A method of environmental perception based on model evolution

Abstract

The present invention discloses an environmental perception method based on model evolution, which uses the channel response data obtained by periodically performing channel estimation tasks in the existing communication system to perceive the environment. First, the interaction mechanism between electromagnetic waves and environmental targets is divided into reflection and transmission, and then a mathematical model of channel response and environmental targets is constructed, and the environmental perception problem is modeled as a compressed sensing optimization problem. Finally, in order to solve the compressed sensing problem, a method is designed to start from the basic model and complete the iteration and evolution of the model by solving the target, and finally realize the perception of the environment. Compared with the existing environmental perception reconstruction method, the model considered by the present invention is closer to the real electromagnetic wave propagation characteristics, and has very little requirements for environmental prior information, and significantly improves the accuracy of environmental perception, which provides an effective method for realizing environmental perception in mobile communication systems.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to an environment perception method based on model evolution. Background Art

[0002] Integrated communication and perception is a key technology for the next generation of wireless communication systems. It requires that while completing normal communication, it can also perceive the physical world, including but not limited to user location, movement speed, human posture, etc. Integrated communication and perception requires the simultaneous realization of communication and perception functions in a system, realizing intensive use of resources, and the communication results and perception results can collaborate with each other to improve the overall performance of the system. It is a key enabling technology for emerging applications such as autonomous driving, non-destructive security inspection and smart cities.

[0003] Using channel response data (also called channel state information CSI) from conventional communication processes can achieve integrated communication and perception without making major changes to the hardware structure and software algorithms of existing communication systems. Therefore, it is an extremely important key technology. Among the many perception tasks and goals, the most important task is to image environmental targets, because imaging can obtain the largest amount of information about the target and the most original information, and various other tasks can be developed and implemented based on it, such as target recognition based on imaging results. At the same time, the imaging task is also the most difficult perception task, because it requires an efficient and accurate model and description of the interaction between the target object and the electromagnetic wave, and the design of an effective algorithm to achieve the imaging of the target.

[0004] The propagation rules of electromagnetic waves, especially the interaction between electromagnetic waves and obstacles, are very complex. When electromagnetic waves encounter obstacles, their energy will be divided into multiple parts. In simple terms, it can be divided into reflected energy, lost heat energy and transmitted energy: part of the energy is reflected back into space, part of the energy enters the object and propagates, part of which will be absorbed by the object and converted into heat energy, but part of the energy eventually passes through the target object, also known as transmission. When imaging the target, the environmental target is reversely solved based on the received channel state information. One of the main difficulties is: the mutual influence between environmental targets, such as the electromagnetic wave propagating to target B is very weak because it is blocked by target A in the middle, which is the coupling between targets A and B. Since there are many targets in the environment, they may have extremely complex coupling relationships, which leads to nonlinear characteristics of the problem, which greatly increases the difficulty of solving. The current environmental perception algorithms have not fully studied the coupling relationship between environmental targets, so they have poor performance in solving imaging problems under the real electromagnetic wave propagation characteristics.

[0005] In summary, considering the environmental perception problem and the real electromagnetic wave propagation model, the research on imaging of environmental targets that requires less prior information is of high difficulty and important practical significance. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art and taking into account the complex coupling relationship between environmental targets under the real electromagnetic wave propagation model, the present invention provides an environmental perception method based on model evolution. The environmental perception method based on model evolution obtains channel response data based on the periodic channel estimation task performed by the communication system to achieve environmental perception. Without making major modifications to the hardware or software of the existing communication system, the integration of perception and communication can be achieved, which solves the problem of how to achieve environmental perception at the base station end under the real electromagnetic wave propagation model.

[0007] The objective of the present invention is achieved through the following technical solutions:

[0008] A method for environmental perception based on model evolution is provided, the method comprising the following steps:

[0009] S1, using a base station to periodically receive pilot sequence signals sent by all active users within a certain space and perform channel estimation on the pilot sequence signals to obtain channel response data;

[0010] S2, using the electromagnetic wave propagation mechanism and the occlusion effect calculation method based on the position relationship, establishes the reflection and transmission channel model under the occlusion effect;

[0011] S3, discretizing the environment space, and combining the reflection and transmission channel models obtained in step S2, constructing a mathematical model of channel response and environmental target, and transforming the environmental perception problem into a generalized compressed sensing optimization problem through the mathematical model of channel response and environmental target;

[0012] S4, using the channel response data obtained in step S1, based on the mathematical model of the channel response and environmental target obtained in step S3, perform model evolution, and use the evolved model and the channel response data obtained in step S1 to solve the generalized compressed sensing optimization problem to complete the perception of the environmental target.

[0013] Furthermore, in step S1, while the base station is using the base station to perform channel estimation on the pilot sequence signal, the absolute position of the user terminal or the relative position of the user terminal to the base station is uploaded to the base station.

