CN117395669A

CN117395669A - Channel clustering and modeling method for 6G unmanned aerial vehicle air-to-ground communication scene

Info

Publication number: CN117395669A
Application number: CN202311341950.6A
Authority: CN
Inventors: 刘玉; 张兆磊
Original assignee: Shandong University
Current assignee: Shandong University
Priority date: 2023-10-17
Filing date: 2023-10-17
Publication date: 2024-01-12

Abstract

The invention relates to a channel clustering and modeling method for a 6G unmanned aerial vehicle air-to-ground communication scene, which comprises the following steps: acquiring a full-frequency-band multi-dimensional channel data set of unmanned aerial vehicle air-to-ground communication; performing channel clustering on the full-band multi-dimensional channel data set based on a VB-GMM algorithm; acquiring an evolution process of a cluster center based on an MCD algorithm; acquiring channel characteristics based on clustering and evolution results; and establishing an unmanned aerial vehicle air-to-ground channel model based on the channel characteristics, and realizing channel clustering and modeling for a 6G unmanned aerial vehicle air-to-ground communication scene. The invention considers mean value and covariance information of MPC at the same time, and realizes clustering of MPC in time delay domain and space domain. The method can realize channel clustering of sub-6GHz and millimeter wave frequency bands and can automatically determine the optimal cluster number. The invention introduces a four-state Markov chain to describe the evolution process of the cluster, and provides a 6G unmanned aerial vehicle air-to-ground communication channel model based on the cluster, and typical channel characteristics are researched.

Description

Channel clustering and modeling method for 6G unmanned aerial vehicle air-to-ground communication scene

Technical Field

The invention relates to a channel clustering and modeling method for a 6G unmanned aerial vehicle air-to-ground communication scene, and belongs to the technical field of channel clustering and modeling.

Background

Unmanned aerial vehicle communication has been attracting attention in industry and academia as an important component of future sixth generation (6G) wireless communication technology and air-space-sea integrated networks. In order to realize the communication requirements of huge flow, huge connection and differentiation of a 6G mobile communication network in the future, the unmanned aerial vehicle can be rapidly deployed as an air base station by utilizing the advantages of wide coverage range, high mobility and the like, and can supplement network flow, enhance network coverage area, provide edge service and the like for a ground base station. For example, unmanned aerial vehicle base stations in large concerts and other places with dense traffic can provide powerful network guarantees. Meanwhile, unmanned aerial vehicle communication can be used as a low-altitude aircraft to accelerate the transition of future 6G mobile communication from two-dimensional plane coverage to three-dimensional full coverage, and the method is widely applied to various scenes such as urban traffic, target detection, emergency rescue, mountain area exploration and the like. The requirements of China push the unmanned aerial vehicle field to break through development, strengthen the construction of the air, the sky, the land and the sea integrated communication network, and make the unmanned aerial vehicle become a new economic growth point for domestic civilians. It is well understood that wireless communication channels are the basis for communication network design, system optimization, and performance assessment. From the above background, it can be clearly found that it is crucial to study the air-to-ground communication channel of a drone.

In the unmanned aerial vehicle air-to-ground communication process, the received multipath signals often show a cluster structure. A cluster is a collection of a set of multipaths in a wireless channel that have similar characteristic properties (e.g., angle of arrival, angle of departure, time delay, etc.). Efficient clustering and analysis of multipath components (multipath components, MPCs) of a channel with similar properties helps to build an accurate and efficient channel model. Current cluster-based channel models also focus primarily on traditional terrestrial communication channels, such as indoor communications, terrestrial vehicle-to-vehicle communications, and high-speed rail communications. However, the unmanned aerial vehicle air-to-ground communication channel has a large elevation angle, any track and the like difference from the conventional ground channel. It results in significant differences in characteristics between clusters and within clusters, such as the number, distribution, and lifetime of clusters. Therefore, developing a proper clustering algorithm research, and analyzing the relevant characteristics of the unmanned aerial vehicle air-to-ground channel is of great importance.

To date, several algorithms have been applied to clustering of MPCs in different scenarios. The earliest clustering method was based on the Saleh-Valenzuela (SV) model, but the model was too simplified to fully reflect the channel characteristics. In the field of wireless communication, a K-power-means (KPM) algorithm for improving performance of MPCs by considering the power thereof is widely applied to channel clustering in various environments, but the method needs to know the number of categories of clusters in advance. The clustering method based on the shape does not need prior information of the number of clusters to be clustered, but only focuses on the distribution of MPCs in a time delay domain, and lacks angle information. In addition to the above conventional algorithms, clustering algorithms based on gaussian mixture models are increasingly used for channel clustering. Such an algorithm is more suitable for channel clustering as it can contain more statistical features. However, the literature for this latest approach is very limited. Furthermore, clusters in the channel may exhibit complex evolutionary behaviour such as birth/death, sports, splitting or merging. Correlation studies have shown that tracking methods based on multipath component distances (multipath component distance, MCD) can achieve better robustness and accuracy in processing time-varying channel data.

In summary, the channel clustering research in unmanned aerial vehicle air-to-ground communication is still in a very early stage. Channel measurements at a frequency of 6.5GHz have been developed and the extracted MPCs are clustered using a conventional KPM algorithm. However, it only considers information of the delay domain, lacking angle information. Furthermore, it does not relate to the millimeter wave (mmWave) band. Also, the clustering method based on MCD analyzes channel characteristics. However, it also focuses on the frequency band below 6GHz (sub-6 GHz) and uses the traditional clustering approach. Therefore, by adopting a channel clustering algorithm with good performance, novelty and effectiveness, it is important to deeply analyze the characteristics of the full-band multi-dimension clusters of the unmanned aerial vehicle air-to-ground communication channel and establish an accurate channel model.

Disclosure of Invention

Aiming at the defects of the prior art, the invention aims to establish a channel clustering and modeling method for a 6G unmanned aerial vehicle air-to-ground communication scene; the invention can provide reference for the deployment of the unmanned aerial vehicle communication network and lay a foundation for the integration of the air, the ground and the sea integrated network in the future.

