Particle filter based beam tracking method and system for mmwave scenarios

Abstract

The present application relates to the technical field of millimeter wave communication, and particularly relates to a particle filtering-based beam tracking method and system for a millimeter wave scene, which dynamically determines whether a tracking step needs to be performed by comparing the signal-to-noise ratio fading of beams formed before and after time slots. If tracking is needed, two perturbation angles are generated near the pointing angle of a data beam in the previous time slot to form a measurement beam, and the received signal strength ratio of the measurement beam is taken as the observation value of an adaptive particle filtering algorithm. Angle particle sets are generated in a half-wave width interval centered on the current beam pointing angle, the particle weights are calculated in combination with double non-central F distribution, and then the optimal particle is selected to update the data beam direction. The present application can realize efficient tracking of millimeter wave beams in different mobility level scenes, guarantee communication performance, reduce tracking overhead, and adaptively adjust the overhead according to the user moving speed, which can significantly improve the beam tracking precision, and is particularly suitable for millimeter wave communication in high-speed scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of millimeter wave communication, and particularly relates to a beam tracking method based on particle filtering for a millimeter wave scene (Adaptive Beam Tracking Method based on Particle Filter, ABTPF) and a system. BACKGROUND

[0002] Millimeter wave communication has become a key technology to support ultra-high-speed data communication due to its high bandwidth characteristics, and is an important direction to promote the evolution of mobile communication systems. However, due to the inherent high path loss of millimeter wave signals, communication devices use large-scale multiple-input multiple-output and beamforming technology to obtain directional gain of signals to ensure the coverage of communication. When using beamforming technology, in order to achieve high directional gain and optimize communication performance, the main path angles of the millimeter wave channel, i.e., the angle of arrival (AoA) and the angle of departure (AoD), must be accurately estimated to ensure that the beam direction is aligned with the main channel path direction. The initial establishment of the millimeter wave communication link requires beam alignment, which usually requires searching in the complete angle space. Existing beam initial alignment methods mainly include compressed sensing and spatial scanning. However, due to the large angle search space, initial alignment usually brings large beam training overhead.

[0003] After completing the initial alignment of the beam, the millimeter wave communication transceiver terminal needs to track the main path direction of the channel through dynamic tracking and update the beam direction in real time to ensure high-quality connection of the communication link. For the beam tracking problem, existing technologies have proposed various solutions, but there are still limitations. For example, the codebook design based on hierarchical search arranges codebooks with different resolutions in space in a hierarchical manner to cover the search needs of wide beams and narrow beams. Although this method effectively reduces the search overhead, its tracking accuracy is limited by the codebook resolution, making it difficult to maintain high efficiency in complex dynamic scenarios. For another example, the beam tracking based on reinforcement learning uses the Q-learning method to evaluate the analog beamforming vector in the wireless channel to find the best beam pair that maximizes the signal-to-interference-plus-noise ratio. This method needs to search a large-scale discrete action space, and when the Q table size is too large, it may cause overfitting, thereby affecting the actual application effect. The beam tracking based on supervised learning uses a long short-term memory network for offline training and realizes beam prediction through a large number of pre-labeled sample data, but it is strongly dependent on data and the offline training process brings high time and cost overhead.

[0004] Machine learning based beam tracking methods usually require a large amount of training data to ensure model effectiveness. However, in data-scarce scenarios, traditional filtering methods are introduced to meet the beam tracking demand. For example, the Kalman Filter (KF) method estimates through a linear Gaussian model, but due to its error propagation problem when dealing with nonlinear estimation problems, the tracking error will gradually accumulate over time. To reduce such errors, the Extended Kalman Filter (EKF) linearizes the nonlinear problem to improve the estimation error, but its performance is still limited by the error introduced by linearization. On the other hand, the Standard Particle Filter (SPF) and its variants are widely used for beam tracking, representing the state probability distribution through a particle set, and approximating the posterior distribution of the signal using particle weights. Compared with EKF, PF shows stronger robustness when dealing with nonlinear and non-Gaussian problems. However, the above schemes usually assume that the statistical properties (such as variance) of angle change and path gain change are known, while in actual scenarios these parameters are often difficult to obtain accurately. In addition, the current particle filtering scheme generally ignores the problem of dynamically adjusting the tracking overhead according to the mobility in the scene, which will lead to waste of computing resources. SUMMARY

[0005] To this end, the present application provides a particle filter based beam tracking method and system for millimeter wave scenarios, which measures through a pair of auxiliary beams, calculates the received signal strength ratio based on this measurement information as the observation value of the adaptive particle filter algorithm, estimates the channel main path angle according to the observation value, and establishes a good communication link; judges whether the tracking step needs to be performed through the signal-to-noise ratio fading of the beamformed adjacent time slots, realizes efficient tracking of millimeter wave beams in different mobility level scenarios, and guarantees the communication performance.