[0014] Furthermore, the step S2 is specifically as follows:

[0015] S21, using the electromagnetic wave propagation mechanism, in the mth r carrier frequency band, starting from the mth uThe user passes through the nth scatterer and reaches the mth base station k Ideal non-line-of-sight channel response for each antenna It can be expressed as:

[0016]

[0017] Among them, e is a natural constant, j is an imaginary unit, is the amplitude of the channel, is the phase of the channel;

[0018]

[0019]

[0020] in, It is the mth r The wavelength of the carrier wave, It is the mth u The distance from the user to the nth scatterer, is the mth base station k The distance from the root antenna to the nth scatterer, σ is the radar reflection cross section of the scatterer;

[0021] S22, the shielding effect is used to model the approximate straight-line propagation characteristics of high-frequency electromagnetic waves, and a shielding effect calculation method based on the position distribution of user terminals, base stations and scatterers is proposed; set point O A , O C is a general point, and point O B represents an obstacle, then the judgment point O B Is it right at point O? A and O C The main basis for judging whether the direct path between them is blocked is the following two points:

[0022] Point O B To point O A , O C The distance d of the connection is less than a certain threshold γ, that is:

[0023]

[0024] Point O B At point O A , O C Between, that is:

[0025]

[0026] When both of the above conditions are met, A From the perspective, O B O CApplying this calculation method to all targets, we get the occlusion indication matrix B:

[0027]

[0028] Where x is the reflection coefficient of the target, is the set of locations of all users and base stations, g is the calculation method function mentioned above;

[0029] S23, based on the ideal non-line-of-sight channel in step S21 The occlusion effect calculation method based on the position relationship in step S22 considers the reflection coefficient x of the environmental target, and the reflection channel model under the occlusion effect is It can be expressed as:

[0030]

[0031] in, is the occlusion indicator matrix of the reflection channel, x n is the reflection coefficient of the nth scatterer; considering the transmission coefficient y of the environmental target, the transmission channel model under the occlusion effect is It can be expressed as:

[0032]

[0033] in, is the occlusion indicator matrix of the transmission channel, y n is the transmission coefficient of the nth scatterer.

[0034] Furthermore, the step S3 is specifically as follows:

[0035] S31, discretize the environment space and divide the entire environment space into small cubes. The environment targets contained in each small cube can be equivalently concentrated at a point in the center of the small cube, which is also called point cloud division. Suppose the size of the environment space to be perceived is L L , L W , L H , the size of the small cube is l l , l w , l h , then there are N=L in the whole environment L / l l ×L W / l w ×L H / l h point cloud; use x n To represent the reflection coefficient of the nth point cloud, if the point cloud does not contain a scatterer, its reflection coefficient is 0. If the point cloud contains a scatterer, its reflection coefficient is a non-zero number. At the same time, use yn To represent the transmission coefficient of the nth point cloud; therefore, use the N-dimensional reflection coefficient vector x = [x1, ..., x N ] T And the N-dimensional transmission coefficient vector y = [y1, ..., y N ] T To represent the environmental space to be perceived;

[0036] S32, combining the reflection coefficient vector x and the transmission coefficient vector y obtained by discretizing the environmental space in step S31 with the reflection and transmission channel model obtained in step S2, to construct a mathematical model of the channel response and the environmental target as follows:

[0037] h=(H⊙B Re )x+(H⊙B Tr )y+v

[0038] =A Re x+A Tr y+v

[0039] Wherein, h is the channel response, x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, H is the ideal non-line-of-sight path matrix calculated in step S31, and B Re and B Tr is the occlusion effect indicator matrix calculated in step S32, v is random additive white Gaussian noise, ⊙ represents the Hadamard product operation, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel;

[0040] S33, based on the mathematical model of the channel response and the environmental target in step S32, the environmental perception problem is transformed into a generalized compressed sensing optimization problem, which is expressed as follows:

[0041]

[0042] st||hA Re xA Tr y||2≤ε

[0043] A Re =H⊙B Re

[0044] A Tr =H⊙B Tr

[0045]

[0046]

[0047] Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, ||·||1, ||·||2 are the first norm and second norm of the vector respectively. It means to find the minimum value of the corresponding expression in the value space of x and y, h is the channel response, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel, ε is the relaxation variable, H is the ideal non-line-of-sight path matrix calculated in step S21, and B Re and B Tr is the occlusion effect indication matrix calculated in step S22, g(·,·) is the occlusion effect calculation function based on the position relationship, It is the location of the base station and all user terminals.

[0048] Furthermore, the step S4 is specifically as follows:

[0049] S41, using a model evolution algorithm, the basic model represents the channel response data obtained in step S1, as follows:

[0050] h=Hx+v (1)

[0051] Where h is the channel response, x represents the reflection coefficient vector of the environment, H is the ideal non-line-of-sight path matrix, and v is the random additive white Gaussian noise. The simple model formula (1) is established as a compressed sensing optimization problem, which is expressed as follows:

[0052]

[0053] st||h-Hx||2≤ε

[0054] in It means to find the minimum value of the corresponding expression in the value space of x, and ε is the slack variable. To solve this problem, the maximum expectation-generalized approximate message passing algorithm is used.