The unmanned aerial vehicle is widely applied to various fields by the advantages of flexible deployment, wide coverage, strong maneuverability and the like, and is oriented to the 6G mobile communication technology. The development of any wireless communication technology leaves away a channel model that accurately describes the characteristics of the channel. Therefore, development of a channel clustering and modeling method research in an unmanned aerial vehicle air-to-ground communication scene is needed. The invention adopts a variable decibel leaf-Gaussian mixed model (variational Bayesian-Gaussian mixture model, VB-GMM) algorithm to carry out multipath clustering, the algorithm considers the space-time characteristics of multipath clustering, and the optimal clustering number can be automatically obtained. And acquiring an evolution process of the cluster center based on the MCD algorithm. In addition, the invention also researches the characteristics of the clusters and the characteristics among the clusters, such as the number of the clusters, the power distribution, the angle and time delay offset and the angle and time delay diffusion characteristics of the clusters, and fully analyzes the evolution characteristics of the clusters. Based on clustering results, the invention provides a novel unmanned aerial vehicle air-to-ground channel model, and introduces a four-state Markov chain to describe the evolution process of clusters. The work can provide important reference for modeling the ground communication channel of the 6G unmanned aerial vehicle in future, and provide accurate and low-complexity solution and data support for the design and evaluation of the unmanned aerial vehicle communication system.

Term interpretation:

1. the ray tracing simulation method comprises the following steps: ray tracing is a deterministic channel modeling technique based on geometrical optics and uniform diffraction theory. It takes into account interactions between rays and scatterers, including reflection, scattering and diffraction phenomena. Ray tracing is widely used in the field of wireless communications to simulate radio wave propagation to derive all possible ray paths between transceivers in multipath channels.

2. Wireless channel measurement system based on software radio USRP: channel measurement refers to the completion of measurement activities of the radio wave propagation channel based on the channel detector under certain specific environments. The wireless channel measurement system based on the USRP of the software radio equipment mainly comprises an air platform and a ground platform. And the real-time transmission of the detection signals is completed when the aerial platform measures, and the ground platform completes the real-time receiving and storing of the detection signals after the signals are transmitted through the wireless channel. The aerial platform consists of an unmanned plane, a USRP, an antenna, a portable power supply and a microcomputer. The transmitted signal, such as a pseudo-noise PN sequence signal, is transmitted and modulated using USRP. The ground platform consists of USRP, an antenna and a computer. The signal is received using USRP and stored in the ground computer.

The technical scheme of the invention is as follows:

a channel clustering and modeling method for a 6G unmanned aerial vehicle air-to-ground communication scene comprises the following steps:

acquiring a full-frequency-band multi-dimensional channel data set of unmanned aerial vehicle air-to-ground communication;

performing channel clustering on the full-band multi-dimensional channel data set based on a VB-GMM algorithm;

acquiring an evolution process of a cluster center based on an MCD algorithm;

acquiring channel characteristics based on clustering and evolution results;

and establishing an unmanned aerial vehicle air-to-ground channel model based on the channel characteristics, and realizing channel clustering and modeling for a 6G unmanned aerial vehicle air-to-ground communication scene.

According to the invention, the method for acquiring the full-band multi-dimensional channel data set of the unmanned aerial vehicle air-to-ground communication comprises the following steps: acquiring a typical channel parameter, namely a full-band multi-dimensional channel data set by adopting a wireless channel measurement system or a ray tracing simulation method based on USRP of software radio equipment; typical channel parameters include time delay τ, arrival pitch angle θ ^R Pitch angle θ ^T Azimuth of arrival phi ^R Azimuth angle phi of departure ^T And a received power P; the full frequency band refers to sub-6GHz and millimeter wave frequency bands.

According to the invention, the method for carrying out channel clustering on the full-band multidimensional channel data set based on the VB-GMM algorithm comprises the following steps:

Assuming that the GMM consists of a linear combination of K gaussian distributions, each with its own mean μ _i Covariance Σ _i And probability coefficient pi _i Wherein i is from 1 to K; each data point x in the GMM is obtained by first applying a probability coefficient pi to _i Randomly sampling, and sampling the density function of the ith component to generate an observed value; thus, the obtained mixed model is shown as the formula (I) and the formula (II)The illustration is:

therefore, the calculation formula of the posterior probability is shown as formula (III):

responsibility p (i|x) represents the probability that each MPCs is generated by the ith gaussian distribution; each MPCs is assigned to a gaussian distribution with the highest posterior probability; formula (I) is rewritten as formula (IV):

wherein T is _i Representing an accuracy matrix; VB-GMM is obtained by applying a priori probability distribution to the parameters { pi, μ, T };

the Dirichlet prior distribution of the parameter pi and the Gauss-Wishare prior distribution of the parameter { mu, T } are introduced;

the Dirichlet a priori distribution is used for pi as shown in formula (V):

wherein Γ (·) is a gamma function; dir (·) is a Dirichlet function; parameter alpha _i Expressed as an effective a priori number of observations associated with the blended components;

the Gauss-Wishare a priori distribution is used for (μ, T), as shown in formula (VI):

wherein,refers to Wishart distribution; v represents the degree of freedom, V represents the scale matrix;

h= { Z, pi, μ, T } using a variational method, an approximation q (h) of VB-GMM is calculated, where Z is a hidden variable, and the approximation is expressed as the product of formula (VII) as follows:

q(h)＝q(Z)·q(π)·q(μ,T)(VII)

the results are shown in the formulas (VIII), (IX), (X), (XI) and (XII), respectively:

q(π)＝Dir(π∣{α _i })(IX)

wherein,ρ _in representing the weight of the nth multipath component in the ith gaussian component; z _in Is a binary variable, and the sum is equal to 1; η (eta) _i And U _i Wishare distribution parameters representing degrees of freedom and scale matrices, respectively; obtaining an optimal variation lower limit through multiple iterations; in addition, the optimal approximation q (h) of the true posterior p (h|x) is also obtained through a simple iterative updating program; thus obtaining a label for each MPCs.

