; and use the horizontal beam differential vector and the vertical beam differential vector to be trained as the first... The model takes input samples and uses the search probabilities of the first and second sub-target beams as first model output samples. Based on the first model input samples and the first model output samples, the current beam tracking model is trained. Alternatively, a joint beam difference vector is determined based on the horizontal and vertical beam difference vectors to be trained. The joint beam difference vector is input into the current beam tracking model to obtain a third search probability matching the joint beam difference vector. Based on the third search probability and the measured value of the search beam matching the joint beam difference vector, the target beam search probability is calculated. The joint beam difference vector is used as a second model input sample, and the target beam search probability is used as a second model output sample. Based on the second model input samples and the second model output samples, the current beam tracking model is trained.

[0147] The model training apparatus provided in this embodiment of the invention can execute the model training method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0148] Example 6

[0149] Figure 8 This is a schematic diagram of a beam tracking device provided in Embodiment Six of the present invention. Figure 8 As shown, the device includes: a differential vector acquisition module 610, a target beam search range determination module 620, a signal transmission and measurement value acquisition module 630, and a target tracking beam determination module 640, wherein,

[0150] The differential vector acquisition module 610 is used to acquire the current millimeter-wave beam differential vector;

[0151] The target beam search range determination module 620 is used to analyze the current millimeter-wave beam differential vector through the target beam tracking model and determine the target beam search range that matches the current millimeter-wave beam differential vector.

[0152] The signal transmission and measurement value acquisition module 630 is used to transmit a signal through the transmitting end according to the target beam search range and acquire the measurement value of the received beam fed back by the receiving end.

[0153] The target tracking beam determination module 640 is used to determine the target tracking beam based on the measurement value of the received beam fed back by the receiver.

[0154] The target beam tracking model is a model trained using the model training method in any embodiment.

[0155] The technical solution of this invention obtains the current millimeter-wave beam differential vector, then analyzes the current millimeter-wave beam differential vector using a target beam tracking model to determine the target beam search range matching the current millimeter-wave beam differential vector. The transmitting end then transmits a signal according to the target beam search range and obtains the measured value of the received beam fed back by the receiving end. Based on the measured value of the received beam fed back by the receiving end, the target tracking beam is determined. This solves the problems of high millimeter-wave beam search overhead and high algorithm complexity in existing technologies, reduces the time overhead of millimeter-wave beam search, maintains stable tracking of millimeter-wave beams transmitted by rapidly moving terminals, and can effectively adapt to highly dynamic communication environments.

[0156] The beam tracking device provided in the embodiments of the present invention can execute the beam tracking method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0157] Example 7

[0158] Figure 9 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0159] like Figure 9As shown, the electronic device 80 includes at least one processor 81 and a memory, such as a read-only memory (ROM) 82 and a random access memory (RAM) 83, communicatively connected to the at least one processor 81. The memory stores computer programs executable by the at least one processor. The processor 81 can perform various appropriate actions and processes based on the computer program stored in the ROM 82 or loaded from storage unit 88 into the RAM 83. The RAM 83 can also store various programs and data required for the operation of the electronic device 80. The processor 81, ROM 82, and RAM 83 are interconnected via a bus 84. An input / output (I / O) interface 85 is also connected to the bus 84.

[0160] Multiple components in electronic device 80 are connected to I / O interface 85, including: input unit 86, such as keyboard, mouse, etc.; output unit 87, such as various types of monitors, speakers, etc.; storage unit 88, such as disk, optical disk, etc.; and communication unit 89, such as network card, modem, wireless transceiver, etc. Communication unit 89 allows electronic device 80 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0161] Processor 81 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 81 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 81 performs the various methods and processes described above, such as model training methods or beam tracking methods.

[0162] In some embodiments, the model training method or beam tracking method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 88. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 80 via ROM 82 and / or communication unit 89. When the computer program is loaded into RAM 83 and executed by processor 81, one or more steps of the model training method or beam tracking method described above may be performed. Alternatively, in other embodiments, processor 81 may be configured to execute the model training method or beam tracking method by any other suitable means (e.g., by means of firmware).