[0055] S42, after obtaining the preliminary solution result of the environmental target x in step S41 Afterwards, the model can be evolved to obtain the estimated occlusion effect indicator matrix as follows:

[0056]

[0057] Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target can be evolved as follows:

[0058]

[0059] Where H is the ideal non-line-of-sight path matrix, v is random additive white Gaussian noise, and the model formula (2) is established as a compressed sensing optimization problem, which is expressed as follows:

[0060]

[0061]

[0062]

[0063]

[0064] Among them, x is the environmental target to be sought, is the estimation result of step S41, It means to find the minimum value of the corresponding expression in the value space of x, ε is the slack variable, h is the channel response data obtained in step S1, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals;

[0065] S43, for the measurement matrix The columns are normalized, and the compressed sensing optimization problem in step S42 is transformed into:

[0066]

[0067] st||hC Re x s ||2≤ε

[0068]

[0069]

[0070]

[0071] x s =diag(w)·x

[0072]

[0073] Among them, x is the environmental target to be sought, is the estimation result of step S41, represents finding the minimum value of the corresponding expression in the value space of x, h is the channel response obtained in step S1, C Reis the normalized effective measurement matrix, x s is the normalized ambient reflectance, ε is the relaxation variable, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals. The function diag(·) represents the diagonal matrix formed by taking vector elements. w represents The vector of column norms of Representation Matrix The nth column of represents a non-zero constant, and N is the number of point clouds in the environment space. After the above preprocessing, the maximum expectation-generalized approximate message passing algorithm is used to solve the problem.

[0074] S44, after obtaining the solution result of the environmental target x in step S43 Afterwards, the model can be evolved to obtain the estimated occlusion effect indicator matrix and as follows:

[0075]

[0076]

[0077] Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target can be evolved as follows:

[0078]

[0079] Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, Yes The calculated reflection channel measurement matrix, v is random additive white Gaussian noise;

[0080] By observing the model, it can be found that both the reflection coefficient x and the transmission coefficient y depend on the existence vector z of the scatterer in the environment, so a unified new model will be established

[0081] h=(H⊙B)z+v (3)

[0082] Among them, B represents the mask matrix of the scatterer existence variable, which is calculated as follows

[0083]

[0084] in and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively. The model formula (3) is modeled as an optimization problem:

[0085]

[0086] st||h-(H⊙B)z||2≤ε

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] Among them, z is the existence vector of the scatterer to be determined, x is the reflection coefficient vector to be determined, and y is the transmission coefficient vector to be determined. represents the search for the minimum value of the corresponding expression in the value space of z, h is the channel response obtained in step S1, H is the ideal non-line-of-sight path matrix, B represents the mask matrix of the scatterer existence variable, ε is the relaxation variable, and is the estimation result of step S44, is the position of the base station and all user terminals, g(·,·) is the occlusion effect calculation function based on the position relationship, and Yes and The calculated occlusion indicator matrix, and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively, and N is the number of point clouds in the ambient space;

[0093] S45, due to the complexity of the optimization problem in step S44, an alternating optimization method of iteratively solving the scatterer existence vector z and correcting the mask matrix B is adopted. The alternating optimization algorithm is described as follows:

[0094] When solving the scatterer existence vector z, the mask matrix B is fixed and the maximum expectation-generalized approximate message passing algorithm is used to solve z;

[0095] When modifying the mask matrix B, first fix the scatterer existence vector z and calculate the reflection coefficient and transmission coefficient Then calculate the occlusion effect matrix and B Tr , and merge into a new mask matrix B;

[0096] S46, alternately iterates and optimizes between the process of solving the scatterer existence vector z and correcting the mask matrix B. The algorithm ends under two conditions: reaching the maximum number of iterations T max Or the algorithm converges below the threshold τ, and the algorithm ends, completing the perception of the environmental target.

[0097] The beneficial effects of the present invention are as follows: first, based on the electromagnetic wave propagation mechanism, the present invention establishes a reflection and transmission channel model under the occlusion effect, and constructs a mathematical model of channel response and environmental targets, then transforms the environmental perception problem into a generalized compressed sensing optimization problem, and finally adopts a model evolution algorithm, starting from a simple basic model, obtains the necessary information of the advanced model by solving the target, realizes the iteration and evolution of the model, realizes the solution of the compressed sensing problem, and completes the perception of the environmental target. Through the strategy of model evolution, it achieves the characteristics of decoupling between environmental targets and very little demand for prior information, realizes the gradual reconstruction of the environment, and compared with the existing environmental perception reconstruction algorithm, the environmental perception algorithm based on model evolution of the present invention significantly improves the accuracy of environmental target perception, and provides an effective method for realizing environmental perception in a mobile communication system. Under the real electromagnetic wave propagation model, the present invention uses the model evolution strategy to realize environmental perception at the base station end, which has the advantages of very little demand for prior information and high accuracy of environmental perception. The present invention realizes environmental perception based on channel response data obtained by periodically performing channel estimation tasks on a communication system, and can realize perception and communication integration without making major modifications to the hardware or software of the existing communication system. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0099] Figure 1 A schematic diagram of an environment perception scenario based on electromagnetic wave propagation characteristics provided by an exemplary embodiment;