According to the invention, the VB-GMM algorithm is used for carrying out channel clustering on the full-band multidimensional channel data set, and the method specifically comprises the following steps:

step 1: carrying out normalization operation on channel original data, namely a full-band multidimensional channel data set, in an unmanned aerial vehicle air-to-ground communication scene;

step 2: setting the maximum iteration times and initializing prior parameters of the mixed model;

step 3: applying prior parameters of the current hybrid model, and calculating a posterior probability value by a formula (III);

step 4: re-estimating the parameter values of the mixed model by using the current posterior probability values according to formulas (IX), (X), (XI) and (XII), so as to maximize the likelihood values of the mixed model;

Step 5: calculating a log likelihood function;

step 6: judging whether the log likelihood function has converged or reaches the maximum iteration number; if not, returning to the step 3 to perform calculation again; if yes, a channel clustering result is obtained, and the total number of clusters and clusters affiliated by each MPC is output;

according to the invention, the evolution process of the cluster center is acquired based on the MCD algorithm, which comprises the following steps:

the calculation formula of the MCD is shown in formulas (XIII), (XIV) and (XV) by adopting a tracking method based on the multipath component distance MCD:

wherein ζ is the weight of the delay distance, τ _std Is the standard deviation of MPCs delay, deltaτ _max Is the maximum value of MPCs delay difference; based on MCD distanceAnd judging whether the cluster center is evolved or not. If the distance between the centers of two nearest clusters exceeds a threshold value, the cluster is considered to have a death phenomenon in the future; also, if the current cluster center cannot find the past cluster center, the cluster is considered to be new.

Further preferably ζ=3.

According to the invention, the channel characteristics are acquired based on clustering and evolution results, wherein the clustering and evolution results provide labels of each MPCs cluster and the number of clusters; the channel characteristics include intra-cluster characteristics, inter-cluster characteristics, and evolution characteristics; the intra-cluster characteristics include: the number of MPCs in each cluster, the power attenuation factor, the intra-cluster rice K factor, the intra-cluster delay distribution and the intra-cluster angle distribution; the inter-cluster characteristics include the number of clusters, inter-cluster power and delay characteristics, and inter-cluster angle characteristics.

Preferably, according to the present invention, the number of MPCs within a cluster represents the abundance of multipath effects in the channel; the method for obtaining the number of MPCs in the cluster comprises the following steps: firstly, obtaining the number of MPCs at each moment according to a VB-GMM algorithm; then, calculating a cumulative distribution function for the MPCs at all times to obtain the number of MPCs in the cluster;

the method for obtaining the power attenuation factor comprises the following steps: when setting the power P of the c-th MPC in the cluster _c In dB, the values of the radiation beams have linear inclined attenuation relations with time delay, and different numbers of radiation beams tend to have different decreasing trends, as shown in the formula (XVI):

P _c (dB)＝-k·τ _c +b(XVI)

wherein k is a power attenuation factor, b is a power attenuation intercept, τ _c Delay for the c-th MPC in the cluster; the power attenuation factor k is found by equation (XVI).

The in-cluster rice K factor lambda characterizes the power relation between MPCs with the strongest power in the cluster and the rest MPCs, and reflects the power distribution relation of the MPCs in the cluster; the calculation formula of the intra-cluster rice K factor Λ is shown in formula (XVII):

the intra-cluster delay distribution comprises two types of distribution relations of intra-cluster root mean square delay spread and intra-cluster delay offset; intra-cluster root mean square delay spreadThe calculation formula is shown as a formula (XVIII);

calculating the intra-cluster root mean square delay spread of each moment according to formula (XVIII);

The intra-cluster delay offset is the difference between the intra-cluster sub-path delay and the cluster average delay, i.e., τ _off ＝τ _c -mean(τ _c )；

The intra-cluster angle distribution comprises two types of distribution relations, namely intra-cluster root mean square angle expansion and intra-cluster angle offset; four types of intra-cluster root mean square angle expansion are provided, and the calculation formulas of the four types of intra-cluster root mean square angle expansion are shown as formula (XIX) and formula (XX);

wherein,ε ^T/R the intra-cluster root mean square azimuth angle diffusion and intra-cluster root mean square pitch angle diffusion of the sending angle and the arrival angle are respectively carried out;

calculating the intra-cluster root mean square angle spread for each time according to formula (XIX) and formula (XX);

the intra-cluster angle offset is the difference between the intra-cluster sub-path angle and the cluster average angle, i.e. satisfiesAnd->

According to the invention, the number of clusters is obtained according to the clustering result, and the richness of the clusters in the unmanned aerial vehicle air-to-ground channel is reflected;

the method for calculating the inter-cluster power and time delay characteristics comprises the following steps: firstly, obtaining original inter-cluster power and original inter-cluster time delay according to different cluster labels obtained by a VB-GMM algorithm; comparing the original inter-cluster power with the time delay and power of an original inter-cluster time delay and line of sight (LoS) path to eliminate the calculated interference influence of the LoS paths at different positions and obtain the relative time delay and relative power of each cluster at each moment; classifying the relative time delay and the relative power belonging to the same cluster at all moments; finally, calculating the average value of the relative time delay and the relative power, namely the inter-cluster power and the time delay characteristic;

The method for calculating the inter-cluster angle characteristics comprises the following steps: firstly, dividing different clusters according to a VG-GMM algorithm to obtain four angle information of each MPC in the clusters, wherein the four angle information comprises an arrival pitch angle, a departure pitch angle, an arrival azimuth angle and a departure azimuth angle; obtaining the angular distribution of the moment by calculating the average value of four angle information of each MPC in different clusters; and carrying out statistical distribution on the angle distribution at all the moments to obtain the inter-cluster angle characteristics.

According to the invention, the evolution characteristic is analyzed based on the cluster identification CLID; assigning a new CLID to a new cluster; the old cluster inherits the CLID of the last cluster; through the evolution rule of a four-state Markov chain reaction channel; the four states in the four-state Markov chain include: no generation of new clusters S0, generation of new clusters S1, death of old clusters S2, generation of new clusters and death of old clusters S3; the state transition probability matrix T of the four-state Markov chain is shown as a formula (XXI);

the statistical distribution calculation is carried out on the survival time of the clusters in the CLID, and as a result, a plurality of clusters only exist in a part of time period, and obvious life-extinguishing phenomenon exists; calculating a cumulative distribution function for the survival time of the cluster in the whole flight process, comparing and analyzing with different probability distribution characteristics, and finding that the survival time of the cluster follows the lognormal distribution;

The survival time distribution of the clusters and the evolution characteristics of the state transition probability matrix reaction clusters of the four-state Markov chain.