[0006] According to the design scheme provided by the present application, on the one hand, a particle filter based beam tracking method for millimeter wave scenarios is provided, which comprises:

[0007] judging whether the difference of the signal-to-noise ratio after beamforming of the current time slot and the previous time slot exceeds the set tracking threshold, if the difference exceeds the tracking threshold, triggering beam tracking;

[0008] adding two specified offset angles in the data beam direction to generate a pair of auxiliary measurement beams, and starting the particle filter algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target and adjust the beam direction;

[0009] In the particle filter algorithm, the received signal strength ratio of the auxiliary measurement beam is calculated as an observation value input into the particle filter, the angle particle set of the current time slot is generated according to the half wave width of the data beam pointing angle of the previous time slot, and the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution. The angle particle with the highest weight is selected as the optimal angle particle of the path angle estimation value, and the data beam direction of the current time slot is updated with the optimal angle particle. The particle weight is used to measure the matching degree of the particle and the observation value.

[0010] As the particle filter-based beam tracking method for the millimeter wave scene, further, for the measurement auxiliary beam, the receiving end in the data beam direction is set as the fixed end and the transmitting end is set as the mobile end, then the data beam pointing angle of the t-1 time slot is plus-minus perturbation angle δ, two auxiliary measurement beams are generated, and the received signals are respectively represented as:

[0011]

[0012] wherein, represents the conjugate transpose; w and f represent the beam forming vectors of the receiving end and the transmitting end respectively, and H t is the channel matrix of the t time slot, and s is the pilot signal, and k represents the pilot length; is the channel path angle of departure; M R and M T represent the number of antennas of the receiving end R and the transmitting end T respectively; z - and z + are additive complex Gaussian noises with mean 0 and variance , and are independent of each other. The received signal contains the angle information of the channel path, and is a key parameter for estimating the main path angle of the channel.

[0013] As the particle filter-based beam tracking method for the millimeter wave scene, further, the process of obtaining the received signal strength ratio of the auxiliary measurement beam includes:

[0014] Based on the received signal, the received signal strength of the two auxiliary measurement beams is obtained, which is respectively represented as: wherein, represents the received signal of the two auxiliary measurement beams respectively;

[0015] The ratio metric is calculated according to the received signal strength, wherein the calculation process of the ratio metric is represented as:

[0016]

[0017] The received signal strength ratio is obtained according to the ratio metric, and the calculation formula of the received signal strength ratio is represented as:

[0018] As the particle filtering-based beam tracking method for the millimeter wave scene, further, the angle particle set of the current time slot is generated according to the half-wave width of the data beam pointing angle of the previous time slot, and the method comprises:

[0019] The angle particle set x of the particle filtering algorithm of the t time slot is set t The N particles are generated, and the generation strategy is represented as:

[0020]

[0021] Wherein, i represents the particle index, The uniform distribution range of the particle is

[0022] As the particle filtering-based beam tracking method for the millimeter wave scene, further, the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution, and the method comprises: setting that the particle is subject to the double non-central F distribution with degrees of freedom υ1 and υ2, and determining the particle weight by using the double non-central F distribution probability density, wherein the calculation process of the particle weight is: for the i th particle, the related double non-central parameters are represented as:

[0023]

[0024] P represents the transmission power, The channel estimation value of the t time slot is calculated by using the i th particle; the weight of the i th particle is calculated by the following formula f F (·) represents the probability density function of the double non-central F distribution.

[0025] As the particle filtering-based beam tracking method for the millimeter wave scene, further, the optimal angle particle serving as the path angle estimation value is selected according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle, and the method comprises:

[0026] The particle weight is normalized by using The particles are resampled according to the normalized particle weight, and in the resampling, the particles with the weight greater than a specified value are reserved according to the weight size and the weight And the new particle set and the weight set Wherein, the new particle set and the weight set are represented as: representing a resampling operation policy function to update the particle set by a resampling operation and a weight set

[0027] With the resampled particle set, an estimated value of a target data beam pointing angle AoA is calculated, and the estimated value calculation formula is represented as: N is the total number of particles.

[0028] As the particle filtering-based beam tracking method for the millimeter wave scene, further, the process of comparing the beamforming post-signal-to-noise ratio of the current time slot with that of the previous time slot comprises:

[0029] obtaining the beamforming post-signal-to-noise ratio γ of the t time slot t and the beamforming post-signal-to-noise ratio γ of the t-1 time slot t-1 if the beamforming post-signal-to-noise ratios of the two time slots satisfy the condition:

[0030] |γ t -γ t-1 |>ι

[0031] then triggering the beam tracking once in the t time slot, wherein the beamforming post-signal-to-noise ratio calculation formula is represented as: P represents the transmission power, and ι is the tracking threshold.