[0100] Figure 2 A schematic diagram of a method for calculating shading effect provided by an exemplary embodiment;

[0101] Figure 3A case diagram showing environmental target perception results comparing the algorithm of the present invention with other reconstruction algorithms provided as an exemplary embodiment;

[0102] Figure 4 A comparison diagram of the environment perception performance MSE of the algorithm of the present invention and the MSE of other reconstruction algorithms under different SNR conditions provided by an exemplary embodiment;

[0103] Figure 5 A diagram showing the relationship between the number of user terminals UE and the environment perception performance MSE when the algorithm of the present invention is compared with other reconstruction algorithms provided by an exemplary embodiment. DETAILED DESCRIPTION

[0104] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0105] It should be clear that the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0106] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0107] like Figure 1 As shown in the figure, the scenario we consider is an uplink communication scenario, where a multi-antenna base station access point (AP) is deployed in a certain area outdoors, and there are multiple active single-antenna user terminals (UEs). In this scenario, the user sends a signal to the AP, and it is assumed that the user will report his or her location information from time to time. The AP will perform periodic channel estimation tasks to overcome the random fading of the channel and obtain channel state information. This channel state information and the user's location will be transmitted to the central computing processor for centralized environmental perception.

[0108] The signals sent by users will be affected by environmental objects. For example, the signal sent by user 1 is reflected by target object 1 and reaches the AP, and the signal sent by user 2 penetrates target object 1 and reaches the AP. In addition, objects in the environment will also affect each other. From the perspective of user 3, target object 1 is blocked by the existence of target object 2.

[0109] The present invention provides an environment perception method based on model evolution, comprising the following steps:

[0110] S1, using a base station to periodically receive pilot sequence signals sent by all active users within a certain space and perform channel estimation on the pilot sequence signals to obtain channel response data;

[0111] In one embodiment, in step S1, while the base station is using the pilot sequence signal to perform channel estimation, the absolute position of the user terminal or the relative position of the user terminal to the base station is uploaded to the base station.

[0112] S2, using the electromagnetic wave propagation mechanism and the occlusion effect calculation method based on the position relationship, establishes the reflection and transmission channel model under the occlusion effect;

[0113] In one embodiment, step S2 is specifically as follows:

[0114] S21, using the electromagnetic wave propagation mechanism, in the mth r carrier frequency band, starting from the mth u The user passes through the nth scatterer and reaches the mth base station k Ideal non-line-of-sight channel response for each antenna It can be expressed as:

[0115]

[0116] Among them, e is a natural constant, j is an imaginary unit, is the amplitude of the channel, is the phase of the channel;

[0117]

[0118]

[0119] in, It is the mth r The wavelength of the carrier wave, It is the mth u The distance from the user to the nth scatterer, is the mth base station k The distance from the root antenna to the nth scatterer, σ is the radar reflection cross section of the scatterer;

[0120] S22, the shielding effect is used to model the approximate straight-line propagation characteristics of high-frequency electromagnetic waves, and a shielding effect calculation method based on the position distribution of user terminals, base stations and scatterers is proposed; Figure 2 As shown, set point O A , O C is a general point, and point O B represents an obstacle, then the judgment point O B Is it right at point O? A and O CThe main basis for judging whether the direct path between them is blocked is the following two points:

[0121] Point O B To point O A , O C The distance d of the connection is less than a certain threshold γ, that is:

[0122]

[0123] Point O B At point O A , O C Between, that is:

[0124]

[0125] When both of the above conditions are met, A From the perspective, O B O C Applying this calculation method to all targets, we get the occlusion indication matrix B:

[0126]

[0127] Where x is the reflection coefficient of the target, is the set of locations of all users and base stations, g is the calculation method function mentioned above;

[0128] S23, based on the ideal non-line-of-sight channel in step S21 The occlusion effect calculation method based on the position relationship in step S22 considers the reflection coefficient x of the environmental target, and the reflection channel model under the occlusion effect is It can be expressed as:

[0129]

[0130] in, is the occlusion indicator matrix of the reflection channel, x n is the reflection coefficient of the nth scatterer; considering the transmission coefficient y of the environmental target, the transmission channel model under the occlusion effect is It can be expressed as:

[0131]

[0132] in, is the occlusion indicator matrix of the transmission channel, y n is the transmission coefficient of the nth scatterer.