[0163] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0164] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0165] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0166] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0167] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0168] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0169] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0170] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A model training method, characterized in that, include: Generate millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences; When the current beam tracking model needs to be trained at the transmitting or receiving end, the current beam tracking model is trained based on the millimeter-wave beam tracking samples to obtain the target beam tracking model. The target beam tracking model is used to track millimeter-wave beams; The process of generating millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences includes: By performing differential operations on the historical horizontal and vertical index sequences of the historical millimeter-wave beam index sequence at either the transmitting or receiving end, a first historical horizontal beam differential vector and a first historical vertical beam differential vector are obtained. A horizontal flip flag value is determined based on the first non-zero element of the first historical horizontal beam differential vector, and a vertical flip flag value is determined based on the first non-zero element of the first historical vertical beam differential vector. The first historical horizontal beam differential vector is updated based on the horizontal flip flag value to obtain a second historical horizontal beam differential vector, and the first historical vertical beam differential vector is updated based on the vertical flip flag value to obtain a second historical vertical beam differential vector. A historical millimeter-wave beam differential vector is determined based on the second historical horizontal and vertical beam differential vectors. The millimeter-wave beam tracking sample is generated based on the historical millimeter-wave beam differential vectors and the current beam tracking model.

2. The method of claim 1, wherein, The step of generating the millimeter-wave beam tracking sample based on the historical millimeter-wave beam differential vector and the current beam tracking model includes: The historical millimeter-wave beam differential vector is input into the current beam tracking model to obtain the search probability of each search beam; The beam search range is determined based on the search probability and search threshold of each search beam; If there are unmeasured search beams outside the beam search range, then obtain the unmeasured associated beams; The measured value of the search beam is obtained, and the unmeasured search beam is estimated using the unmeasured associated beam to obtain the estimated value of the unmeasured search beam; The millimeter-wave beam tracking sample is generated based on the historical millimeter-wave beam differential vector, the measured value of the search beam, and the estimated value of the unmeasured search beam.

3. The method according to claim 1, characterized in that, Before training the current beam tracking model based on the millimeter-wave beam tracking samples, the method further includes: The position of the target search beam is determined based on the measured value of the search beam and the estimated value of the unmeasured search beam. The target horizontal beam index difference and target vertical beam index difference are determined to match the target search beam position. If the target horizontal beam index difference is not equal to the horizontal beam index difference of the measured maximum beam to be compared, or if the target vertical beam index difference is not equal to the vertical beam index difference of the measured maximum beam to be compared, then the millimeter-wave beam tracking sample is determined to be a valid millimeter-wave beam tracking sample; or... Calculate the equivalent measurement value of the search beam; determine the comparison error based on the search probability of the search beam and the equivalent measurement value; and determine the millimeter-wave beam tracking sample as a valid millimeter-wave beam tracking sample when the comparison error is greater than or equal to the comparison error threshold.

4. The method according to claim 3, characterized in that, The transmitting or receiving end needs to train the current beam tracking model, including: Within the target time period matched with historical millimeter-wave beam differential vectors, obtain the continuous duration of the target search beam measurement value that is less than the target measurement threshold; if the continuous duration of the target is greater than the target duration threshold, determine whether the transmitter or receiver needs to train the current beam tracking model; or, Obtain the number of target samples of the effective millimeter-wave beam tracking samples. If the number of target samples is greater than the target number threshold, it is determined that the transmitting end or the receiving end needs to train the current beam tracking model.

