WIFI wave beam fast training method and device

Abstract

The invention provides a WIFI beam fast training method and device, and the method comprises the steps: determining an optimal sector through STA feedback position information and received signal strength; performing beam training preprocessing to obtain a wide transmitting beam; determining an optimal wide beam pair through a wide beam training process assisted by sensing information; based on the optimal wide beam, determining an optimal narrow beam pair through a narrow beam training process assisted by sensing information; in the beam training process, the sensing process is to obtain the position and shape sensing information of the STA by using a pre-trained convolutional neural network (CNN) model. According to the WIFI beam fast training method and device provided by the invention, perceptual information of the WIFI communication equipment can be utilized during communication beam training, and targeted training can be performed in the direction in which the optimal communication beam possibly exists, so that not only can the overhead of beam training be reduced, but also the equipment access time can be shortened, the communication delay can be reduced, and the communication quality can be improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and in particular, to a method and device for fast training of WIFI millimeter wave beams. Background Art

[0002] As a short-range wireless communication technology, WIFI is one of the key technologies commonly used in the Internet of Things, and is particularly suitable for fields such as smart homes and smart cities. Millimeter wave WIFI is mainly based on the IEEE 802.11ad and 802.11ay standard specifications. Millimeter wave chips developed in accordance with the above standard specifications can provide high-speed wireless transmission and are applicable to applications such as high-definition video, AR / VR, and industrial automation. With the popularization of WIFI devices and the ubiquity of WIFI networks, when WIFI signals propagate in physical space, wave phenomena such as reflection, diffraction, and scattering will occur.

[0003] The 60GHz millimeter wave band has rich available spectrum resources. However, due to the increase in path loss and very serious attenuation, a directional communication technology using beamforming (BF) is introduced to provide additional transmit antenna gain or receive antenna gain to overcome signal attenuation. When using the beamforming technology, beamforming training (BFT) must be carried out first. The process of BFT specifically refers to the process of aligning the transmit and receive beams between an access point (AP) and a station (STA). In the existing IEEE 802.11ay standard, the communication beam training mainly finds the best communication beam pair by continuously trying and making mistakes in each direction according to criteria such as the maximum received signal strength at the signal transmitter and receiver. This method is a traversal algorithm, which requires all beams of the STA and the AP to be tried once. The STA and the AP find the best transmit beam and the best receive beam, which has a certain degree of blindness, and this process is a traversal process, which takes a long time and brings a large communication delay. Summary of the Invention

[0004] The present invention provides a method and device for fast training of WIFI beams to solve the technical problem of the long time consumption and large communication delay caused by the WIFI beam training method in the existing IEEE802.11ad / ay standard.

[0005] In a first aspect, the present invention provides a method for fast training of WIFI millimeter wave beams, including: Determining the best sector by the STA feedback position information and received signal strength; Performing beam training preprocessing to obtain a wide transmit beam; Determine the optimal wide beam pair through a wide beam training process assisted by sensing information; based on the optimal wide beam, determine the optimal narrow beam pair through a narrow beam training process assisted by sensing information; during the beam training process, use a trained convolutional neural network (CNN) model to obtain the sensing information of the STA, including the STA position and shape.

[0006] In some embodiments, determining the optimal sector by the STA feedback of position information and received signal strength includes: In the sector scanning stage when the initiator sends, obtain the beam request message sent by the STA. The beam request message carries position information and the signal strength of the signal that can receive the receiver's signal, and sends beams at the beam selection interval; In the sector scanning stage when the responder sends, send a beam response message to the STA. The beam response message carries position information and the signal strength of the signal that can receive the receiver's signal, and sends beams at the beam selection interval; Determine the optimal transmission sector and the optimal reception sector according to the position information and the signal strength.

[0007] In some embodiments, determining the optimal wide beam pair through the beam training process assisted by sensing information includes: adopting a beam training process assisted by sensing information during the wide beam training process to obtain the optimal wide beam pair; based on the optimal wide beam, determining the optimal narrow beam pair through the beam training process assisted by sensing information includes: within the coverage area of the optimal wide beam, adjust the beam width to a narrow beam, and adopt a beam training process assisted by sensing information during the narrow beam training process to obtain the optimal wide beam pair.

[0008] In some embodiments, the frame format in the sensing process includes those for representing request and response messages.