[0133] S3, discretizing the environment space, and combining the reflection and transmission channel models obtained in step S2, constructing a mathematical model of channel response and environmental target, and transforming the environmental perception problem into a generalized compressed sensing optimization problem through the mathematical model of channel response and environmental target;

[0134] In one embodiment, step S3 is specifically as follows:

[0135] S31, discretize the environment space and divide the entire environment space into small cubes. The environment targets contained in each small cube can be equivalently concentrated at a point in the center of the small cube, which is also called point cloud division. Suppose the size of the environment space to be perceived is L L , L W , L H , the size of the small cube is l l , l w , l h , then there are N=L in the whole environment L / l l ×L W / l w ×L H / l h point cloud; use x n To represent the reflection coefficient of the nth point cloud, if the point cloud does not contain a scatterer, its reflection coefficient is 0. If the point cloud contains a scatterer, its reflection coefficient is a non-zero number. At the same time, use y n To represent the transmission coefficient of the nth point cloud; therefore, use the N-dimensional reflection coefficient vector x = [x1, ..., x N ] T And the N-dimensional transmission coefficient vector y = [y1, ..., y N ] T To represent the environmental space to be perceived;

[0136] S32, combining the reflection coefficient vector x and the transmission coefficient vector y obtained by discretizing the environmental space in step S31 with the reflection and transmission channel model obtained in step S2, to construct a mathematical model of the channel response and the environmental target as follows:

[0137] h=(H⊙B Re )x+(H⊙B Tr )y+v

[0138] =A Re x+A Tr y+v

[0139] Wherein, h is the channel response, x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, H is the ideal non-line-of-sight path matrix calculated in step S31, and BRe and B Tr is the occlusion effect indicator matrix calculated in step S32, v is random additive white Gaussian noise, ⊙ represents the Hadamard product operation, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel;

[0140] S33, based on the mathematical model of the channel response and the environmental target in step S32, the environmental perception problem is transformed into a generalized compressed sensing optimization problem, which is expressed as follows:

[0141]

[0142] st||hA Re xA Tr y||2≤ε

[0143] A Re =H⊙B Re

[0144] A Tr =H⊙B Tr

[0145]

[0146]

[0147] Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, ||·||1, ||·||2 are the first norm and second norm of the vector respectively. It means to find the minimum value of the corresponding expression in the value space of x and y, h is the channel response, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel, ε is the relaxation variable, H is the ideal non-line-of-sight path matrix calculated in step S21, and B Re and B Tr is the occlusion effect indication matrix calculated in step S22, g(·,·) is the occlusion effect calculation function based on the position relationship, It is the location of the base station and all user terminals.

[0148] S4, using the channel response data obtained in step S1, based on the mathematical model of the channel response and environmental target obtained in step S3, perform model evolution, and use the evolved model and the channel response data obtained in step S1 to solve the generalized compressed sensing optimization problem to complete the perception of the environmental target.

[0149] In one embodiment, step S4 is specifically as follows:

[0150] S41, using a model evolution algorithm, the basic model represents the channel response data obtained in step S1, as follows:

[0151] h=Hx+v (1)

[0152] Where h is the channel response, x represents the reflection coefficient vector of the environment, H is the ideal non-line-of-sight path matrix, and v is the random additive white Gaussian noise. The simple model formula (1) is established as a compressed sensing optimization problem, which is expressed as follows:

[0153]

[0154] st||h-Hx||2≤ε

[0155] in It means to find the minimum value of the corresponding expression in the value space of x, and ε is the slack variable. To solve this problem, the Expectation-Maximum Generalized Approximate Message Passing (EM-GAMP) algorithm is used.

[0156] S42, after obtaining the preliminary solution result of the environmental target x in step S41 Afterwards, the model can be evolved to obtain the estimated occlusion effect indicator matrix as follows:

[0157]

[0158] Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target can be evolved as follows:

[0159]

[0160] Where H is the ideal non-line-of-sight path matrix, v is random additive white Gaussian noise, and the model formula (2) is established as a compressed sensing optimization problem, which is expressed as follows:

[0161]

[0162]

[0163]

[0164]

[0165] Among them, x is the environmental target to be sought, is the estimation result of step S41, It means to find the minimum value of the corresponding expression in the value space of x, ε is the slack variable, h is the channel response data obtained in step S1, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals;

[0166] S43, for the measurement matrix The columns are normalized, and the compressed sensing optimization problem in step S42 is transformed into:

[0167]

[0168] st||hC Re x s ||2≤ε

[0169]

[0170]

[0171]

[0172] x s =diag(w)·x

[0173]

[0174] Among them, x is the environmental target to be sought, is the estimation result of step S41, represents finding the minimum value of the corresponding expression in the value space of x, h is the channel response obtained in step S1, C Re is the normalized effective measurement matrix, x s is the normalized ambient reflectance, ε is the relaxation variable, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals. The function diag(·) represents the diagonal matrix formed by taking vector elements. w represents The vector of column norms of Representation Matrix The nth column of represents a non-zero constant, and N is the number of point clouds in the environment space. After the above preprocessing, the Expectation-Maximum Generalized Approximate Message Passing (EM-GAMP) algorithm is used to solve the problem.