[0032] In another aspect, the application further provides a particle filtering-based beam tracking system for a millimeter wave scene, comprising a judgment module and a tracking module, wherein:

[0033] the judgment module is configured to judge whether the difference between the beamforming post-signal-to-noise ratios of the current time slot and the previous time slot exceeds the set tracking threshold, and if the difference exceeds the tracking threshold, it is determined that the beam tracking is needed in the current time slot;

[0034] the tracking module is configured to add two specified offset angles in the data beam direction to generate a pair of auxiliary measurement beams, and to start the particle filtering algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target and adjust the beam direction.

[0035] In the particle filtering algorithm, the received signal strength ratio of the auxiliary measurement beam is obtained as the observation value of the particle filtering algorithm, the angle particle set of the current time slot is generated according to the half beam width of the data beam pointing angle of the previous time slot, the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution, the optimal angle particle serving as the path angle estimated value is selected according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle, and the particle weight is used to measure the matching degree of the particle and the observation value.

[0036] Advantages of the present application:

[0037] The present application adds two perturbation angles to the data beam pointing angle of the previous time slot to form a measurement beam, and uses the received signal strength ratio as the observation value of the adaptive particle filtering algorithm. A set of angle particles is generated within a half wave width interval, the particle weights are calculated in combination with the double non-central F distribution, and the optimal particle is selected to update the data beam direction. By comparing the signal-to-noise ratios after beamforming in the previous and subsequent time slots, it is determined whether to perform particle filtering, thereby reducing the tracking overhead. By comparing the difference in signal-to-noise ratio after beamforming between consecutive time slots, the tracking mechanism is triggered only when the difference exceeds a preset threshold. This scheme significantly reduces unnecessary tracking overhead, allowing the tracking strategy to adapt to the user's moving speed, while maintaining high-precision angle tracking and low loss rate in high-dynamic scenarios. Further experiments one and two in the embodiments show that the ABTPF scheme can adapt to various mobility communication scenarios and meet the high-quality requirements of millimeter wave communication links under different dynamic conditions. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 Flowchart of particle filtering-based beam tracking for millimeter wave scenarios in embodiments;

[0039] Figure 2 Motion model and measurement model in embodiments;

[0040] Figure 3 Angle tracking results and tracking error for each time in embodiments;

[0041] Figure 4 Influence of different tracking thresholds on tracking error and tracking times of the scheme in embodiments;

[0042] Figure 5 Beamforming signal-to-noise ratio and tracking error of different algorithms under different parameters in embodiments;

[0043] Figure 6 Comparison of tracking error for each time of different algorithms under different motion speeds in embodiments;

[0044] Figure 7 Comparison of tracking times of different algorithms under different motion speeds in embodiments;

[0045] Figure 8 Scenario route in embodiments;

[0046] Figure 9 Beamforming signal-to-noise ratio in the scenario in embodiments;

[0047] Figure 10Fig. 1 shows a schematic diagram of beam tracking result analysis in a scenario route in the embodiment. DETAILED DESCRIPTION

[0048] In order to make the objects, technical solutions and advantages of the present application clearer, more apparent, the present application will be further described in detail below with reference to the drawings and technical solutions.

[0049] In the millimeter wave mobile communication scenario, due to the mobility of the communication terminal and the dynamic change of the communication environment, the main path direction of the communication channel can change rapidly. In order to effectively cope with the rapid change of the channel path direction and reduce the beam training overhead, the communication transceiver terminal usually adopts the method of beam tracking. However, how to design a low overhead and high tracking accuracy beam tracking algorithm has become a technical problem to be solved. The embodiment of the present application, as shown in Figure 1 The present application provides a particle filtering based beam tracking method for millimeter wave scenario, which comprises:

[0050] S101, judging whether the difference of the signal-to-noise ratio after beamforming between the current time slot and the previous time slot exceeds the set tracking threshold value, if the difference exceeds the tracking threshold value, it is determined that the current time slot needs beam tracking;

[0051] S102, adding two specified offset angles in the data beam direction to generate a pair of auxiliary measurement beams, and starting the particle filtering algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target and adjust the beam direction;

[0052] In the particle filtering algorithm, the received signal strength ratio of the auxiliary measurement beam is obtained as the observation value of the particle filtering algorithm, the angle particle set of the current time slot is generated according to the half wave width of the data beam pointing angle of the previous time slot, the angle particle weight of the current time slot is calculated based on the probability density function of double non-central F distribution, the optimal angle particle serving as the path angle estimation value is selected according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle, and the particle weight is used to measure the matching degree of the particle and the observation value.