According to the invention, preferably, the method for establishing the unmanned aerial vehicle air-to-ground channel model based on the channel characteristics comprises the following steps:

(1) Setting basic information including environmental parameters, frequency bands, unmanned aerial vehicle flight trajectories, three-dimensional distances between a transmitter (Tx) and a receiver (Rx);

(2) Generating a LoS path and a scatterer cluster; generating time delay and power parameters of the clusters through a random process according to the characteristics among the clusters; according to the characteristics in the cluster, specific parameters of MPCs are obtained through corresponding distribution, wherein the specific parameters comprise the number, time delay and power of the MPCs in the cluster;

(3) In the evolution process, different clusters are endowed with different survival times;

(4) Generating a next state according to the current state based on the four-state Markov chain transfer matrix;

(5) Judging whether birth and death occur or not according to the generated state; when the state of the markov chain is 0, it is assumed that the cluster has not changed; when the state of the Markov chain is 1, generating new clusters randomly, and generating MPCs in the clusters according to the new clusters; when the state is 2, the cluster to be extinguished is emptied according to the survival time of the cluster; when the state of the Markov chain is 3, completing the process of extinction of the state 1 and the state 2;

(6) Combining all generated clusters and corresponding MPCs in the clusters to obtain a channel impulse response CIR, wherein the CIR is shown as a formula (XXII):

where K (t) is the number of clusters, C _i (t) isThe number of MPCs in the ith cluster at time t, delta (·) is the dirac impact function, and the power, phase and delay of the c-th ray in the ith cluster are P _i,c 、φ _i,c And τ _i,c The method comprises the steps of carrying out a first treatment on the surface of the Similarly, the LoS path parameters are respectively set to P ₀ 、φ ₀ And τ ₀ 。

The beneficial effects of the invention are as follows:

based on the background and the defects of the prior art, the invention aims to provide a novel 6G time-varying unmanned aerial vehicle air-to-ground communication channel clustering and modeling method based on machine learning. The main contributions and novel aspects of the invention may be summarized as follows:

1) And based on the unmanned aerial vehicle air-to-ground communication channel data set, adopting a VB-GMM algorithm to perform channel clustering. The full-frequency unmanned aerial vehicle air-to-ground channel clustering research is carried out in the sub-6GHz and millimeter wave frequency bands for the first time. The algorithm considers the mean value and covariance information of the MPC at the same time, and realizes the clustering of the MPC in a time delay domain and a space domain. The method can realize the full-band channel clustering of sub-6GHz and millimeter waves. In addition, the method can automatically determine the optimal cluster number.

2) The inter-cluster and intra-cluster characteristics of the unmanned aerial vehicle air-to-ground communication are comprehensively analyzed, including the number of clusters, cluster delay, cluster power, intra-cluster angle and delay offset. And describing evolution behavior of a time-varying channel by adopting a cluster tracking method based on the MCD, and analyzing the survival time of the cluster.

3) Four-state Markov chains are introduced to describe the evolution process of the clusters, and a cluster-based unmanned aerial vehicle air-to-ground communication channel model is provided. From this channel model, typical channel characteristics were studied.

Drawings

Fig. 1 is a schematic diagram of a clustering and evolution architecture of an unmanned aerial vehicle air-to-ground communication channel;

fig. 2 is a schematic flow chart of a clustering algorithm of an unmanned aerial vehicle air-to-ground channel based on VB-GMM;

FIG. 3 is a schematic diagram of a four-state Markov chain;

FIG. 4 (a) is a schematic diagram of a simulation environment layout of a ray tracing model;

FIG. 4 (b) is a satellite view of a typical campus scenario and a detailed view of unmanned aerial vehicle flight;

fig. 5 is a schematic diagram of a clustering result of an air-to-ground channel of an unmanned aerial vehicle;

FIG. 6 (a) is a sub-6GHz band delay offset distribution histogram;

fig. 6 (b) is a millimeter wave frequency band delay offset distribution histogram;

FIG. 7 is a schematic diagram of an implementation flow of the evolution and extinction process of clusters;

fig. 8 (a) is a schematic diagram of power delay profiles of different moments of a sub-6GHz band cluster-based unmanned aerial vehicle air-to-ground channel model;

fig. 8 (b) is a schematic diagram of power delay profiles of different moments of a millimeter wave frequency band cluster-based unmanned aerial vehicle air-to-ground channel model.

Detailed Description

The invention is further defined by, but is not limited to, the following drawings and examples in conjunction with the specification.

Example 1

acquiring an evolution process of a cluster center based on an MCD algorithm;

acquiring channel characteristics based on clustering and evolution results;

Example 2

According to the channel clustering and modeling method for the 6G unmanned aerial vehicle air-to-ground communication scene in the embodiment 1, the acquisition of the unmanned aerial vehicle air-to-ground communication full-band multi-dimensional channel dataset comprises the following steps: obtaining typical channel parameters, namely full-band multi-dimensional channel number by adopting wireless channel measurement system or ray tracing simulation method based on USRP of software radio equipmentA data set; typical channel parameters include time delay τ, arrival pitch angle θ ^R Pitch angle θ ^T Azimuth of arrival phi ^R Azimuth angle phi of departure ^T And a received power P; the full frequency band refers to sub-6GHz and millimeter wave frequency bands. The frequency band has typical significance and rich application space in future 6G unmanned aerial vehicle air-to-ground communication scenes. Multi-dimensional refers to channel data that includes both the time delay domain and the spatial domain, while also taking into account power information. I.e., the data set contains sub-6GHz and millimeter wave space-time channel multipath information.

Fig. 1 shows the clustering and evolution phenomenon of MPCs in unmanned aerial vehicle air-to-ground communication. Different types of scatterers often affect the propagation of the radio. For example, rx typically detects signals from various propagation paths, such as high-rise buildings, flat houses, floors, trees, vegetation, and the like. In this case, the drone typically acts as Tx, while the ground end acts as Rx; furthermore, the clustering method has excluded the LoS path and only involves analysis of non-light-of-sight (NLoS) paths. This is more consistent with the concept of standardized channel modeling. In addition, if the LoS path is added to the clustering process, the clustering performance is reduced due to the large power difference between the LoS path and other multipaths.

Channel clustering is carried out on the full-band multidimensional channel data set based on a VB-GMM algorithm, and the method comprises the following steps:

along with the continuous improvement of the communication system bandwidth, the diversification of scenes and the high mobility of the terminal, the multipath of the channel presents large data characteristics, and the channel data also has a statistical distribution mode. Traditional distance-based clustering methods (such as K-means and FCM methods) cannot effectively mine statistical distribution patterns. The invention selects a clustering method based on GMM.