[0175] S44, after obtaining the solution result of the environmental target x in step S43 Afterwards, the model can be evolved to obtain the estimated occlusion effect indicator matrix and as follows:

[0176]

[0177]

[0178] Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target can be evolved as follows:

[0179]

[0180] Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, Yes The calculated reflection channel measurement matrix, v is random additive white Gaussian noise;

[0181] By observing the model, it can be found that both the reflection coefficient x and the transmission coefficient y depend on the existence vector z of the scatterer in the environment, so a unified new model will be established

[0182] h=(H⊙B)z+v (3)

[0183] Among them, B represents the mask matrix of the scatterer existence variable, which is calculated as follows

[0184]

[0185] in and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively. The model formula (3) is modeled as an optimization problem:

[0186]

[0187] st||h-(H⊙B)z||2≤ε

[0188]

[0189]

[0190]

[0191]

[0192]

[0193] Among them, z is the existence vector of the scatterer to be determined, x is the reflection coefficient vector to be determined, and y is the transmission coefficient vector to be determined. represents the search for the minimum value of the corresponding expression in the value space of z, h is the channel response obtained in step S1, H is the ideal non-line-of-sight path matrix, B represents the mask matrix of the scatterer existence variable, ε is the relaxation variable, and is the estimation result of step S44, is the position of the base station and all user terminals, g(·,·) is the occlusion effect calculation function based on the position relationship, and Yes and The calculated occlusion indicator matrix, and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively, and N is the number of point clouds in the ambient space;

[0194] S45, due to the complexity of the optimization problem in step S44, an alternating optimization method of iteratively solving the scatterer existence vector z and correcting the mask matrix B is adopted. The alternating optimization algorithm is described as follows:

[0195] When solving the scatterer existence vector z, the mask matrix B is fixed and the expectation-maximization-generalized approximate message passing (EM-GAMP) algorithm is used to solve z;

[0196] When modifying the mask matrix B, first fix the scatterer existence vector z and calculate the reflection coefficient and transmission coefficient Then calculate the occlusion effect matrix and B Tr , and merge into a new mask matrix B;

[0197] S46, alternately iterates and optimizes between the process of solving the scatterer existence vector z and correcting the mask matrix B. The algorithm ends under two conditions: reaching the maximum number of iterations T max Or the algorithm converges below the threshold τ, and the algorithm ends, completing the perception of the environmental target.

[0198] Through computer simulation, we can see that: Figure 3As shown in the figure, the EM-GAMP algorithm completely ignores the occlusion effect and the mutual coupling relationship between the target objects, so the imaging effect is poor. The algorithm of the present invention is based on the model evolution method, and the model is iteratively corrected by calculation, thereby solving the problems such as the occlusion effect and the coupling relationship between the target objects. Compared with the EM-AGMP algorithm, the environment perception algorithm based on model evolution of the present invention significantly improves the accuracy of environment perception. Figure 4 This shows that the environmental perception performance of the algorithm of the present invention is significantly better than that of other algorithms, and the leading advantage becomes larger and larger with the increase of SNR. Figure 5 This indicates that the environmental perception performance of the method of the present invention improves as the number of users increases, and its advantage over other algorithms becomes increasingly greater.

[0199] In summary, under the real electromagnetic wave propagation model, the base station uses the strategy of model evolution to achieve environmental perception, which has the advantages of very little demand for prior information and high accuracy of environmental perception. The present invention is based on the communication system periodically performing channel estimation tasks to obtain channel response data to achieve environmental perception, and no major modifications are made to the hardware or software of the existing communication system, so as to achieve the integration of perception and communication. First, based on the electromagnetic wave propagation mechanism, the present invention establishes a reflection and transmission channel model under the occlusion effect, and constructs a mathematical model of channel response and environmental targets, and then transforms the environmental perception problem into a generalized compressed sensing optimization problem. Finally, the model evolution algorithm is adopted. Starting from a simple basic model, the necessary information of the advanced model is obtained by solving the target, and the iteration and evolution of the model are realized, which realizes the solution of the compressed sensing problem and completes the perception of the environmental target. Through the strategy of model evolution, it achieves the characteristics of decoupling between environmental targets and extremely low demand for prior information, thereby realizing the gradual reconstruction of the environment. Compared with the existing environmental perception reconstruction algorithm, the environmental perception algorithm based on model evolution of the present invention has significantly improved the accuracy of environmental target perception, and provides an effective method for realizing environmental perception in mobile communication systems.

[0200] The above description is merely a preferred embodiment of one or more embodiments of the present specification and is not intended to limit one or more embodiments of the present specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present specification shall be included in the scope of protection of one or more embodiments of the present specification.