[0053] By comparing the difference of the signal-to-noise ratio after beamforming between the current time slot and the previous time slot, if the difference exceeds the tracking threshold, the particle filter algorithm is started to estimate the target motion state. In the particle filter algorithm, a pair of auxiliary beams is introduced based on the data beam direction for beam measurement, the auxiliary beams are formed by adding a specific offset angle to the data beam pointing angle at the transceiver end, and the auxiliary beams include two beams with different directions; the received signal strength ratio of the auxiliary beam pair is taken as the observation value of the particle filter algorithm to estimate the channel main path angle; the particle filter algorithm generates a particle set within the half beam width interval of the data beam pointing angle of the previous time slot; the particle weight is calculated based on the probability density function of the double non-central F distribution; the optimal particle is calculated as the estimated value of the channel main path angle according to the particle weight, and the data beam direction is updated.

[0054] As shown in Figure 2 , a pair of auxiliary beams with different directions is formed by adding a predefined offset angle to the direction of the data beam.

[0055] Define M R and M T represent the number of antennas of the base station (receiver end) and the mobile user (transmitter end) respectively. The channel matrix between the receiver end and the transmitter end at time slot t is represented as Define the channel main path angles AoA and AoD as θ and Consider using a uniform linear antenna array at the receiver end and the transmitter end, and the data beam steering vectors of the receiver end and the transmitter end are represented as:

[0056]

[0057] wherein and are the estimated values of θ and respectively. For the convenience of understanding, only the uplink scenario is analyzed, that is, only the AoA is estimated at the receiver end, and the AoD of the transmitter end is known (i.e. ). The downlink scenario analysis is the same as this. A perturbation angle is added to the data beam direction of the receiver end at time slot t-1 to form a pair of auxiliary beams for beam measurement at time slot t, and the auxiliary beams can be represented as:

[0058]

[0059] δ is the perturbation angle size, and is taken as is the half beam width. Then the received signals corresponding to the pair of auxiliary beams at time slot t are represented as:

[0060]

[0061] wherein, represents the conjugate transpose, denotes the pilot signal, z - and z + are additive complex Gaussian noises with mean 0 and variance The received signal contains the angle information of the channel path, which is the key parameter to estimate the main path angle of the channel. The corresponding received signal strength can be expressed as:

[0062]

[0063] The ratio metric is calculated as:

[0064]

[0065] Let have:

[0066]

[0067] The received signal strength ratio ε t is the observation of the ABTPF. The particle set x t of the t-th time slot ABTPF algorithm contains N particles, and the generation strategy is:

[0068]

[0069] i represents the particle index, represents a uniform distribution. The weight of the particle is used to measure the matching degree of the particle and the observation value ε t . The particle is subject to a double non-central F distribution with degrees of freedom υ1 and υ2, and the particle weight is determined by calculating the probability density of the distribution. For the i-th particle, the related double non-central parameters are:

[0070]

[0071] where P represents the received power, represents the t-th time slot using the i-th particle to calculate the channel estimation value, which is expressed as:

[0072]

[0073] where represents the t-th time slot path gain estimation value, which is estimated using the received signal obtained by the data beam of the t-1 time slot

[0074]

[0075] is the channel steering vector estimated using the i-th particle, which is expressed as:

[0076]

[0077] The weight of the ith particle in the tth time slot is calculated by the following formula:

[0078]

[0079] where f F (·) represents the probability density function of the doubly non-central F distribution. The particle weights are normalized:

[0080]

[0081] The particles are resampled according to the normalized weights, and the particles with larger weights are reserved and their weights are copied as a new particle set and a weight set are expressed as:

[0082]

[0083] where represents a resampling operation strategy function for updating the particle set and the weight set Using the resampled particle set, the estimated value of the target AoA is calculated according to the following formula:

[0084]

[0085] In order to adapt to different speed change frequencies of motion scenes and reduce unnecessary tracking overhead, a tracking threshold i is introduced. If the signal-to-noise ratio γ t after beamforming in the tth time slot satisfies: t-1

[0086] |γ t -γ t-1 |>ι

[0087] tracking will be performed. Ignoring the time slot subscript, the expression of the signal-to-noise ratio after beamforming is:

[0088]

[0089] Further, based on the above method, the embodiment of the present application also provides a particle filtering-based beam tracking system for a millimeter wave scene, comprising: a judgment module and a tracking module, wherein,

[0090] The judgment module is used for judging whether the difference between the signal-to-noise ratios after beamforming in the current time slot and the previous time slot exceeds the set tracking threshold. If the difference exceeds the tracking threshold, it is determined that the current time slot needs beam tracking.