5. The method according to claim 4, characterized in that, The step of training the current beam tracking model based on the millimeter-wave beam tracking samples includes: Obtain the millimeter-wave beam tracking sample to be trained from the effective millimeter-wave beam tracking sample, and determine the horizontal beam differential vector and the vertical beam differential vector to be trained of the millimeter-wave beam tracking sample to be trained. The horizontal beam differential vector to be trained and the vertical beam differential vector to be trained are respectively input into the current beam tracking model to obtain a first search probability that matches the horizontal beam differential vector to be trained and a second search probability that matches the vertical beam differential vector to be trained. The first sub-target beam search probability is calculated based on the first search probability and the equivalent measurement value matching the horizontal beam differential vector to be trained, and the second sub-target beam search probability is calculated based on the second search probability and the equivalent measurement value matching the vertical beam differential vector to be trained. The horizontal beam difference vector and the vertical beam difference vector to be trained are used as input samples of the first model, and the first sub-target beam search probability and the second sub-target beam search probability are used as output samples of the first model; beam tracking training is performed on the current beam tracking model based on the first model input samples and the first model output samples; or... The joint beam differential vector is determined based on the horizontal beam differential vector to be trained and the vertical beam differential vector to be trained. The joint beam differential vector is input into the current beam tracking model to obtain the third search probability that matches the joint beam differential vector. Based on the third search probability and the measurement value of the search beam that matches the joint beam differential vector, the target beam search probability is calculated. The joint beam difference vector is used as the input sample of the second model, and the target beam search probability is used as the output sample of the second model; the current beam tracking model is trained based on the input sample and the output sample of the second model.

6. A beam tracking method, characterized in that, include: Obtain the current millimeter-wave beam differential vector; The target beam search range matching the current millimeter-wave beam differential vector is determined by analyzing the current millimeter-wave beam differential vector using a target beam tracking model. The transmitting end transmits a signal according to the target beam search range and obtains the measured value of the received beam fed back by the receiving end. The target tracking beam is determined based on the measured value of the received beam fed back by the receiver. The target beam tracking model is a model trained using any one of the model training methods described in claims 1-5.

7. A model training device, characterized in that, include: The sample generation module is used to generate millimeter-wave beam tracking samples based on historical millimeter-wave beam index sequences. The target beam tracking model acquisition module is used to train the current beam tracking model based on the millimeter-wave beam tracking samples to obtain the target beam tracking model when the current beam tracking model needs to be trained at the transmitting end or receiving end. The target beam tracking model is used to track millimeter-wave beams; The sample generation module includes: a first historical differential vector acquisition unit, a flip flag value determination unit, a second historical differential vector acquisition unit, a historical millimeter-wave beam differential vector determination unit, and a sample generation unit. The first historical differential vector acquisition unit is used to perform differential operations on the historical horizontal beam index sequence and the historical vertical beam index sequence in the historical millimeter-wave beam index sequence via a transmitter or receiver to obtain a first historical horizontal beam differential vector and a first historical vertical beam differential vector. The flip flag value determination unit is used to determine the horizontal flip flag value based on the first non-zero element of the first historical horizontal beam differential vector, and to determine the value based on the first historical vertical beam differential vector. The system comprises: a first non-zero element to determine the vertical flip flag value; a second historical differential vector acquisition unit, used to update the first historical horizontal beam differential vector according to the horizontal flip flag value to obtain the second historical horizontal beam differential vector, and updating the first historical vertical beam differential vector according to the vertical flip flag value to obtain the second historical vertical beam differential vector; a historical millimeter-wave beam differential vector determination unit, used to determine the historical millimeter-wave beam differential vector according to the second historical horizontal beam differential vector and the second historical vertical beam differential vector; and a sample generation unit, used to generate the millimeter-wave beam tracking sample according to the historical millimeter-wave beam differential vector and the current beam tracking model.

8. A beam tracking device, characterized in that, include: The differential vector acquisition module is used to acquire the current millimeter-wave beam differential vector; The target beam search range determination module is used to analyze the current millimeter-wave beam differential vector through the target beam tracking model to determine the target beam search range that matches the current millimeter-wave beam differential vector. The signal transmission and measurement value acquisition module is used to transmit a signal through the transmitting end according to the target beam search range, and acquire the measurement value of the received beam fed back by the receiving end. The target tracking beam determination module is used to determine the target tracking beam based on the measurement value of the received beam fed back by the receiver. The target beam tracking model is a model trained using any one of the model training methods described in claims 1-5.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the model training method of any one of claims 1-5, or to perform the beam tracking method of claim 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the model training method of any one of claims 1-5, or to execute the beam tracking method of claim 6.

Images