[0009] In some embodiments, the beam training process assisted by sensing information includes: Step 1: The station STA sends a beam training request assisted by sensing information to the access point AP. After receiving the request, the AP performs the STA sensing process to obtain the STA sensing information, and adjusts the beam direction based on the STA sensing information and the strength of the received signal; the AP calculates the transmission / reception beam pair with stronger beam signals according to the CSI magnitude, the shape and position of the STA, and the beam width, and obtains a list of optimal transmission / reception communication beam pairs; Step 2: The AP sends a beam training response assisted by sensing information to the STA, carrying the list of optimal reception communication beam pairs calculated by the AP based on the sensing information assistance; Step 3: The AP sequentially sends training frames to the STA in the optimal transmission beam direction, and the STA then receives the training frame in the optimal reception communication beam list received in Step 2, and records the signal reception strength corresponding to the beam pair; Step 4: The STA sends a feedback frame to the AP, feeding back information such as the best transmission beam obtained in the beam training in Step 3 and the CSI of the STA. Step 5: The STA and the AP complete the fast communication beam training assisted by sensing information and determine the best beam pair.

[0010] In some embodiments, the sensing information includes the location information and shape information of the STA, which are obtained through the STA sensing process; the STA sensing process is a process of performing online inference based on a pre-trained CNN network model; more specifically, the input data of the CNN model inference process is the real-time channel matrix and CSI data reported by the STA, as well as the type and model of the STA reported during the initial registration of the STA; the output of the CNN model inference process is the STA sensing information: the location information and shape information of the STA. The pre-trained CNN model is obtained by an offline training method, and the offline training includes: Construct a training data set according to the historical data collected by the AP, including the channel matrix reported by the STA, the type of the STA, the model of the STA, and the CSI data fed back by the STA; input the training data in the training data set into the CNN model, complete the training of the CNN model, and obtain an available CNN model.

[0011] In a second aspect, the present invention further provides a WIFI beam fast training device, including: A sector scanning module, configured to determine the best sector by feeding back the location information and received signal strength by the STA. A preprocessing module, configured to perform beam training preprocessing to obtain a wide transmission beam. A wide beam scanning module, configured to determine the best wide beam pair through a wide beam training process assisted by sensing information. A narrow beam scanning module, configured to determine the best narrow beam pair based on the best wide beam through a narrow beam training process assisted by sensing information.

[0012] A sensing module, configured to obtain the sensing information of the STA, including the STA location and shape, by using a trained convolutional neural network (CNN) model during the beam training process.

[0013] In a third aspect, the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the WIFI beam fast training method as described in any one of the above when executing the computer program.

[0014] Fourthly, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the WIFI beam rapid training method as described in any one of the above.

[0015] Fifthly, the present invention further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the WIFI beam rapid training method as described in any one of the above.

[0016] The WIFI beam rapid training method and device provided by the present invention can utilize the sensing information of WIFI communication devices during communication beam training, and perform targeted training in the direction where the best communication beam may exist, which can not only reduce the overhead of beam training, but also reduce the device access time, reduce communication latency, and improve communication quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0018] Figure 1 is a schematic flowchart of the WIFI beam rapid training method provided by the present invention.

[0019] Figure 2 is a schematic diagram of the sector-level beam scanning process provided by the present invention.

[0020] Figure 3 is a schematic diagram in which the beacon period in the 802.11ad / ay protocol is divided into multiple time periods.

[0021] Figure 4 is a schematic diagram of the I-TXSS stage provided by the present invention.

[0022] Figure 5 is a schematic diagram of the R-TXSS stage provided by the present invention.

[0023] Figure 6 is a schematic diagram of the SSW Feedback stage provided by the present invention.

[0024] Figure 7 is a schematic diagram of the SSW ACK stage provided by the present invention.

[0025] Figure 8 is a schematic flowchart of the beam training process assisted by sensing information provided by the present invention.

[0026] Figure 9 It is a schematic diagram of the process for offline training of the CNN model provided by the present invention.

[0027] Figure 10 It is a schematic diagram of the CNN model provided by the present invention.

[0028] Figure 11 It is a schematic diagram of the frame format with sensing information provided by the present invention.

[0029] Figure 12 It is a schematic diagram of the structure of the WIFI beam rapid training device provided by the present invention.

[0030] Figure 13 It is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed implementation manners

[0031] STA devices in the WIFI communication network are diverse and have various shapes. For example, the STA device can be a mobile phone, a sweeping robot, a laptop computer, etc. The communication node AP can obtain the size, shape, and contour of the STA device through sensing. When performing communication beam training, these sensing information can be utilized to perform targeted training in the direction where the best communication beam may exist, which can not only reduce the overhead of beam training, but also reduce the device access time, reduce the communication delay, and improve the communication quality. It can assist the AP in narrowing the beam training range. The present application proposes a rapid beam training method combining communication and sensing. The WIFI device detects and senses the WIFI communication environment and communication nodes through wireless sensing technology, such as sensing information of the communication node position, moving speed, etc. The sensing information and sensing results can be used to assist algorithm selection, algorithm parameter setting, and algorithm optimization of modules such as channel estimation, equalization, and beam management at the receiving end and sending end of the communication system.