Claims

1. A method for environmental perception based on model evolution, characterized in that: The following steps are involved: S1, using a base station to periodically receive pilot sequence signals sent by all active users within a certain space and perform channel estimation on the pilot sequence signals to obtain channel response data; S2, using the electromagnetic wave propagation mechanism and the shielding effect calculation method based on the position relationship, establish a reflection and transmission channel model under the shielding effect; step S2 is specifically: S21, using the electromagnetic wave propagation mechanism, in the mth r carrier frequency band, starting from the mth u The user passes through the nth scatterer and reaches the mth base station k Ideal non-line-of-sight channel response for each antenna Expressed as: Among them, e is a natural constant, j is an imaginary unit, is the amplitude of the channel, is the phase of the channel; in, It is the mth r The wavelength of the carrier wave, It is the mth u The distance from the user to the nth scatterer, is the mth base station k The distance from the root antenna to the nth scatterer, σ is the radar reflection cross section of the scatterer; S22, the occlusion effect is used to model the approximate straight-line propagation characteristics of high-frequency electromagnetic waves; set point O A ,O C is a general point, and point O B represents an obstacle, then the judgment point O B Is it right at point O? A and O C The main basis for judging whether the direct path between them is blocked is the following two points: Point O B To point O A ,O C The distance d of the connection is less than a certain threshold γ, that is: Point O B At point O A ,O C Between, that is: When both of the above conditions are met, A From the perspective, O B O C Applying this calculation method to all targets, we get the occlusion indication matrix B: Where x is the reflection coefficient of the target, is the set of locations of all users and base stations, g is the calculation method function mentioned above; S23, based on the ideal non-line-of-sight channel in step S21 The occlusion effect calculation method based on the position relationship in step S22 considers the reflection coefficient x of the environmental target, and the reflection channel model under the occlusion effect is Expressed as: in, is the occlusion indicator matrix of the reflection channel, x n is the reflection coefficient of the nth scatterer; considering the transmission coefficient y of the environmental target, the transmission channel model under the occlusion effect is Expressed as: in, is the occlusion indicator matrix of the transmission channel, y n is the transmission coefficient of the nth scatterer; S3, discretizing the environment space, and combining the reflection and transmission channel models obtained in step S2, constructing a mathematical model of channel response and environmental target, and transforming the environmental perception problem into a generalized compressed sensing optimization problem through the mathematical model of channel response and environmental target; S4, using the channel response data obtained in step S1, based on the mathematical model of the channel response and environmental target obtained in step S3, perform model evolution, and use the evolved model and the channel response data obtained in step S1 to solve the generalized compressed sensing optimization problem to complete the perception of the environmental target.

2. The environment perception method based on model evolution according to claim 1, characterized in that: In step S1, the base station is used to perform channel estimation on the pilot sequence signal, and at the same time, the absolute position of the user terminal or the relative position to the base station is uploaded to the base station.

3. The method according to claim 1, characterized in that The step S3 is specifically as follows: S31, discretize the environment space and divide the entire environment space into small cubes. The environment targets contained in each small cube are equivalently concentrated at a point in the center of the small cube, which is also called point cloud division. Suppose the size of the environment space to be perceived is L L ,L W ,L H , the size of the small cube is l l ,l w ,l h , then there are N=L in the whole environment L / l l ×L W / l w ×L H / l h point cloud; use x n To represent the reflection coefficient of the nth point cloud, if the point cloud does not contain a scatterer, its reflection coefficient is 0. If the point cloud contains a scatterer, its reflection coefficient is a non-zero number. At the same time, use y n To represent the transmission coefficient of the nth point cloud; therefore, the N-dimensional reflection coefficient vector x = [x1,…,x N ] T And the N-dimensional transmission coefficient vector y=[y1,…,y N ] T To represent the environmental space to be perceived; S32, combining the reflection coefficient vector x and the transmission coefficient vector y obtained by discretizing the environmental space in step S31 with the reflection and transmission channel model obtained in step S2, to construct a mathematical model of the channel response and the environmental target as follows: h=(H⊙B Re )x+(H⊙B Tr )y+v =A Re x+A Tr y+v Wherein, h is the channel response, x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, H is the ideal non-line-of-sight path matrix calculated in step S31, and B Re and B Tr is the occlusion effect indicator matrix calculated in step S32, v is random additive white Gaussian noise, ⊙ represents the Hadamard product operation, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel; S33, based on the mathematical model of the channel response and the environmental target in step S32, the environmental perception problem is transformed into a generalized compressed sensing optimization problem, which is expressed as follows: s.t.||h-A Re x-A Tr y||2≤ε A Re =H⊙B Re A Tr =H⊙B Tr Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, ||·||1,||·||2 are the first norm and second norm of the vector respectively. It means to find the minimum value of the corresponding expression in the value space of x and y, h is the channel response, A Re represents the measurement matrix of the reflection channel, A Tr represents the measurement matrix of the transmission channel, ε is the relaxation variable, H is the ideal non-line-of-sight path matrix calculated in step S21, and B Re and B Tr is the occlusion effect indication matrix calculated in step S22, g(·,·) is the occlusion effect calculation function based on the position relationship, It is the location of the base station and all user terminals.