[0091] ​A tracking module is configured to add two specified offset angles in the data beam direction to generate a pair of auxiliary measurement beams, and to start a particle filter algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target and adjust the beam direction.

[0092] In the particle filter algorithm, the received signal strength ratio of the auxiliary measurement beam is obtained as an observation value of the particle filter algorithm, an angle particle set of the current time slot is generated according to the half beam width of the data beam pointing angle of the previous time slot, the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution, the optimal angle particle serving as the path angle estimation value is selected according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle. The particle weight is used to measure the matching degree of the particle and the observation value.

[0093] To verify the effectiveness of the scheme, the following further explains and describes the data of Experiment I and Experiment II:

[0094] Experiment I: The beam tracking problem between the fixed millimeter wave communication base station (receiving end) and the millimeter wave communication mobile terminal (transmitting end) is studied. The transmitting end is an omnidirectional antenna, and only the beam tracking of the receiving end is considered. It is assumed that the communication channel follows the block fading model, which is defined as follows:

[0095]

[0096] where g l,t is the path complex coefficient, and t represents the influence of the t-th time slot scattering path. The mobile terminal motion model is established as:

[0097] θ t = θ t-1 + Ω + n t

[0098] where Ω is a fixed value of the channel angle change between adjacent time slots, and is a constant; n t is a random disturbance angle, which follows a Gaussian distribution with a mean of 0 and a variance of , that is, The simulation parameter settings are shown in Table 1.

[0099] Table 1 Simulation parameter settings

[0100]

[0101]

[0102] The experiment shows the tracking error, signal-to-noise ratio fading, tracking times, realignment times, and measurement overhead of the scheme ABTPF and the existing SPF and EKF under different conditions.

[0103] Figure 3 The tracking results of ABTPF in LoS scenario are shown in the following figures, where σ θ = 0.05δ, Ω = 0.1δ, ι = 1dB, and the SNR before beamforming. The angle tracking results of ABTPF in 1000 time slots are shown in the following figures. Figure 3 (a) The comparison of the angle changes between ABTPF tracking and channel angle is shown in the following figure, where ABTPF keeps the angle unchanged in some time slots because the SNR fading does not reach the preset tracking threshold ι, thus the tracking is not triggered. Figure 3 (b) The normalized tracking error of ABTPF is shown in the following figure, which is obtained by calculating the ratio of the error of each tracking time slot to the half beamwidth δ. In the simulation of 1000 time slots, ABTPF triggers tracking 165 times, accounting for only 16.5% of the total time slots, while SPF and EKF schemes need to track in each time slot, i.e. 1000 times of tracking.

[0104] Figure 4 The tracking error and tracking times of ABTPF under different tracking thresholds ι (ι = 0.5dB, 1.0dB, 1.5dB) in LoS communication scenario are shown in the following figures. The motion model parameters are: σ θ = 0.05δ, Ω = 0.1δ. The tracking error is the average value of 1000 repeated simulations. It can be seen from the figure that the larger the tracking threshold, the smaller the average tracking times. Under the tracking thresholds of ι = 1.0dB and ι = 1.5dB, the ABTPF tracking error curves are close to each other, which may be because the SNR drop caused by ABTPF tracking error is mostly concentrated in 1dB or below.

[0105] Figure 5 The SNR after beamforming and tracking error are shown in the following figures in LoS communication scenario when ε = 10dB, σ θ = 0.15δ, Ω = 0.2δ, ι = 0. Since SPF and EKF are model-driven tracking methods, σ θ and Ω are known in the simulation. Figure 5 It can be seen that when ABTPF tracks in each time slot, the SNR after beamforming is almost coincided with the upper bound without tracking error, and the tracking error is basically stable below 0.1. While the tracking error of SPF is stable at 0.25, and for EKF algorithm, due to the error accumulation problem, its tracking error reaches a maximum of 0.9.

[0106] Figure 6 and Figure 7This paper demonstrates the performance of different algorithms (ABTPF, SPF, and EKF) under various conditions, aiming to evaluate their performance in the face of varying degrees of "velocity" changes (measured by angle increment Ω and angle perturbation variance σ). θ (This reflects) robustness against noise interference. Figure 6 It shows that when σ θ The effect of the angle increment Ω on the tracking error and number of tracking iterations of ABTPF, SPF, and EKF when Ω = 0.05δ and ι = 1dB. Note that the SPF and EKF algorithms assume the parameter σ... θ Ω and ε are known. When Ω = 0.3δ and ι = 1dB, Figure 7 Showing different σ θ The tracking errors and number of tracking attempts for ABTPF, SPF, and EKF are calculated. Combined with... Figure 6 and Figure 7 It can be seen that even with rapid changes in angle (Ω = 0.3δ or σ), θ With a signal-to-noise ratio (SNR) of 0.25δ, the maximum tracking error of ABTPF does not exceed 0.18. In contrast, although SPF exhibits higher stability under different SNRs and "moving speeds," its tracking error is twice or more than that of ABTPF in this case. EKF's tracking error does not exceed 0.4, but this is achieved under conditions of multiple realignments. From the perspective of the number of tracking attempts, ABTPF can dynamically adjust the number of tracking attempts according to the tracking threshold, and its maximum number of tracking attempts is only about half that of the other two methods.