[0032] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0033] Figure 1 It is a schematic diagram of the process of the WIFI beam rapid training method provided by the present invention. As Figure 1 shown, the method includes the following: Step 101: Determine the optimal sector by means of the STA feedback of the position information and the received signal strength.

[0034] Specifically, the process of beamforming (BFT) is standardized in the 802.11 protocol. The process includes sector-level sweep (SLS) and beam refinement protocol (BRP). Figure 2 It is a schematic diagram of the sector-level beam scanning process provided by the present invention. As Figure 2 shown, the SLS process can occur in the "Beacon Transmission Interval (BTI) + Adaptive Beamforming Training (A-BFT)" phase and the Data Transfer Interval (DTI) phase. The BRP process can only occur in the DTI phase.

[0035] The SLS process is a process of interaction between the Initiator and the Responder, usually including four sub-phases, namely Initiator Sector Sweep (ISS), Responder Sector Sweep (RSS), Sector Sweep Feedback (SSW-Feedback, SSW-FBCK), and Sector Sweep ACK (SSW-ACK). The SLS phase only trains the transmitting sectors of the Initiator and the Responder, and is mandatory. BRP is used to train the receiving sectors and perform beam pairing. In the existing protocol, a traversal algorithm is adopted in the BRP process to achieve beam pairing. The technical solution of this application optimizes the existing method. In the sector scanning phase, sectors are selected according to the STA feedback position information and the received signal strength. Figure 3 It is a schematic diagram in which the beacon interval in the 802.11ad / ay protocol is divided into multiple time intervals. As Figure 3As shown in the figure, a beacon period can be divided into a Beacon Header Interval (BHI) and a Data Transfer Interval (DTI). The BHI includes a Beacon Transmission Interval (BTI), an Association Beamforming Training (A-BFT), and an Announcement Transmission Interval (ATI). The DTI contains a Scheduling Period (SP) and a Contention-Based Access Period (CBAP). The Sector-Level Sector Scan (SLS) training phase mainly completes the transmission sector training of the initiator and the responder. The training process includes four sub-phases: Initiator-Transmit Sector Sweep (I-TXSS), Responder Transmit Sector Sweep (R-TXSS), Sector Sweep Feedback (SSW-Feedback), and Sector Sweep ACK (SSW-ACK). In the existing protocol, the beam training process is executed according to the traversal method. This application proposes a secondary beam training method for collaborative sensing information.

[0036] Figure 4 It is a schematic diagram of the I-TXSS stage provided by the present invention. As Figure 4 shown, in the I-TXSS stage: the initiator sends beacon frames or dedicated training frames in each sector to discover the responder. At this time, the responder needs to receive in a quasi-omnidirectional mode. This stage is used to train the transmission beam of the sender and generally occurs during the BTI. Different from the method of the sender in the existing protocol sending messages in sequence, this application proposes that the beam request message in the training process sent by the sender carries position information and the signal strength capable of receiving the signal of the receiver, and the beam in the transmission beam selection interval is selected.

[0037] Figure 5 It is a schematic diagram of the R-TXSS stage provided by the present invention. As Figure 5As shown in the figure, in the R-TXSS phase: The responder sends SSW frames on each of its sectors, and carries the best transmission sector information of the initiator in the previous phase in the SSW frames. The initiator receives in an omnidirectional mode at this time. This phase is used to train the transmission sectors of the responder and generally occurs during A-BFT. The beam request message in the training process sent by the responder carries position information and the signal strength at which the signal of the receiver can be received, and sends beams at the beam selection interval.

[0038] The sender and the receiver determine a smaller range according to the position information and the signal strength and repeat the steps of the I-TXSS phase and the R-TXSS phase, so that the best transmission sector and the best reception sector can be quickly determined.

[0039] Figure 6 is a schematic diagram of the SSW Feedback phase provided by the present invention. As Figure 6 shown, in the SSW-Feedback phase: The initiator sends an SSW-Feedback frame on the best sector specified in the previous step. The frame carries a list of the responder's transmission sectors sorted by reception quality and the best sector of the responder in the previous phase; when the sender is a STA, the SSW-Feedback frame carries the channel state information (Channel State Information, CSI) and the channel matrix of each sector in the sector list. The AP receives and saves the CSI information and the channel matrix; the responder receives in an omnidirectional mode at this time. This phase generally occurs during A-BFT.