Assuming that the GMM consists of a linear combination of K gaussian distributions, each with its own mean μ _i Covariance Σ _i And probability coefficient pi _i Wherein i is from 1 to K; each data point x in the GMM is obtained by first applying a probability coefficient pi to _i Randomly sampling, and sampling the density function of the ith component to generate an observed value;

thus, the obtained hybrid model is shown in the formula (I) and the formula (II):

in order to cluster all MPCs, it is necessary to determine a posterior probability, also known as responsibility. Responsibility p (i|x) represents the probability that each MPCs is generated by the ith gaussian distribution; each MPCs is assigned to a gaussian distribution with the highest posterior probability; finding responsibility (posterior probability) is equivalent to solving a hidden variable problem. The parameters required for solving are: { pi _i ,μ _i ,σ _i }. An expectation-maximization (EM) algorithm is typically used to solve the maximum log-likelihood function by iterative optimization. However, EM algorithms require cross-validation to determine the best clusters. Furthermore, the covariance matrix may become singular. In order to solve the GMM, the invention uses a variational Bayesian method, the method can automatically determine the optimal cluster number without cross verification, and the convergence is good. For convenience, formula (I) is rewritten to formula (IV) without loss of generality:

the Dirichlet prior distribution of the parameter pi and the Gauss-Wishare prior distribution of the parameter { mu, T } are introduced; a more detailed description is as follows.

The Dirichlet a priori distribution is used for pi as shown in formula (V):

from the prior probability distribution of the parameters, the VB-GMM does not need to estimate any parameters, but rather relies on hidden random variables, i.e. h= { Z, pi, μ, T } using a variational method, an approximation q (h) of the VB-GMM is calculated, where Z is a hidden variable, expressed as the product of the form:

q(h)＝q(Z)·q(π)·q(μ,T)(VII)

q(π)＝Dir(π∣{α _i })(IX)

the method comprises the steps of respectively calculating the attribution relation of MPCs in a channel by using VB-GMM algorithm for unmanned aerial vehicle air-to-ground channel data at each moment, as shown in figure 2, specifically comprising the following steps:

step 5: calculating a log likelihood function;

the evolution process for acquiring the cluster center based on the MCD algorithm comprises the following steps:

for unmanned aerial vehicle air-to-ground communications, it is unique in the time-variability of the channel. Clusters often experience a kill phenomenon during flight. However, the number of clusters and the variation of cluster positions may have an influence on some indexes of the mobile communication system, such as channel capacity and delay spread. Therefore, it is necessary to describe the evolution process of the cluster. In previous methods, euclidean distance was typically used to measure the distance between the parts. However, for the channel dataset, the data units of the MPCs parameters are not consistent, and direct use of euclidean distance does not provide objective and accurate tracking results. In addition, since the cluster changes rapidly in the unmanned aerial vehicle air-to-ground communication, only considering the nearest two sampling points often leads to track interruption, and is not suitable for further analysis. Therefore, the invention adopts a tracking method based on multipath component distance MCD, and the calculation formula of the MCD is shown in formulas (XIII), (XIV) and (XV):

Wherein ζ is the weight of the delay distance, τ _std Is the standard deviation of MPCs delay, deltaτ _max Is the maximum value of MPCs delay difference; and judging whether the cluster center evolves or not based on the MCD distance. If the distance between the centers of two nearest clusters exceeds a threshold value, the cluster is considered to have a death phenomenon in the future; also, if the current cluster center cannot find the past cluster center, the cluster is considered to be new. ζ=3.

Acquiring channel characteristics based on clustering and evolution results, wherein the clustering and evolution results provide the attribution relation and the number of clusters of each MPCs; the channel characteristics include intra-cluster characteristics, inter-cluster characteristics, and evolution characteristics; the intra-cluster characteristics include: the number of MPCs in each cluster, the power attenuation factor, the intra-cluster rice K factor, the intra-cluster delay distribution and the intra-cluster angle distribution; the inter-cluster characteristics include the number of clusters, inter-cluster power and delay characteristics, and inter-cluster angle characteristics.

The number of MPCs within a cluster represents the abundance of multipath effects in the channel; the method for obtaining the number of MPCs in the cluster comprises the following steps: firstly, obtaining the number of MPCs at each moment according to a VB-GMM algorithm; then, calculating a cumulative distribution function for the MPCs at all times to obtain the number of MPCs in the cluster; and compared with normal distribution characteristics, the method finds that the distribution rules are basically consistent, and the fitting effect is good.

P _c (dB)＝-k·τ _c +b(XVI)

wherein k is a power attenuation factor, b is a power attenuation intercept, τ _c Delay for the c-th MPC in the cluster; through analysis of clustering results, the bilateral index model can well model the relation. The power attenuation factor k is found by equation (XVI).

The in-cluster rice K factor lambda characterizes the power relation between MPCs with strongest power in the cluster and the rest MPCs, and the reaction describes the power distribution relation of the MPCs in the cluster; the calculation formula of the intra-cluster rice K factor Λ is shown in formula (XVII):

and calculating the cumulative distribution function of the Less K factors in the clusters at all moments, and comparing and analyzing the cumulative distribution function with normal distribution characteristics to find that the fitting effect is better. Finally, the statistical characteristic of the rice K factor in the cluster can be found to basically meet the normal distribution rule.

The intra-cluster delay profile includes intra-cluster root mean square delay spreadAnd intra-cluster delay offset of two types of distribution relations; intra-cluster root mean square delay spreadThe calculation formula is shown as a formula (XVIII);

Calculating the intra-cluster root mean square delay spread of each moment according to formula (XVIII); and then the cumulative distribution function of the root mean square delay spread in the cluster at all moments is calculated, and compared with different probability distribution characteristics, the root mean square delay spread in the cluster is found to follow the lognormal distribution, and the fitting effect is good, namelyWherein mu _DS Sum sigma _DS Is thatMean and standard deviation of (a);

the intra-cluster delay offset is the difference between the intra-cluster sub-path delay and the cluster average delay, i.e., τ _off ＝τ _c -mean(τ _c ) The method comprises the steps of carrying out a first treatment on the surface of the And calculating the intra-cluster delay offset at each moment, and then calculating the cumulative distribution function of the intra-cluster delay offsets at all moments to find that the intra-cluster delay offset follows zero-mean Laplace distribution.