[0107] Tables 2 and 3 compare the performance of the three algorithms (ABTPF, SPF, and EKF) in this proposal under different degrees of "speed" variation under LosS and NLoS conditions, including realignment ratio, average tracking error, and average measurement overhead. The simulation was performed 1000 times, with the parameters σ for SPF and EKF being known. θ The algorithm for Ω. The realignment criterion is: the signal-to-noise ratio γ after the current beamforming. t Compared to the signal-to-noise ratio after beamforming realignment P(t RE If the fading exceeds 3dB compared to the time slot index where realignment is performed, the communication link is considered lost, triggering realignment. The realignment ratio represents the proportion of realignment attempts to the total tracking time slots, and the average tracking error is the average tracking error per time slot in the simulation. Average measurement overhead is measured by the average number of beam measurements per time slot. SPF and EKF use one measurement beam per time slot with 1 beam measurement attempt, while ABTPF uses two measurement beams with 2 beam measurements. For time slot t where realignment is performed... RE, the number of beam measurements is 64, because a detailed search is needed on a codebook of 64 discrete Fourier transform (DFT) beams. Table 2 is the simulation result under the condition of σ θ = 0.15δ, Ω = 0.2δ, i = 1dB, ε = 10dB. In this case, the channel angle changes slowly, and all three methods do not need to be realigned and can stably track the main path angle of the channel. In contrast, the average tracking error of ABTPF is about 25% of the other two schemes, and the average measurement overhead is only 34% of the other two schemes. Table 3 is the performance under the condition of σ θ = 0.35δ, Ω = 0.4δ, i = 1dB, ε = 10dB. In this case, the channel angle changes rapidly. In the LoS and NLoS scenarios, SPF and EKF have 2.9% and 17.5% realignment rates, respectively, while the realignment rate of ABTPF is less than 1% (the highest is only 0.8%). In terms of tracking error, the average error of ABTPF is not more than 0.159, while the minimum tracking error of SPF and EKF is 0.441 and 0.573, respectively, which is about 2.7 times and 3.6 times of ABTPF. At the same time, the average measurement overhead of ABTBF is lower than that of EKF and SPF. It can be seen that even under the condition of rapid channel change, ABTPF still maintains high tracking accuracy and low overhead.

[0108] Table 2 Performance comparison of different algorithms: σ θ = 0.15δ, Ω = 0.2δ, i = 1dB, ε = 10dB

[0109]

[0110]

[0111] Table 3 Performance comparison of different algorithms: σ θ = 0.35δ, Ω = 0.4δ, i = 1dB, ε = 10dB

[0112]

[0113] From the above experimental data, it can be seen that the ABTPF scheme of the present application can adaptively adjust the tracking overhead according to the set tracking threshold without relying on the known motion model, while maintaining high tracking accuracy. Under the typical statistical channel model, ABTPF exhibits superior performance compared to SPF and EKF.

[0114] Experiment 2: In order to further verify the performance of the ABTPF scheme of the present application, a real scene is modeled using a ray tracing software. This scene data contains both LoS and NLoS channels, making the changes of channel gain and path angle more complex. In the simulation scene, the base station (BS) is the receiving end and the mobile terminal (UE) is the transmitting end, both of which use auxiliary beam pairs for beam tracking to estimate AoA and AoD, respectively.Figure 8 The environment and UE moving route map of the embodiment two scenario are shown. The specific scenario and ABTPF algorithm related parameters are shown in Table 4.

[0115] Table 4 Scenario and ABTPF related parameters

[0116]

[0117] Figure 9 The beamforming post SNR variation of the scenario is shown. The effective channel gain of both BS and UE end has a sudden change around 40m. The reason for this change is that there is no building blocking between BS and UE, AoA and AoD transmit sudden change, the channel changes from NLoS to LoS. From 380m, the path gain drops suddenly and is rapidly decaying due to the increase of distance and there is building blocking, AoD changes greatly, the channel changes from LoS to NLoS, which causes the path gain to drop suddenly and is rapidly decaying.