[0040] Figure 7 is a schematic diagram of the SSW ACK phase provided by the present invention. As Figure 7 shown, in the SSW-ACK phase: The responder sends an SSW-ACK frame on the best sector specified in the previous step. The frame carries a list of the initiator's transmission sectors sorted by quality. The initiator should receive in an omnidirectional mode at this time. The training of the SLS phase is completed at this step. SLS is generally carried out in BHI, and the subsequent BRP training is generally carried out in SP.

[0041] Step 102: Perform beam training preprocessing to obtain a wide transmission beam.

[0042] Specifically, after selecting the sector, beam-level training is performed. The beam-level training is divided into two stages: wide beam training and narrow beam training. A preprocessing process is added during the process. The beam width is adjusted by controlling the antenna switch, so that the first-stage beam training process is performed on a wide beam, thereby selecting a wider beam alignment.

[0043] The sender and the receiver use antenna switch control and power control technologies to turn off the antennas outside the optimal sector, rendering most of the antennas in an inactive state, which is equivalent to only using a small number of antenna elements in the antenna array to transmit training beams. Since the number of antennas participating in the operation decreases, with other conditions remaining unchanged, the width of the transmitted beam will broaden, and the coverage range of its main lobe will become wider. At this time, only a small number of training beams are required to search and obtain the approximate interval where the target user is located.

[0044] Step 103: Determine the optimal wide beam through beam-level beam training and scanning.

[0045] Specifically, the perception information-assisted communication beam training proposed by the present invention is adopted under the existing IEEE 802.11ad / ay standard system.

[0046] Figure 8 It is a schematic diagram of the perception information-assisted beam training process provided by the present invention. As Figure 8 shown, it includes the following steps: Step 1: The station STA sends a perception information-assisted beam training request to the access point AP. After receiving the request, the AP performs the STA perception process to obtain the STA perception information, and then adjusts the beam direction based on the perception information. The AP calculates the transmission / reception beam pairs with stronger beam signals according to the CSI magnitude, the shape and position of the STA, and the beam width, and obtains a list of optimal transmission / reception communication beam pairs.

[0047] The STA perception process is as follows; The AP uses a Convolutional Neural Networks (CNN) model to perceive the shape and location of STA devices. The STA perception process is a process of performing online inference based on a pre-trained CNN network model. The input data for the CNN model inference process includes the real-time channel matrix reported by the STA during the SSW-Feedback phase, the CSI data feedback by the STA, as well as the type and model of the STA reported during the STA's initial registration; the output of the CNN model inference process is the STA perception information: the location information and shape information of the STA. The training of the CNN model is divided into two phases: offline training and online inference according to the generation and usage process of the CNN model. In the offline training phase, an offline training method is used to obtain a pre-trained CNN model. During the offline training process, the input data dataset required for the CNN network training for STA device shape and location information perception is generated based on the historical data collected by the AP. The input data of the training dataset of the CNN model includes, but is not limited to, the channel matrix reported by the STA device, the CSI data feedback by the STA device, the type of the STA device, and the model of the STA device; the output data of the training dataset includes the shape of the STA device and the location of the STA device; then the CNN model training set dataset is used for the offline training of the model, and finally the parameters of the trained CNN network model are saved; in the online inference phase, the type and model of the STA device are reported to the AP during the initial connection registration and saved in the AP. The STA device reports the real-time channel matrix and CSI data. The AP processes the received input data and transmits it to the trained CNN model, and can directly output the perception results: the shape of the STA device and the location of the STA device.

[0048] Since the shapes of STA devices of the same type and model are the same, the quantization of the STA device shape is independently measured by the sum of the maximum distances in the x, y, and z directions of the device, and the location of the STA device is measured by a binary tuple of longitude and latitude information.

[0049] Step 2: The AP sends a beam training response assisted by perception information to the STA, carrying a list of the best receive communication beam pairs calculated by the AP based on the perception information.

[0050] Step 3: The AP sequentially sends training frames to the STA directionally according to the best transmit beam determined in the previous step. The STA then receives the training frame directionally according to the list of the best receive communication beams received in Step 2 and records the signal reception strength corresponding to the beam pair.

[0051] Step 4: The STA sends a feedback frame to the AP, feeding back the best transmit beam obtained in the beam training in Step 3 and the CSI information of the STA.

[0052] Step 5: The STA and AP complete the perception-based fast communication beam training and determine the optimal beam.