Similar to intra-cluster delay distribution, intra-cluster delay distribution comprises two types of distribution relations of intra-cluster root mean square angle expansion and intra-cluster angle offset; four types of intra-cluster root mean square angle expansion are provided, and the calculation formulas of the four types of intra-cluster root mean square angle expansion are shown as formula (XIX) and formula (XX);

/>

wherein,ε ^T/R the intra-cluster root mean square azimuth angle diffusion and intra-cluster root mean square pitch angle diffusion of the sending and arrival angles are respectively carried out. The characteristic of the intra-cluster root mean square delay spread is consistent, and the intra-cluster root mean square angle spread can also be fitted by adopting log normal distribution; the intra-cluster angle offset is the difference between the intra-cluster sub-path angle and the cluster average angle, i.e. satisfies +. >Andand the characteristic of the intra-cluster delay offset is consistent, and the intra-cluster delay offset is fitted by using zero-mean Laplace distribution.

The number of the clusters is obtained according to the clustering result, and the richness of the clusters in the unmanned aerial vehicle air-to-ground channel is reflected;

the method for calculating the inter-cluster power and time delay characteristics comprises the following steps: firstly, obtaining original inter-cluster power and original inter-cluster time delay according to different cluster labels obtained by a VB-GMM algorithm; comparing the original inter-cluster power and the original inter-cluster time delay with the time delay and the power of the LoS path to eliminate the calculated interference influence of the LoS paths at different positions and obtain the relative time delay and the relative power of each cluster at each moment; classifying the relative time delay and the relative power belonging to the same cluster at all moments; statistical analysis is performed on the relative time delay and the relative power of the same category, and the results are found to be basically consistent. Finally, calculating the average value of the relative time delay and the relative power to obtain the inter-cluster power and the time delay characteristic;

Analyzing evolution characteristics based on the cluster identification CLID; assigning a new CLID to a new cluster; the old cluster inherits the CLID of the last cluster; through the evolution rule of a four-state Markov chain reaction channel; through analysis of the CLID in the whole flight process, the evolution rule of a reaction channel with a good four-state Markov chain is found, and the characteristics of unmanned aerial vehicle air-to-ground channel evolution are met. FIG. 3 is a schematic diagram of a four-state Markov chain; the four states in the four-state Markov chain include: no generation of new clusters S0, generation of new clusters S1, death of old clusters S2, generation of new clusters and death of old clusters S3; the state transition probability matrix T of the four-state Markov chain is shown as a formula (XXI);

Establishing an unmanned aerial vehicle air-to-ground channel model based on channel characteristics, comprising:

based on the proposed comprehensive and rich channel characteristics, the invention provides a novel unmanned aerial vehicle air-to-ground communication channel model based on clusters. The LoS paths, scatterer clusters, and MPCs within the clusters are derived using the relevant channel parameters, as shown in fig. 7.

(1) Setting basic information, including environmental parameters, frequency bands, unmanned aerial vehicle flight tracks, and three-dimensional distances between Tx and Rx;

where K (t) is the number of clusters, C _i (t) is the number of MPCs in the ith cluster at time t, delta (. Cndot.) is the Dirac impulse function, and the power, phase and delay of the c-th ray in the ith cluster are P _i,c 、φ _i,c And τ _i,c The method comprises the steps of carrying out a first treatment on the surface of the Similarly, the relevant parameters of the LoS path are respectively set as P ₀ 、φ ₀ And τ ₀ 。

Example 3

The channel clustering and modeling method for the 6G unmanned aerial vehicle air-to-ground communication scene according to embodiment 1 or 2 is characterized in that:

acquiring an unmanned aerial vehicle air-to-ground communication full-band multi-dimensional channel data set, comprising:

reconstructing the air-to-ground communication scene of the unmanned aerial vehicle in the same proportion to obtain ray tracing simulation data; the method mainly adopts the same-proportion scene reconstruction by the Wireless institute software. By tracking all possible ray paths, MPCs between Rx and Tx can be determined, thus accurately describing the entire propagation process.

In the embodiment, in combination with the early-stage investigation, a software garden school area of Shandong university in Jinan, shandong province is selected as an experimental address; the specific implementation method is to obtain a map of a typical Shandong university campus scene, as shown in fig. 4 (b). In particular to the dimensions of all objects in the scene, the material electromagnetic parameters. The 3D pro-proportional modeling is performed in ray tracing software as shown in fig. 4 (a). And maximally restoring the original scene, and setting a plurality of parameters in the simulation scene, including various buildings, vegetation and the like. Topography and topography are also contemplated. According to the real scene, the tree height is set to 5 to 12 meters, the grassland height is set to 0.2 meters, and the building height is set to about 15 to 70 meters. According to the actual situation, the material of part of the positions is set as glass. The choice of material type and electromagnetic parameters for each scatterer is referred to the international telecommunications union database. The flight path of the drone, which spans many buildings, is labeled in fig. 4 (b). Radio wave propagation parameters in the simulation scene are set, including frequencies set to 28GHz and 3.8GHz, transmission power set to 10dBm, antenna type set to isotropic, maximum gain set to 0dBi for the antenna, antenna polarization set to vertical polarization, reception threshold set to-160 dBm, reflection diffraction and transmission coefficients set to 6/0/1.

the drone acts as Tx and the ground side acts as Rx. The flight distance of the unmanned aerial vehicle is 300 meters, the flight speed is 10 meters/second, the sampling distance interval of the whole flight process is 1 meter, and the sampling time interval is 0.1 second. The receiver detects signals from various propagation paths, including high-rise buildings, flat houses, floors, trees, vegetation. For each time instant, the channel characteristics include time delay τ, arrival pitch angle θ ^R Pitch angle θ ^T Azimuth of arrival phi ^R Azimuth angle phi of departure ^T And a received power P.