[0118] Figure 10 The successful application of ABTPF scheme in high-precision ray tracing scenario (UE moving speed is 72km / h) is shown. Figure 10 The SNR after ABTPF beamforming (red curve) is shown in (a) of FIG. 12, compared with the maximum SNR (black curve) obtained without tracking error. Figure 10 (b) of FIG. 12 shows the indicator of ABTPF realignment during the tracking process, where “1” means there is realignment and “0” means there is no realignment. From the beginning of the tracking to about time slot 3700, ABTPF algorithm successfully tracks the main path of the channel in NLoS scenario (SNR after beamforming is close to the optimal value) and there is no realignment. From about time slot 3700, since NLoS path and LoS path exist in the channel at the same time and NLoS path still has higher path strength, the effective channel gain of ABTPF drops greatly at about 300 time slots of SNR. At about time slot 4100, NLoS path weakens and the channel path switches to LoS path, ABTPF needs to realign, see (b) of FIG. 12. At about time slot 35000, LoS path weakens and the channel switches to NLoS path, ABTPF also successfully tracks the main path angle of the channel. As the distance of UE becomes far and the angle changes rapidly, the channel gain becomes very weak, after time slot 37500 at the end of the trajectory, the communication link is often lost and frequent beam alignment is needed. In fact, ABTPF only tracks 215 times during the whole tracking process, accounting for about 0.53% of the total time slots, and the measurement overhead is 1.06x10 Figure 10 Figure 10 During the whole tracking process, ABTPF only tracks 215 times, accounting for about 0.53% of the total time slots, and the measurement overhead is 1.06x10 -2 . ​

[0119] Table 5 shows the tracking performance of ABTPF algorithm under the UE moving speed of 20km / h, 43.2km / h and 72km / h. The tracking accuracy κ is defined as the proportion of the times that the beamforming post-SNR drops more than 3dB in the total tracking slots. From Table 5, it can be seen that the tracking accuracy of ABTPF is almost the same under the three moving speeds. When the moving speed is slow, ABTPF reduces the measurement overhead by reducing the tracking times; while when the moving speed is fast, ABTPF ensures the stability of the tracking accuracy by increasing the tracking times.

[0120] Table 5 Performance under different moving speeds

[0121] Moving speed (km / h) 20 43.2 72 Tracking time slot ratio 0.039％ 0.32％ 0.53％ Measurement overhead 0.078 x 10 -2 ]] 0.65 x 10 -2 ]]> 1.06 x 10 -2 ]]> Kappa 2.00％ 2.35％ 2.17％

[0122] The above two experimental data show that the scheme of the present application can adapt to various mobility communication scenarios and meet the high-quality demand of millimeter wave communication link under different dynamic conditions.

[0123] Unless specifically stated otherwise, the relative steps, numerical expressions, and numerical values of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0124] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0125] The units and method steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been described in the above description in general terms. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation does not exceed the scope of the present application.

[0126] Those skilled in the art can understand that all or part of the steps in the foregoing method can be instructed by programs to the relevant hardware to complete, and the programs can be stored in a computer readable storage medium, such as a read-only memory, a magnetic disk or an optical disk. Alternatively, all or part of the steps of the foregoing embodiments can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the foregoing embodiments can be implemented in the form of hardware or in the form of a software functional module. The present application is not limited to any specific form of combination of hardware and software.

[0127] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present application, which are used to illustrate the technical solutions of the present application, rather than limit the same. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can make modifications or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some of the technical features, within the technical scope disclosed by the present application. Such modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A particle filter based beam tracking method for millimeter wave scenarios, characterized in that, The method comprises: determining whether the difference between the beamforming post-signal-to-noise ratio of the current time slot and the previous time slot exceeds a set tracking threshold, and if the difference exceeds the tracking threshold, determining that the current time slot needs beam tracking; on the basis of the data beam direction, adding two specified offset angles and generating a pair of auxiliary measurement beams to start a particle filtering algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target in the current time slot and adjust the beam direction; in the particle filtering algorithm, the received signal strength ratio of the auxiliary measurement beam is obtained, the received signal strength ratio is taken as the observation value of the particle filtering algorithm, the angle particle set of the current time slot is generated according to the half wave width of the data beam pointing angle of the previous time slot, the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution, the optimal angle particle is selected as the path angle estimation value according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle, and the particle weight is used to measure the matching degree of the particle and the observation value.