[0053] In some embodiments, the method further includes: Construct a training dataset based on the channel matrix reported by the STA, the CSI data reported by the STA device, the type of the STA device, and the model of the STA device; Input the training data in the training dataset into the CNN model to complete the training of the CNN model.

[0054] Process of offline training the CNN model: The offline training process refers to training the CNN model using large-scale data in a non-real-time environment, and then deploying the trained model to the AP for inference after completion. Figure 9 It is a schematic diagram of the offline training process of the CNN model provided by the present invention. As Figure 9 shown, the offline training process includes several steps such as data preparation and preprocessing, forward propagation, loss calculation, backpropagation, and model optimization and update to obtain a trained CNN model.

[0055] Data preparation and preprocessing: In this stage, data acquisition, data augmentation, data normalization, and data partitioning required for model training are completed. The specific process is as follows: Data acquisition: Obtain training sample data from the database formed by historical data. The data in this embodiment includes, but is not limited to, the channel matrix, the CSI data reported by the STA device, the type of the STA device, and the model of the STA device, and an error threshold T is set.

[0056] Data augmentation: Augment the acquired sample data by methods such as random cropping, rotation, flipping, and noise addition to increase data diversity and prevent overfitting during model training.

[0057] Data normalization: Standardize the augmented data. In this embodiment, normalization processing is adopted to normalize the data to the [0,1] interval to accelerate the convergence speed.

[0058] Data partitioning: Divide the normalized data into a training set (about 70 - 80%) and a test set (about 20 - 30%); where the training set data is used as the input data for training the CNN model, and the test set is used to evaluate the accuracy of the model and model tuning; and a set of data is taken as the label data.

[0059] Forward propagation: Determine the CNN model and input the training set data into the model.

[0060] Figure 10It is a schematic diagram of the CNN model provided by the present invention. As Figure 10 shown, the CNN model adopted in this embodiment is composed of an input layer (Input Layer), a convolutional layer (Convolutional Layer), a pooling layer (Pooling Layer), a fully connected layer (Fully Connected Layer), and an output layer (Output Layer). The forward propagation process is as follows: The input layer inputs the model input data obtained; in the convolutional layer, local features are extracted, multiple filters are applied for feature mapping. In this embodiment, a 3×3 convolutional kernel is adopted for feature extraction; the activation function adopts the ReLU (Rectified Linear Unit) function: f(x) = max(0, x) to increase the non-linear characteristics; the pooling layer uses average pooling for feature dimensionality reduction to improve the calculation efficiency; the fully connected layer (FC) maps the extracted high-level features to the output features of the final perception task; the output layer outputs the output value finally obtained from the model training.

[0061] Loss calculation: The test set is used as the input data and input into the trained CNN model to obtain the shape of the STA device and the position of the STA device output during training. The output data is compared with the real value data used as the label to calculate the error; the shape error of the device is the difference between the perceived device shape metric and the real value; the position error of the device is the sum of the perceived device position error and the difference in longitude and latitude of the real device position; when the sum of the errors is greater than the preset threshold T, the model input is adjusted and retrained until the error is less than the threshold.

[0062] Backpropagation: The chain rule is used to calculate the gradient of the error with respect to the weight parameters of each layer of the CNN model, and the gradient is backpropagated to each layer (convolutional layer, fully connected layer).

[0063] Model optimization and update: Update the training data and model parameters (such as weights, biases) and retrain.

[0064] Figure 11 It is a schematic diagram of the frame format with sensing information provided by the present invention. As Figure 11 shown, in the sensing process of this application, the frame format in the protocol is modified, and bits are added to the frame format to represent request and response messages.

[0065] Step 104: Based on the optimal wide beam, determine the optimal narrow beam pair through the narrow beam training process assisted by sensing information.

[0066] Specifically, since some STAs may be located between the coverage ranges of two wide beams or narrow beams, in order to provide the best beam direction for all STAs, perceptual information training is added during the beam training process. Information such as the volume and position of the STA is obtained through perceptual information training. After the AP determines the wide beam range with stronger signals, the perceptual information-assisted narrow beam beam training shown in Figure 8 is performed on several wide beams in the coarsely grained space selected according to the wide beam perceptual information. Further, the narrow beam width is adjusted according to the more accurate position and volume information of the perceived STA, and the beam direction is selected by adjusting the codebook used for narrow beam transmission. The optimal narrow beam pairing is determined through the narrow beam training process. Figure 8 During the narrow beam training process, the AP obtains the channel matrix reported by the STA, the type of the STA, the model of the STA, and the CSI, and inputs the channel matrix, the type of the STA, the model of the STA, and the CSI into the CNN network model. The CSI information fed back by the STA under the condition of the coarse beam transmission signal is obtained by using the convolutional kernels of different scales of the CNN network model. In addition, the accurate position information and volume information of the STA can be perceived. Through the perceptual information, the AP dynamically adjusts the width of the narrow beam and obtains the narrow beam pair with the best transmitted and received signals.