The clustering implementation process comprises the following steps: for the first moment, the prior parameters of the model are initialized. And entering a program circulation process, and firstly, calculating by using the current model parameter value to obtain a posterior probability value. And then the latest posterior probability value is obtained by applying the current calculation, and the model parameter value is estimated again, so that the model likelihood value is maximized. And calculating the log-likelihood function. It is determined whether the log-likelihood function has converged or the maximum number of iterations is reached. If yes, the cycle is ended, and a channel clustering result at the first moment is obtained; if not, returning to the calculation again. And for the next moment, completing a circulation step similar to the first moment, and obtaining a channel clustering result of the next moment. A schematic diagram of the channel clustering result is shown in fig. 5.

and calculating the cluster center according to the obtained clustering result. And judging the MCD between the centers of the clusters at different moments by adopting a tracking method based on the multipath component distance MCD. Wherein the ith cluster center m at m time _i And the j-th cluster center n at time n _j Multipath component distance betweenThe calculation formula of (2) is as follows:

wherein,and->The method comprises the following steps:

wherein ζ is the weight of the delay distance, τ _std Is the standard deviation of MPCs delay, deltaτ _max Is the maximum value of MPCs delay difference;

and matching the cluster closest to the m time and the n time according to the MCD distances between all clusters at the m time and all clusters at the n time. And calculating the time domain delay and angle domain difference between the nearest clusters of the two moments. Wherein the threshold of the time delay domain is set to 0.2us and the threshold of the angle domain is set to 0.5rad; if the threshold value is exceeded, the cluster is considered to have a death phenomenon; also, if the past cluster center cannot be found, the cluster is considered to be newly created. And calculating all time channel tracking results, and matching the relations among different time clusters.

Acquiring channel characteristics based on clustering and evolution results, including:

and analyzing the obtained clustering and evolution results, wherein the characteristics in the clusters, the characteristics among the clusters and the evolution characteristics are mainly analyzed.

The characteristics in the clusters comprise the number of MPCs in each cluster, a power attenuation factor, time delay in the clusters and angle distribution;

through statistical analysis of clustering results at each moment, it is found that under the condition that the receiving threshold values are the same, the path loss of MPCs in the millimeter wave clusters is larger, the receiving end is difficult to detect, and therefore fewer MPCs in the millimeter wave frequency band clusters are caused.

For the millimeter wave band, the relationship between the power attenuation factor k and the number of multipaths C within the cluster is expressed as:

k＝-0.72·exp(0.01·C)-9.05·exp(-0.12·C)

for the sub-6GHz band, the relationship between the power attenuation factor k and the number of multipaths C in the cluster is expressed as:

k＝-32.15·exp(-0.45·C)-3.62·exp(-0.03·C)

the intra-cluster delay profile includes intra-cluster root mean square delay spread and delay offset.

Through data analysis, the time delay of sub paths in the cluster is similar, and the time delay expansion in the cluster is not obvious. The root mean square delay spread in the cluster follows the lognormal distribution; compared with the millimeter wave frequency band, the intra-cluster time delay of the sub-6GHz frequency band is more dispersed. The intra-cluster delay bias specific effect is shown in fig. 6 (a) and 6 (b), which follow a zero-mean laplace distribution.

The intra-cluster angular distribution includes root mean square angular spread and angular offset. The root mean square angle diffusion in the cluster can be found to have a better fitting effect by using lognormal distribution. From the average of the root mean square angle spread, the arrival azimuth spread > the departure azimuth spread > the arrival pitch angle spread > the departure pitch angle spread are met. This is because the azimuth angles in the clusters come from different distributions of surrounding buildings, resulting in a dispersed azimuth angle distribution. Also, the pitch angle is mainly affected by the unmanned aerial vehicle height and ground Rx, resulting in an angle concentration. The angular offset conditions are consistent with the root mean square angular spread.

The inter-cluster characteristics comprise the number of clusters, inter-cluster power and time delay characteristics and angle distribution; according to the VB-GMM clustering result, the change relation between the clustering number of the air-to-ground communication of the unmanned aerial vehicle in two frequency bands and the MPCs number can be seen. Under the same threshold value, the number of MPCs and clusters of millimeter waves is relatively small; the maximum number of clusters in the sub-6GHz frequency band is 9, and the maximum number of clusters in the millimeter wave frequency band is 8; compared with the traditional ground channel model, the power and time delay characteristics among the unmanned aerial vehicle air-to-ground channel clusters are wider in time delay range, and the influence of a remote high building on propagation can be perceived. For the spatial distribution of clusters, the angular domain characteristics of the clusters of sub-6GHz and millimeter wave bands are found to be substantially similar. In the horizontal direction, the scatterer distribution is substantially uniform. The arrival azimuth and departure azimuth are thus distributed over the entire angular domain. In the vertical direction, the scatterers are distributed near the ground. The arrival pitch angle and the departure pitch angle are thus distributed over a range of angles.

For evolution characteristics, the time to live is analyzed using lognormal distribution and modeled using a four-state Markov chain. The survival time of the clusters is relatively short and the clusters are frequently extinguished in the whole unmanned aerial vehicle flight process.

and generating basic information of the clusters and the MPCs through the obtained channel characteristics. Firstly, setting a three-dimensional distance between Tx and Rx; generating time delay and power parameters of the clusters through corresponding random processes according to the characteristics among the clusters; according to the characteristics in the cluster, specific parameters of MPCs are obtained through corresponding random distribution, wherein the specific parameters comprise the number, time delay and power of the MPCs in the cluster; in the evolution process, different clusters are firstly endowed with different survival times, and the evolution of the channel state is carried out based on a Markov chain in a formula (XXI); judging whether birth and death occur or not according to the generated state; and combining all randomly generated clusters and corresponding intra-cluster MPCs thereof, and carrying into a formula (XXII) to obtain the channel impulse response CIR. Fig. 8 (a) and 8 (b) are schematic diagrams of power delay profiles of a cluster-based unmanned aerial vehicle air-to-ground channel model generated randomly based on the evolution process. Fig. 8 (a) is a schematic diagram of power delay profiles at two different moments in the sub-6GHz frequency band, and fig. 8 (b) is a schematic diagram of power delay profiles at two different moments in the millimeter wave frequency band.

Example 4

According to embodiment 3, the channel clustering and modeling method for the 6G unmanned aerial vehicle air-to-ground communication scene is characterized in that:

Acquiring unmanned aerial vehicle air-to-ground communication channel data, comprising:

unmanned aerial vehicle air-to-ground communication channel measurements based on the software radio USRP were developed. Based on the unmanned aerial vehicle side carried aerial platform, the detection signal is continuously sent. The selected detection signal is a pseudo random PN sequence. The unmanned aerial vehicle flies according to the planned track, and after the signal propagates through the specific channel environment, the ground receiving station receives and stores the detection signal in real time. And constructing a transmitting and receiving flow diagram of the detection signal at a transmitting end and a receiving end respectively through GNU Radio or Labview software, and setting corresponding measurement parameters to realize the transmitting and receiving of the detection signal. The transmitting end transmits the detection signal, and the receiving end extracts the multipath information of the channel. And acquiring the multipath information of the channel at each moment based on the actually measured data of the air-to-ground communication channel of the unmanned aerial vehicle. All time channel multipath information forms an unmanned aerial vehicle air-to-ground communication channel data set.