2. The particle filter based beam tracking method for millimeter wave scenarios according to claim 1, characterized in that, For the measurement auxiliary beam, the receiving end in the data beam direction is set as the fixed end and the transmitting end is the mobile end. The data beam pointing angle at the t-1 time slot is Then, the receiving signals of the two auxiliary measurement beams formed by adding and subtracting two perturbation angles δ are respectively represented as: wherein, denotes the conjugate transpose; w, f denote the receive and transmit beamforming vectors, respectively, and H t is the channel matrix for the t-th time slot, and s is the pilot signal, and k denotes the pilot length; is the channel path angle of departure; M R and M T denote the number of antennas at the receiver R and transmitter T, respectively; z - and z + are additive complex Gaussian noises with mean 0 and variance , and are mutually independent.

3. The particle filter based beam tracking method for millimeter wave scenarios according to claim 2, characterized in that, The process of obtaining the received signal strength ratio of the auxiliary measurement beam comprises: Based on the received signal, the received signal strength of the auxiliary measurement beam is obtained, and the received signal strengths of the two auxiliary measurement beams are respectively represented as: wherein, respectively represent the received signals of the two auxiliary measurement beams. calculating the ratio metric according to the received signal strength, wherein the calculation process of the ratio metric is represented as: The received signal strength ratio is calculated in accordance with the ratio metric, which is given by:

4. The particle filter based beam tracking method for millimeter wave scenario according to claim 2, characterized in that, generating the angle particle set of the current time slot according to the half wave width of the data beam pointing angle of the previous time slot comprises: Set the angle particle set x of the t time slot particle filter algorithm t Contains N particles, the generation strategy is represented as: where i denotes the particle index, The uniform distribution range of the particles is represented by 5. The particle filter based beam tracking method for millimeter wave scenario according to claim 3, characterized in that, calculating the angle particle weight of the current time slot based on the probability density function of the double non-central F distribution comprises: setting the particle to follow the double non-central F distribution with degrees of freedom υ1 and υ2, and determining the particle weight by using the double non-central F distribution probability density, wherein the particle weight calculation process is: for the i th particle, the related double non-central parameters are represented as: P denotes a transmit power, denotes that the channel estimate value is calculated using the i-th particle at time slot t; the weight of the i-th particle is calculated by denotes that the channel estimate value is calculated using the i-th particle at time slot t; the weight of the i-th particle is calculated by F denotes a probability density function of a doubly non-central F distribution.

6. The particle filter based beam tracking method for millimeter wave scenarios according to claim 5, characterized in that, selecting the optimal angle particle as the path angle estimation value according to the angle particle weight, and updating the data beam direction of the current time slot based on the optimal angle particle comprises: Utilizing normalizing the particle weights, resampling the particles according to the normalized particle weights, in the resampling, retaining the particles with weights greater than a specified value according to the weight size and the weights and obtaining a new particle set and a weight set according to the retained particles through copying and the weights wherein the new particle set and the weight set are represented as: representing a resampling operation strategy function, updating the particle set and the weight set through the resampling operation and the weights Using the resampled particle set, the estimated value of the target data beam pointing angle AoA is calculated, and the estimated value calculation formula is represented as: N is the total number of particles.

7. The particle filter based beam tracking method for millimeter wave scenarios of claim 2, wherein, the process of comparing the beamforming post-signal-to-noise ratio of the current time slot and the previous time slot comprises: computing the post-beamforming signal-to-noise ratio γ in the tth time slot t and the post-beamforming signal-to-noise ratio γ in the t-1th time slot t-1 if the post-beamforming signal-to-noise ratios of the two time slots satisfy the condition: |γ t -γ t-1 |>ι Then beam tracking is triggered once in the tth time slot, where the formula of the calculation of the beamforming post-SNR is expressed as: P represents the transmit power, and i is the tracking threshold.

8. A particle filter based beam tracking system for millimeter wave scenarios, characterized in that, comprises: a judgment module and a tracking module, wherein the judgment module is used to determine whether the current time slot needs beam tracking by comparing the beamforming post-signal-to-noise ratio of the current time slot and the previous time slot; the tracking module is used to add two specified offset angles in the data beam direction, generate a pair of auxiliary measurement beams, and start a particle filtering algorithm based on the auxiliary measurement beams to dynamically estimate the motion state of the target and adjust the beam direction; in the particle filtering algorithm, the received signal strength ratio of the auxiliary measurement beam is obtained, the received signal strength ratio is taken as the observation value of the particle filtering algorithm, the angle particle set of the current time slot is generated according to the half wave width of the data beam pointing angle of the previous time slot, the angle particle weight of the current time slot is calculated based on the probability density function of the double non-central F distribution, the optimal angle particle is selected as the path angle estimation value according to the angle particle weight, and the data beam direction of the current time slot is updated based on the optimal angle particle, and the particle weight is used to measure the matching degree of the particle and the observation value.

9. An electronic device, comprising: comprises: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program can be executed to implement the method in any one of claims 1-7.

Images