[0067] Step 105: During the beam training process, the trained convolutional neural network (CNN) model is used to obtain the perceptual information of the STA, including the position and shape of the STA.

[0068] WIFI signals rely on wireless channels for transmission. CSI is a measure of the attributes of the wireless channel, which records the changes in the wireless channel under the influence of various factors such as distance attenuation, environmental attenuation, and signal scattering. WIFI systems based on the 802.11n and above standards incorporate MIMO and OFDM technologies, that is, both the transmitter and the receiver have multiple antennas, and the OFDM technology is used to modulate the WIFI signal. At this time, the CSI measured at the receiver is a complex matrix of size

[0069] where represents the number of transmit antennas, represents the number of receive antennas, and represents the number of subcarriers.

[0070]

[0070] In a wireless transmission environment, the transmission of WIFI signals is affected by obstacle factors, resulting in phenomena such as reflection, scattering, refraction, and diffraction. As a result, the receiving end receives WIFI signals on multiple paths in different directions, which is called the multipath effect. A typical WIFI communication system consists of a WIFI signal transmitter (such as a wireless router), a receiving end (such as devices like mobile phones and computers), and users. The WIFI signals propagate in a multipath manner, including the line-of-sight path (LoS) from the transmitter to the receiver, and paths reflected by stationary objects such as walls. Such paths are called static paths. When a person in the WIFI environment performs a gesture action, it will cause dynamic changes in a part of the propagation paths of the WIFI signals, and these paths are called dynamic paths. Due to the influence of the multipath effect, the receiving end receives signals on multiple paths. Among them, the signals on each path are affected by different channel attenuations, phase delays, and scattering, etc. The receiving end can obtain the CSI information of the signals through channel estimation technology.

[0071] When the volume of the obstacle is greater than a certain threshold, different positions of the obstacle have different effects on the propagation path of WIFI. Especially when the obstacle moves or its shape changes, different positions will also have different effects on the propagation of WIFI. Correspondingly, the receiving end will collect different CSI information. The characteristic information related to the volume and position of the obstacle can be extracted from the CSI information according to the correlation between the shape, position of the obstacle and the CSI, and then input into the machine learning algorithm to realize the perception of the position and volume information of the obstacle.

[0072] This application proposes to introduce the perception information of the STA in the beam training and beam tracking processes. The perception information can perceive information such as the position and shape of an object, so as to narrow the beam range in the beam training and beam tracking processes; furthermore, this application proposes a multi-level scanning beam training scheme. In the existing protocol, the beam training process specification proposes two-level scanning at the sector level and the beam level. This application proposes a three-level beam scanning process, including sector scanning, coarse-grained beam scanning, and fine-grained beam scanning; among them, the sector scanning follows the sector-level scanning of the existing technology. The coarse-grained beam level scanning process is a beam-level scanning, which uses wide beam training to find the best transmit beam and receive wide beam in a short time, and then uses narrow beam training within the coverage range of the wide beam. This consumes less time than directly using narrow beam training in a large range, thus reducing the communication delay.

[0073] The WIFI beam fast training device provided by the present invention will be described below. The WIFI beam fast training device described below can be correspondingly referred to the WIFI beam fast training method described above.

[0074] Figure 12It is a schematic structural diagram of the WIFI beam rapid training device provided by the present invention. As Figure 12 shown, the present invention provides a WIFI beam rapid training device, including: The sector scanning module 1001 is used to determine the optimal sector by means of the STA feedback position information and the received signal strength. The preprocessing module 1002 is used to perform beam training preprocessing to obtain a wide transmission beam. The wide beam scanning module 1003 is used to determine the optimal wide beam pair through a wide beam training process assisted by sensing information. The narrow beam scanning module 1004 is used to determine the optimal narrow beam pair based on the optimal wide beam through a narrow beam training process assisted by sensing information.

[0075] The sensing module 1005 is used to obtain the sensing information of the STA, including the STA position and shape, by using a trained convolutional neural network CNN model during the beam training process.