Claims

1. A6G unmanned aerial vehicle air-to-ground communication scene-oriented channel clustering and modeling method is characterized by comprising the following steps:

Acquiring an evolution process of a cluster center based on an MCD algorithm;

acquiring channel characteristics based on clustering and evolution results;

2. The method for clustering and modeling channels for a 6G unmanned aerial vehicle air-to-ground communication scenario of claim 1, wherein obtaining an unmanned aerial vehicle air-to-ground communication full-band multi-dimensional channel dataset comprises: acquiring a typical channel parameter, namely a full-band multi-dimensional channel data set by adopting a wireless channel measurement system or a ray tracing simulation method based on USRP of software radio equipment; typical channel parameters include time delay τ, arrival pitch angle θ ^R Pitch angle θ ^T Azimuth of arrival phi ^R Azimuth angle phi of departure ^T And a received power P; the full frequency band refers to sub-6GHz and millimeter wave frequency bands.

3. The method for clustering and modeling the channels of the 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the method for clustering the channels of the full-band multi-dimensional channel data set based on the VB-GMM algorithm comprises the following steps:

wherein T is _i Representing an accuracy matrix; VB-GMM is obtained by applying a priori distribution to the parameters { pi, μ, T };

the Dirichlet a priori distribution is used for pi as shown in formula (V):

wherein W (T) _i |v, V) refers to Wishart distribution; v represents the degree of freedom, V represents the scale matrix;

q(h)＝q(Z)·q(π)·q(μ,T) (VII)

q(π)＝Dir(π∣{α _i }) (IX)

4. The method for clustering and modeling the channels of the 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the method for clustering the channels of the full-band multi-dimensional channel data set based on the VB-GMM algorithm specifically comprises the following steps:

Step 5: calculating a log likelihood function;

step 6: judging whether the log likelihood function has converged or reaches the maximum iteration number; if not, returning to the step 3 to perform calculation again; if yes, a channel clustering result is obtained, and the cluster and the total number of clusters affiliated by each MPC are output.

5. The method for channel clustering and modeling for a 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the MCD algorithm-based acquisition of the evolution process of the cluster center comprises:

wherein ζ is the weight of the delay distance, τ _std Is the standard deviation of MPCs delay, deltaτ _max Is the maximum value of MPCs delay difference; judging whether the cluster center evolves or not based on the MCD distance; if the distance between the centers of two nearest clusters exceeds a threshold value, the cluster is considered to have a death phenomenon in the future; also, if the current cluster center cannot find the past cluster center, the cluster is considered to be new;

further preferably ζ=3.

6. The method for clustering and modeling the channels of the 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the channel characteristics are obtained based on clustering and evolution results, and the clustering and evolution results provide labels of each MPCs cluster and the number of clusters; the channel characteristics include intra-cluster characteristics, inter-cluster characteristics, and evolution characteristics; the intra-cluster characteristics include: the number of MPCs in each cluster, the power attenuation factor, the intra-cluster rice K factor, the intra-cluster delay distribution and the intra-cluster angle distribution; the inter-cluster characteristics include the number of clusters, inter-cluster power and delay characteristics, and inter-cluster angle characteristics.

7. The method for clustering and modeling channels for a 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the number of MPCs in the clusters represents the abundance of multipath effects in the channels; the method for obtaining the number of MPCs in the cluster comprises the following steps: firstly, obtaining the number of MPCs at each moment according to a VB-GMM algorithm; then, calculating a cumulative distribution function for the MPCs at all times to obtain the number of MPCs in the cluster;

P _c (dB)＝-k·τ _c +b (XVI)

wherein k is a power attenuation factor, b is a power attenuation intercept, τ _c Delay for the c-th MPC in the cluster; obtaining a power attenuation factor through a formula (XVI);

the intra-cluster delay distribution comprises two types of distribution relations of intra-cluster root mean square delay spread and intra-cluster delay offset; intra-cluster root mean square delay spread The calculation formula is shown as a formula (XVIII);

wherein,ε ^T/R the intra-cluster root mean square azimuth angle diffusion and intra-cluster root mean square pitch angle diffusion of the sending and arrival angles are respectively carried out.

8. The channel clustering and modeling method for the 6G unmanned aerial vehicle air-to-ground communication scene of claim 1, wherein the number of clusters is obtained according to a clustering result, and the richness of the clusters in the unmanned aerial vehicle air-to-ground channel is reflected;

the method for calculating the inter-cluster power and time delay characteristics comprises the following steps: firstly, obtaining original inter-cluster power and original inter-cluster time delay according to different cluster labels obtained by a VB-GMM algorithm; comparing the original inter-cluster power and the original inter-cluster time delay with the time delay and the power of the LoS path to eliminate the calculated interference influence of the LoS paths at different positions, and comparing the relative time delay and the relative power of each cluster at each moment; classifying the relative time delay and the relative power belonging to the same cluster at all moments; finally, calculating the average value of the relative time delay and the relative power, namely the inter-cluster power and the time delay characteristic;

9. The method for channel clustering and modeling for a 6G unmanned aerial vehicle air-to-ground communication scene according to claim 1, wherein the analysis evolution characteristic is performed based on a cluster identification CLID; assigning a new CLID to a new cluster; the old cluster inherits the CLID of the last cluster; through the evolution rule of a four-state Markov chain reaction channel; the four states in the four-state Markov chain include: no generation of new clusters S0, generation of new clusters S1, death of old clusters S2, generation of new clusters and death of old clusters S3; the state transition probability matrix T of the four-state Markov chain is shown as a formula (XXI);

10. The method for clustering and modeling channels for a 6G unmanned aerial vehicle air-to-ground communication scenario according to any one of claims 1-9, wherein the method for building an unmanned aerial vehicle air-to-ground channel model based on channel characteristics comprises:

where K (t) is the number of clusters, C _i (t) is the number of MPCs in the ith cluster at time t, delta (. Cndot.) is the Dirac impulse function, and the power, phase and delay of the c-th ray in the ith cluster are P _i,c 、φ _i,c And τ _i,c The method comprises the steps of carrying out a first treatment on the surface of the Similarly, the LoS path parameters are respectively set to P ₀ 、φ ₀ And τ ₀ 。