4. The method according to claim 1, characterized in that: The step S4 is specifically as follows: S41, using a model evolution algorithm, the basic model represents the channel response data obtained in step S1, as follows: h=Hx+v (1) Where h is the channel response, x represents the reflection coefficient vector of the environment, H is the ideal non-line-of-sight path matrix, and v is the random additive white Gaussian noise. The simple model formula (1) is established as a compressed sensing optimization problem, which is expressed as follows: st||h-Hx||2≤ε in It means to find the minimum value of the corresponding expression in the value space of x, and ε is the slack variable. To solve this problem, the maximum expectation-generalized approximate message passing algorithm is used. S42, after obtaining the preliminary solution result of the environmental target x in step S41 Afterwards, the model is evolved to obtain the estimated occlusion effect indicator matrix as follows: Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target is evolved as follows: Where H is the ideal non-line-of-sight path matrix, v is random additive white Gaussian noise, and the model formula (2) is established as a compressed sensing optimization problem, which is expressed as follows: Among them, x is the environmental target to be sought, is the estimation result of step S41, It means to find the minimum value of the corresponding expression in the value space of x, ε is the slack variable, h is the channel response data obtained in step S1, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals; S43, for the measurement matrix The columns are normalized, and the compressed sensing optimization problem in step S42 is transformed into: s.t.||h-C Re x s ||2≤ε x s =diag(w)·x Among them, x is the environmental target to be sought, is the estimation result of step S41, represents finding the minimum value of the corresponding expression in the value space of x, h is the channel response obtained in step S1, C Re is the normalized effective measurement matrix, x s is the normalized ambient reflectance, ε is the relaxation variable, Yes The calculated reflection channel measurement matrix, H is the ideal non-line-of-sight path matrix, Yes The calculated occlusion indicator matrix, g(·,·), is a function for calculating the occlusion effect based on the position relationship. is the location of the base station and all user terminals. The function diag(·) represents the diagonal matrix formed by taking vector elements. w represents The vector of column norms of Representation Matrix The nth column of represents a non-zero constant, and N is the number of point clouds in the environment space. After the above preprocessing, the maximum expectation-generalized approximate message passing algorithm is used to solve the problem. S44, after obtaining the solution result of the environmental target x in step S43 Afterwards, the model is evolved to obtain the estimated occlusion effect indicator matrix and as follows: Among them, g(·,·) is the occlusion effect calculation function based on the position relationship, is the location of the base station and all user terminals. At this time, the mathematical model of channel response and environmental target is evolved as follows: Where x represents the reflection coefficient vector of the environment, y represents the transmission coefficient vector of the environment, Yes The calculated reflection channel measurement matrix, v is random additive white Gaussian noise; By observing the model, it is found that both the reflection coefficient x and the transmission coefficient y depend on the existence vector z of the scatterer in the environment, so a unified new model will be established: h=(H⊙B)z+v, (3) Among them, B represents the mask matrix of the scatterer existence variable, which is calculated as follows in and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively. The model formula (3) is modeled as an optimization problem: st||h-(H⊙B)z||2≤ε Among them, z is the existence vector of the scatterer to be determined, x is the reflection coefficient vector to be determined, and y is the transmission coefficient vector to be determined. represents the search for the minimum value of the corresponding expression in the value space of z, h is the channel response obtained in step S1, H is the ideal non-line-of-sight path matrix, B represents the mask matrix of the scatterer existence variable, ε is the relaxation variable, and is the estimation result of step S44, is the position of the base station and all user terminals, g(·,·) is the occlusion effect calculation function based on the position relationship, and Yes and The calculated occlusion indicator matrix, and are the means of the non-zero parts of the reflection coefficient x and the transmission coefficient y, respectively, and N is the number of point clouds in the ambient space; S45, due to the complexity of the optimization problem in step S44, an alternating optimization method of iteratively solving the scatterer existence vector z and correcting the mask matrix B is adopted. The alternating optimization algorithm is described as follows: When solving the scatterer existence vector z, the mask matrix B is fixed and the maximum expectation-generalized approximate message passing algorithm is used to solve z; When modifying the mask matrix B, first fix the scatterer existence vector z and calculate the reflection coefficient and transmission coefficient Then calculate the occlusion effect matrix and B Tr , and merge into a new mask matrix B; S46, alternately iterates and optimizes between the process of solving the scatterer existence vector z and correcting the mask matrix B. The algorithm ends under two conditions: reaching the maximum number of iterations T max Or the algorithm converges below the threshold τ, and the algorithm ends, completing the perception of the environmental target.

Images