[0076] In some embodiments, the determining the optimal sector by means of the STA feedback position information and the received signal strength includes: In the sector scanning stage sent by the initiator, obtain the beam request message sent by the STA. The beam request message carries the position information and the signal strength capable of receiving the signal of the receiver, and send the beam at the beam selection interval. In the sector scanning stage sent by the responder, send a beam response message to the STA. The beam response message carries the position information and the signal strength capable of receiving the signal of the receiver, and send the beam at the beam selection interval. Determine the optimal transmitting sector and the optimal receiving sector according to the position information and the signal strength.

[0077] In some embodiments, the determining the optimal wide beam pair through the beam training process assisted by sensing information includes: adopting a beam training process assisted by sensing information during the wide beam training process to obtain the optimal wide beam pair; the determining the optimal narrow beam pair based on the optimal wide beam through the beam training process assisted by sensing information includes: within the coverage area of the optimal wide beam, adjust the beam width to a narrow beam, and adopt a beam training process assisted by sensing information during the narrow beam training process to obtain the optimal wide beam pair.

[0078] In some embodiments, the frame format in the sensing process includes messages for indicating requests and responses.

[0079] In some embodiments, the beam training process assisted by sensing information includes: Step 1: The station STA sends a sensing information-assisted beam training request to the access point AP. After receiving the request, the AP performs the STA sensing process to obtain STA sensing information, and adjusts the beam direction based on the STA sensing information and the strength of the received signal; the AP calculates the transmit / receive beam pair with stronger beam signals according to the CSI magnitude, the shape and position of the STA, and the beam width, and obtains a list of the best transmit / receive communication beam pairs; Step 2: The AP sends a sensing information-assisted beam training response to the STA, carrying the list of the best receive communication beam pairs calculated by the AP based on the sensing information assistance; Step 3: The AP sequentially sends training frames to the STA in the best transmit beam direction. The STA then receives the training frame in the best receive communication beam list received in Step 2, and records the signal reception strength corresponding to the beam pair; Step 4: The STA sends a feedback frame to the AP, feeding back the best transmit beam obtained in the beam training in Step 3 and information such as the CSI of the STA; Step 5: The STA and the AP complete the sensing information-assisted fast communication beam training and determine the best beam pair.

[0080] In some embodiments, the sensing information includes the position information and shape information of the STA, which are obtained through the STA sensing process; the STA sensing process is a process of performing online inference based on a pre-trained CNN network model; more specifically, the input data of the CNN model inference process is the real-time channel matrix and CSI data reported by the STA, as well as the type and model of the STA reported during the initial registration of the STA; the output of the CNN model inference process is the STA sensing information: the position information and shape information of the STA; The pre-trained CNN model is obtained by an offline training method, and the offline training includes: Construct a training data set according to the historical data collected by the AP, including the channel matrix reported by the STA, the type of the STA, the model of the STA, and the CSI data feedback by the STA; input the training data in the training data set into the CNN model, complete the training of the CNN model, and obtain an available CNN model.

[0081] Specifically, the above WIFI beam fast training device provided by the embodiments of the present application can implement all the method steps implemented by the above WIFI beam fast training method embodiments, and can achieve the same technical effects. The same parts and beneficial effects as those in the method embodiments in this embodiment will not be specifically described here.

[0082] Figure 13 Illustrates a schematic physical structure diagram of an electronic device, such as Figure 13As shown in the figure, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete communication with each other through the communication bus 840. The processor 810 may call logic instructions in the memory 830 to execute the WIFI beam rapid training method, and this method includes: Determine the optimal sector by the STA feedback of position information and received signal strength; Perform beam training preprocessing to obtain a wide transmission beam; Determine the optimal wide beam pair through a wide beam training process assisted by sensing information; Based on the optimal wide beam, determine the optimal narrow beam pair through a narrow beam training process assisted by sensing information; During the beam training process, use the trained convolutional neural network CNN model to obtain the sensing information of the STA, including the STA position and shape.

[0083] In addition, when the logic instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0084] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the WIFI beam rapid training method provided by the above-mentioned various methods. This method includes: Determine the optimal sector by the STA feedback of position information and received signal strength; Perform beam training preprocessing to obtain a wide transmission beam; Determine the optimal wide beam pair through a wide beam training process assisted by sensing information; Based on the optimal wide beam, determine the optimal narrow beam pair through a sensing information-assisted narrow beam training process; During the beam training process, use the trained convolutional neural network (CNN) model to obtain the sensing information of the STA, including the STA position and shape.

[0085] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the WIFI beam fast training method provided by the above-mentioned various methods. The method includes: Determine the optimal sector by the STA feedback of the position information and the received signal strength; Perform beam training preprocessing to obtain a wide transmission beam; Determine the optimal wide beam pair through a sensing information-assisted wide beam training process; Based on the optimal wide beam, determine the optimal narrow beam pair through a sensing information-assisted narrow beam training process; During the beam training process, use the trained convolutional neural network (CNN) model to obtain the sensing information of the STA, including the STA position and shape.

[0086] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0087] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A WIFI beam fast training method, characterized in that: include: Determine the best sector through STA feedback of location information and received signal strength; Perform beam training preprocessing to obtain a wide transmit beam; Determine the best wide beam pair through a wide beam training process assisted by perceptual information; Based on the optimal wide beam, determining an optimal narrow beam pair through a narrow beam training process assisted by perceptual information; During the beam training process, the trained convolutional neural network (CNN) model is used to obtain the STA perception information, including the STA position and shape.

2. The WIFI beam fast training method according to claim 1, characterized in that: The determining the best sector by feedback of the position information and the received signal strength by the STA includes: In the sector scanning phase sent by the initiator, the beam request message sent by the STA is obtained. The beam request message carries the location information and the signal strength of the receiving party's signal, and the beam of the sending beam selection interval; During the sector scanning phase, the responder sends a beam response message to the STA. The beam response message carries the location information and the signal strength of the receiving party's signal, and sends the beam of the beam selection interval. The best transmitting sector and the best receiving sector are determined based on the location information and signal strength.

3. The WIFI beam fast training method according to claim 1, characterized in that: The method of determining the optimal wide beam pair through the beam training process assisted by perceptual information includes: adopting the beam training process assisted by perceptual information in the wide beam training process to obtain the optimal wide beam pair; the method of determining the optimal narrow beam pair through the beam training process assisted by perceptual information based on the optimal wide beam includes: adjusting the beam width to a narrow beam within the coverage area of ​​the optimal wide beam, and adopting the beam training process assisted by perceptual information in the narrow beam training process to obtain the optimal wide beam pair.

4. The WIFI beam fast training method according to claim 1, characterized in that: The frame format used in the sensing process includes messages for representing request and response.

5. The WIFI beam fast training method according to claim 3, characterized in that: The perception information-assisted beam training process includes: Step 1: The station STA sends a beam training request assisted by perception information to the access point AP. After receiving the request, the AP performs the STA perception process to obtain the STA perception information, and adjusts the beam direction based on the STA perception information and the strength of the received signal; the AP calculates the transmit / receive beam pair with a stronger beam signal according to the CSI size, the shape and position of the STA, and the beam width, and obtains the best transmit / receive communication beam pair list; Step 2: The AP sends a beam training response assisted by sensing information to the STA, carrying a list of the best receiving communication beam pairs calculated by the AP based on the sensing information; Step 3: The AP sends training frames to the STA in a directional manner according to the best transmission beam. The STA receives the training frames in a directional manner according to the best receiving communication beam list received in step 2, and records the signal receiving strength corresponding to the beam pair. Step 4: The STA sends a feedback frame to the AP, feeding back the best transmit beam obtained in the beam training in step 3 and the STA's CSI and other information; Step 5: STA and AP complete fast communication beam training assisted by perception information and determine the best beam pair.

6. The WIFI beam fast training method according to claim 5, characterized in that: The perception information includes the position information and shape information of the STA, which is obtained through the STA perception process; the STA perception process is a process of performing online reasoning based on a pre-trained CNN network model; More specifically, the input data of the CNN model reasoning process is the real-time channel matrix and CSI data reported by the STA, as well as the STA type and STA model reported during the initial registration of the STA; the CNN model reasoning process outputs the STA perception information: the location information and shape information of the STA; The pre-trained CNN model is obtained by an offline training method, and the offline training includes: Build a training data set based on historical data collected by the AP, including the channel matrix reported by the STA, the type of the STA, the model of the STA, and the CSI data fed back by the STA; The training data in the training data set is input into the CNN model to complete the training of the CNN model and obtain a usable CNN model.

7. A WIFI beam fast training device, characterized in that: include: The sector scanning module is used to determine the best sector through STA feedback of location information and received signal strength; A preprocessing module is used to perform beam training preprocessing to obtain a wide transmit beam; A wide beam scanning module for determining an optimal wide beam pair through a wide beam training process assisted by perception information; A narrow beam scanning module, configured to determine an optimal narrow beam pair based on the optimal wide beam through a narrow beam training process assisted by perception information; The perception module is used to obtain the perception information of STA, including the position and shape of STA, by using the trained convolutional neural network (CNN) model during the beam training process.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the WIFI beam fast training method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the WIFI beam fast training method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the WIFI beam fast training method according to any one of claims 1 to 6 is implemented.

Images